Apply SanitizeFunctionName on request and RestoreSanitizedToolName on
response for: gemini/claude, gemini/openai/chat-completions,
gemini/openai/responses, antigravity/openai/chat-completions,
gemini-cli/openai/chat-completions.
Also update SanitizedToolNameMap to handle OpenAI format
(tools[].function.name) in addition to Claude format (tools[].name).
Claude Code and MCP clients may send tool names containing characters
invalid for Gemini's function_declarations (e.g. '/', '@', spaces).
Sanitize on request via SanitizeFunctionName and restore original names
on response for both antigravity/claude and gemini-cli/claude translators.
- Introduced a new method `buildRecord` in `usageReporter` to encapsulate record creation, improving code readability and maintainability.
- Added latency tracking to usage records, ensuring accurate reporting of request latencies.
- Updated tests to validate the inclusion of latency in usage records and ensure proper functionality of the new reporting structure.
When Claude sends redacted thinking with empty text, the translator
was omitting the "text" field from thought parts. Google Antigravity
API requires this field, causing 500 "Unknown Error" responses.
Verified: 129/129 error logs with empty thought → 500, 0/97 success
logs had empty thought. After fix: 0 new "Unknown Error" 500s.
- Log a warning instead of silently ignoring sjson.SetBytes errors in the TranslateRequest fallback path
- Add registry_test.go with tests covering the fallback model normalization and verifying registered transforms take precedence
When no request translator is registered for a format pair (e.g.
openai-response → openai-response), TranslateRequest returned the raw
payload unchanged. This caused client-side model prefixes (e.g.
"copilot/gpt-5-mini") to leak into upstream requests, resulting in
"The requested model is not supported" errors from providers.
The fallback path now updates the "model" field in the payload to
match the resolved model name before returning.
When enabled, the server uses only the embedded models.json loaded at
init() and skips registry.StartModelsUpdater(), disabling the initial
remote fetch and 3-hour periodic refresh. The management panel
auto-updater (managementasset.StartAutoUpdater) is unaffected.
gjson.Parse() marks any string starting with { or [ as gjson.JSON type,
even when the content is not valid JSON (e.g. macOS plist format, truncated
tool results). This caused sjson.SetRaw to embed non-JSON content directly
into the Gemini API request payload, producing 400 errors.
Add json.Valid() check before using SetRaw to ensure only actually valid
JSON is embedded raw. Non-JSON content now falls through to sjson.Set
which properly escapes it as a JSON string.
Fixes#2161
The Claude-to-Gemini translator derived function names by splitting
tool_use_id on "-", which produced empty strings for IDs with exactly
2 segments (e.g. toolu_tool-<uuid>). Replace the string-splitting
heuristic with a lookup map built from tool_use blocks during the
main processing loop, with fallback to the raw ID on miss.
Only emit note in listAuthFilesFromDisk when the JSON value is actually
a string (gjson.String), matching the synthesizer/buildAuthFileEntry
behavior. Non-string values like numbers or booleans are now ignored
instead of being coerced via gjson.String().
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
buildAuthFileEntry now falls back to reading priority/note from
auth.Metadata when Attributes lacks them. This covers auths registered
via UploadAuthFile which bypass the synthesizer and only populate
Metadata from the raw JSON.
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
- Read note from Attributes (consistent with priority) in buildAuthFileEntry,
fixing missing note on Gemini multi-project virtual auth cards.
- Propagate note from primary to virtual auths in SynthesizeGeminiVirtualAuths,
mirroring existing priority propagation.
- Sync note/priority writes to both Metadata and Attributes in PatchAuthFileFields,
with refactored nil-check to reduce duplication (review feedback).
- Validate priority type in fallback disk-read path instead of coercing all values
to 0 via gjson.Int(), aligning with the auth-manager code path.
- Add regression tests for note synthesis, virtual-auth note propagation, and
end-to-end multi-project Gemini note inheritance.
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
Gemini API rejects uniqueItems in tool schemas with 400. Add it to
unsupportedConstraints alongside minItems/maxItems where it belongs.
Same class of fix as #1424 and #1531.
When assistant has tool_calls but no text content, the translator
emitted an empty message into the Responses API input array before
function_call items. The API then couldn't match function_call_output
to its function_call by call_id, returning:
No tool output found for function call ...
Only emit assistant messages that have content parts. Tool-call-only
messages now produce function_call items directly.
Added 9 tests for tool calling translation covering single/parallel
calls, multi-turn conversations, name shortening, empty content
edge cases, and call_id integrity.
The list endpoint previously omitted priority and note, which are stored
inside each auth file's JSON content. This adds them to both the normal
(auth-manager) and fallback (disk-read) code paths, and extends
PATCH /auth-files/fields to support writing the note field.
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
1. Claude translator: add CleanJSONSchemaForGemini() to sanitize tool
input schemas (removes $schema, anyOf, const, format, etc.) and
delete eager_input_streaming from tool declarations. Remove fragile
bytes.Replace for format:"uri" now covered by schema cleaner.
2. Gemini native translator: filter out content entries with empty or
missing parts arrays to prevent Gemini API 400 error "required
oneof field 'data' must have one initialized field".
Both fixes align gemini-cli with protections already present in the
antigravity translator.
when Amp or Claude Code sends functionResponse with an empty name in Gemini
conversation history, the Gemini API rejects the request with 400
"Name cannot be empty". this fix backfills empty names from the
corresponding preceding functionCall parts using positional matching.
covers all three Gemini translator paths:
- gemini/gemini (direct API key)
- antigravity/gemini (OAuth)
- gemini-cli/gemini (Gemini CLI)
also switches fixCLIToolResponse pending group matching from LIFO to
FIFO to correctly handle multiple sequential tool call groups.
fixes#1903
feat(proxy): centralize proxy handling with `proxyutil` package and enhance test coverage
- Added `proxyutil` package to simplify proxy handling across the codebase.
- Refactored various components (`executor`, `cliproxy`, `auth`, etc.) to use `proxyutil` for consistent and reusable proxy logic.
- Introduced support for "direct" proxy mode to explicitly bypass all proxies.
- Updated tests to validate proxy behavior (e.g., `direct`, HTTP/HTTPS, and SOCKS5).
- Enhanced YAML configuration documentation for proxy options.
test(auth-scheduler): add benchmarks and priority-based scheduling improvements
- Added `BenchmarkManagerPickNextMixedPriority500` for mixed-priority performance assessment.
- Updated `pickNextMixed` to prioritize highest ready priority tiers.
- Introduced `highestReadyPriorityLocked` and `pickReadyAtPriorityLocked` for better scheduling logic.
- Added unit test to validate selection of highest priority tiers in mixed provider scenarios.
When new OAuth auth files are added while the service is running,
`applyCoreAuthAddOrUpdate` calls `coreManager.Register()` (which upserts
into the scheduler) BEFORE `registerModelsForAuth()`. At upsert time,
`buildScheduledAuthMeta` snapshots `supportedModelSetForAuth` from the
global model registry — but models haven't been registered yet, so the
set is empty. With an empty `supportedModelSet`, `supportsModel()`
always returns false and the new auth is never added to any model shard.
Additionally, when all existing accounts are in cooldown, the scheduler
returns `modelCooldownError`, but `shouldRetrySchedulerPick` only
handles `*Error` types — so the `syncScheduler` safety-net rebuild
never triggers and the new accounts remain invisible.
Fix:
1. Add `RefreshSchedulerEntry()` to re-upsert a single auth after its
models are registered, rebuilding `supportedModelSet` from the
now-populated registry.
2. Call it from `applyCoreAuthAddOrUpdate` after `registerModelsForAuth`.
3. Make `shouldRetrySchedulerPick` also match `*modelCooldownError` so
the full scheduler rebuild triggers when all credentials are cooling
down — catching any similar stale-snapshot edge cases.
The Gemini v1internal (cloudcode-pa) and Antigravity Manager endpoints
require camelCase "systemInstruction" in request JSON. The current
snake_case "system_instruction" causes system prompts to be silently
ignored when routing through these endpoints.
Replace all "system_instruction" JSON keys with "systemInstruction" in
chat-completions and responses request translators.
- Added comprehensive unit tests for `authScheduler` and related components.
- Implemented `authScheduler` with support for Round Robin, Fill First, and custom selector strategies.
- Improved tracking of auth states, cooldowns, and recovery logic in scheduler.
test(websocket): add tests for incremental input and prewarm handling logic
- Added test cases for incremental input support based on upstream capabilities.
- Introduced validation for prewarm handling of `response.create` messages locally.
- Enhanced test coverage for websocket executor behavior, including payload forwarding checks.
- Updated websocket implementation with prewarm and incremental input logic for better testability.
test(translator): add unit tests for OpenAI to Claude requests and tool result handling
- Introduced tests for converting OpenAI requests to Claude with text, base64 images, and URL images in tool results.
- Refactored `convertClaudeToolResultContent` and related functionality to properly handle raw content with images and text.
- Updated conversion logic to streamline image handling for both base64 and URL formats.
- Simplified connection logic by removing `connCreateSent` and related state handling.
- Updated `buildCodexWebsocketRequestBody` to always use `response.create`.
- Added unit tests to validate `response.create` behavior and beta header preservation.
- Dropped unsupported `response.append` and outdated `response.done` event types.
- Add model to GetGeminiModels()
- Add model to GetGeminiVertexModels()
- Add model to GetGeminiCLIModels()
- Add model to GetAIStudioModels()
- Add to AntigravityModelConfig with thinking levels
- Update gemini-3-flash-preview description
Registers the new lightweight Gemini model across all provider
endpoints for cost-effective high-volume usage scenarios.
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
SetOAuthSessionError previously sent generic messages to the management
panel (e.g. "Failed to complete Gemini CLI onboarding"), hiding the
actual error returned by Google APIs. The specific error was only
written to the server log via log.Errorf, which is often inaccessible
in headless/Docker deployments.
Include the upstream error in all 8 OAuth error paths so the
management panel shows actionable messages like "no Google Cloud
projects available for this account" instead of a generic failure.
normalizeCacheControlTTL unconditionally re-serializes the entire request
body through json.Unmarshal/json.Marshal even when no TTL normalization
is needed. Go's json.Marshal randomizes map key order and HTML-escapes
<, >, & characters (to \u003c, \u003e, \u0026), producing different raw
bytes on every call.
Anthropic's prompt caching uses byte-prefix matching, so any byte-level
difference causes a cache miss. This means the ~119K system prompt and
tools are re-processed on every request when routed through CPA.
The fix adds a bool return to normalizeTTLForBlock to indicate whether
it actually modified anything, and skips the marshal step in
normalizeCacheControlTTL when no blocks were changed.
Add util.SanitizeClaudeToolID() to replace non-conforming characters in
tool_use.id fields across all five response translators (gemini, codex,
openai, antigravity, gemini-cli).
Upstream tool names may contain dots or other special characters
(e.g. "fs.readFile") that violate Claude's ID validation regex.
The sanitizer replaces such characters with underscores and provides
a generated fallback for empty IDs.
Fixes#1872, Fixes#1849
Made-with: Cursor
- Add `TestTriggerServerUpdateCancelsPendingTimerOnImmediate` to verify proper handling of server update debounce and timer cancellation.
- Fix logic in `triggerServerUpdate` to prevent duplicate timers and ensure proper cleanup of pending state.
The instructions restore logic was originally needed when the proxy
injected custom instructions (per-model system prompts) into requests.
Since ac802a46 removed the injection system, the proxy no longer
modifies instructions before forwarding. The upstream response's
instructions field now matches the client's original value, making
the restore a no-op.
Also removes unused sjson import.
Closesrouter-for-me/CLIProxyAPI#1868
fix(gemini): add `deprecated` to unsupported schema keywords
Add `deprecated` to the list of unsupported schema metadata fields in Gemini and update tests to verify its removal.
- Set Accept-Encoding: identity for SSE streams; upstream must not compress
line-delimited SSE bodies that bufio.Scanner reads directly
- Re-enforce identity after ApplyCustomHeadersFromAttrs to prevent auth
attribute injection from re-enabling compression on the stream path
- Add peekableBody type wrapping bufio.Reader for non-consuming magic-byte
inspection of the first 4 bytes without affecting downstream readers
- Detect gzip (0x1f 0x8b) and zstd (0x28 0xb5 0x2f 0xfd) by magic bytes
when Content-Encoding header is absent, covering misbehaving upstreams
- Remove if-Content-Encoding guard on all three error paths (Execute,
ExecuteStream, CountTokens); unconditionally delegate to decodeResponseBody
so magic-byte detection applies consistently to all response paths
- Add 10 tests covering stream identity enforcement, compressed success bodies,
magic-byte detection without headers, error path decoding, and
auth attribute override prevention
Fixes#5 - The 'tui' flag was registered twice via flag.BoolVar(),
causing a 'panic: flag redefined: tui' crash on startup.
Removed the duplicate at line 208, keeping the original at line 181.
Claude's Tool Search feature (advanced-tool-use-2025-11-20 beta) adds
defer_loading field to tool definitions. When proxying Claude requests
to Codex or Gemini, this unknown field causes 400 errors upstream.
Strip defer_loading (and cache_control where missing) in all three
Claude-to-upstream translation paths:
- codex/claude: defer_loading + cache_control
- gemini-cli/claude: defer_loading
- gemini/claude: defer_loading
Fixes#1725, Fixes#1375
Add support for Claude's "adaptive" and "auto" thinking modes using `output_config.effort`. Introduce support for new effort level "max" in adaptive thinking. Update thinking logic, validate model capabilities, and extend converters and handling to ensure compatibility with adaptive modes. Adjust static model data with supported levels and refine handling across translators and executors.
## Problem
When using Antigravity Claude models through CLIProxyAPI, the thinking
chain (reasoning content) does not display in the Amp client.
## Root Cause
The Amp client sends `thinking: {"type": "auto"}` in its requests,
but `ConvertClaudeRequestToAntigravity` only handled `"enabled"` and
`"adaptive"` types in its switch statement. The `"auto"` type was
silently ignored, resulting in no `thinkingConfig` being set in the
translated Gemini request. Without `thinkingConfig`, the Antigravity
API returns responses without any thinking content.
Additionally, the Antigravity API for Claude models does not support
`thinkingBudget: -1` (auto mode sentinel). It requires a concrete
positive budget value. The fix uses 128000 as the budget for "auto"
mode, which `ApplyThinking` will then normalize to stay within the
model's actual limits (e.g., capped to `maxOutputTokens - 1`).
## Changes
### internal/translator/antigravity/claude/antigravity_claude_request.go
1. **Add "auto" case** to the thinking type switch statement.
Sets `thinkingBudget: 128000` and `includeThoughts: true`.
The budget is subsequently normalized by `ApplyThinking` based
on model-specific limits.
2. **Add "auto" to hasThinking check** so that interleaved thinking
hints are injected for tool-use scenarios when Amp sends
`thinking.type="auto"`.
### internal/registry/model_definitions_static_data.go
3. **Add Thinking configuration** for `claude-sonnet-4-6`,
`claude-sonnet-4-5`, and `claude-opus-4-6` in
`GetAntigravityModelConfig()` -- these were previously missing,
causing `ApplyThinking` to skip thinking config entirely.
## Testing
- Deployed to Railway test instance (cpa-thinking-test)
- Verified via debug logging that:
- Amp sends `thinking: {"type": "auto"}`
- CPA now translates this to `thinkingConfig: {thinkingBudget: 128000, includeThoughts: true}`
- `ApplyThinking` normalizes the budget to model-specific limits
- Antigravity API receives the correct thinkingConfig
Amp-Thread-ID: https://ampcode.com/threads/T-019ca511-710d-776d-a07c-4b750f871a93
Co-authored-by: Amp <amp@ampcode.com>
Captured and compared outgoing requests from CLIProxyAPI against real
Claude Code 2.1.63 and fixed all detectable differences:
Headers:
- Update anthropic-beta to match 2.1.63: replace fine-grained-tool-streaming
and prompt-caching-2024-07-31 with context-management-2025-06-27 and
prompt-caching-scope-2026-01-05
- Remove X-Stainless-Helper-Method header (real Claude Code does not send it)
- Update default User-Agent from "claude-cli/2.1.44 (external, sdk-cli)" to
"claude-cli/2.1.63 (external, cli)"
- Force Claude Code User-Agent for non-Claude clients to avoid leaking
real client identity (e.g. curl, OpenAI SDKs) during cloaking
Body:
- Inject x-anthropic-billing-header as system[0] (matches real format)
- Change system prompt identifier from "You are Claude Code..." to
"You are a Claude agent, built on Anthropic's Claude Agent SDK."
- Add cache_control with ttl:"1h" to match real request format
- Fix user_id format: user_[64hex]_account_[uuid]_session_[uuid]
(was missing account UUID)
- Disable tool name prefix (set claudeToolPrefix to empty string)
TLS:
- Switch utls fingerprint from HelloFirefox_Auto to HelloChrome_Auto
(closer to Node.js/OpenSSL used by real Claude Code)
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
Public Google OAuth client IDs and secrets from upstream CLIProxyAPI.
These are intentionally embedded (installed/CLI app OAuth flow) and
identical to the upstream public repo. Allowlisting prevents GitHub
push protection from blocking merges.
Guard the openai-response streaming path against truncated/invalid SSE data payloads by validating data: JSON before forwarding; surface a 502 terminal error instead of letting clients crash with JSON parse errors.
Move base64 image data from Claude tool_result into functionResponse.parts
as inlineData instead of outer sibling parts, preventing context bloat.
Unify all inlineData field naming to camelCase mimeType across Claude,
OpenAI, and Gemini translators. Add comprehensive edge case tests and
Gemini-side regression test for functionResponse.parts preservation.
CI blocks PRs that modify internal/translator. Revert translator edits and keep only the /v1/responses streaming error-chunk fix; file an issue for translator conformance work.
Ensure response.created and response.completed chunks produced by the OpenAI/Gemini/Claude translators always include required fields (response.model and response.usage) so clients validating Responses SSE do not fail schema validation.
When /v1/responses streaming fails after headers are sent, we now emit a type=error chunk instead of an HTTP-style {error:{...}} payload, preventing AI SDK chunk validation errors.
- Add 60 requests/minute rate limiting per credential using sliding window
- Detect insufficient_quota errors and set cooldown until next day (Beijing time)
- Map quota errors (HTTP 403/429) to 429 with retryAfter for conductor integration
- Cache Beijing timezone at package level to avoid repeated syscalls
- Add redactAuthID function to protect credentials in logs
- Extract wrapQwenError helper to consolidate error handling
- force http/1.1 instead of http/2
- explicit connection close
- strip proxy headers X-Forwarded-For and X-Real-IP
- add project id to fetch models payload
Changes the RoundRobinSelector to use two-level round-robin when
gemini-cli virtual auths are detected (via gemini_virtual_parent attr):
- Level 1: cycle across credential groups (parent accounts)
- Level 2: cycle within each group's project auths
Credentials start from a random offset (rand.IntN) for fair distribution.
Non-virtual auths and single-credential scenarios fall back to flat RR.
Adds 3 test cases covering multi-credential grouping, single-parent
fallback, and mixed virtual/non-virtual fallback.
Default to generating a fresh random user_id per request instead of
reusing cached IDs. Add cache-user-id config option to opt in to the
previous caching behavior.
- Add CacheUserID field to CloakConfig
- Extract user_id cache logic to dedicated file
- Generate fresh user_id by default, cache only when enabled
- Add tests for both paths
console.anthropic.com is now protected by a Cloudflare managed challenge
that blocks all non-browser POST requests to /v1/oauth/token, causing
`-claude-login` to fail with a 403 error.
Switch to api.anthropic.com which hosts the same OAuth token endpoint
without the Cloudflare managed challenge.
Fixes#1659
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
Non-spark codex models (gpt-5.3-codex, gpt-5.2-codex) stream function call
arguments via multiple delta events followed by a done event. The done handler
unconditionally emitted the full arguments, duplicating what deltas already
streamed. This produced invalid double JSON that Claude Code couldn't parse,
causing tool calls to fail with missing parameters and infinite retry loops.
Add HasReceivedArgumentsDelta flag to track whether delta events were received.
The done handler now only emits arguments when no deltas preceded it (spark
models), while delta-based streaming continues to work for non-spark models.
Some Codex models (e.g. gpt-5.3-codex-spark) send function call arguments
in a single "done" event without preceding "delta" events. The streaming
translator only handled "delta" events, causing tool call arguments to be
lost — resulting in empty tool inputs and infinite retry loops in clients
like Claude Code.
Emit the full arguments from the "done" event as a single input_json_delta
so downstream clients receive the complete tool input.
- Introduced `passthrough-headers` option in configuration to control forwarding of upstream response headers.
- Updated handlers to respect the passthrough headers setting.
- Added tests to verify behavior when passthrough is enabled or disabled.
sjson.SetRaw with an empty string produces malformed JSON (e.g. "result":}).
This happens when a Claude tool_result block has no content field, causing
functionResponseResult.Raw to be "". Guard against this by falling back to
sjson.Set with an empty string only when .Raw is empty.
- Implemented `--standalone` mode to launch an embedded server for TUI.
- Enhanced TUI client to support API-based log polling when log hooks are unavailable.
- Added authentication gate for password input and connection handling.
- Improved localization and UX for logs, authentication, and status bar rendering.
- Implemented `--standalone` mode to launch an embedded server for TUI.
- Enhanced TUI client to support API-based log polling when log hooks are unavailable.
- Added authentication gate for password input and connection handling.
- Improved localization and UX for logs, authentication, and status bar rendering.
- Introduced unit tests for request logging middleware to enhance coverage.
- Added WebSocket-based Codex executor to support Responses API upgrade.
- Updated middleware logic to selectively capture request bodies for memory efficiency.
- Enhanced Codex configuration handling with new WebSocket attributes.
The OpenAI Chat Completions translator was silently dropping
response.function_call_arguments.delta and
response.function_call_arguments.done Codex SSE events, meaning
tool call arguments were never streamed incrementally to clients.
Add proper handling mirroring the proven Claude translator pattern:
- response.output_item.added: announce tool call (id, name, empty args)
- response.function_call_arguments.delta: stream argument chunks
- response.function_call_arguments.done: emit full args if no deltas
- response.output_item.done: defensive fallback for backward compat
State tracking via HasReceivedArgumentsDelta and HasToolCallAnnounced
ensures no duplicate argument emission and correct behavior for models
like codex-spark that skip delta events entirely.
Update hardcoded X-Stainless-* and User-Agent defaults to match
Claude Code 2.1.44 / @anthropic-ai/sdk 0.74.0 (verified via
diagnostic proxy capture 2026-02-17).
Changes:
- X-Stainless-Os/Arch: dynamic via runtime.GOOS/GOARCH
- X-Stainless-Package-Version: 0.55.1 → 0.74.0
- X-Stainless-Timeout: 60 → 600
- User-Agent: claude-cli/1.0.83 (external, cli) → claude-cli/2.1.44 (external, sdk-cli)
Add claude-header-defaults config section so values can be updated
without recompilation when Claude Code releases new versions.
CPA-internal thinking levels like 'xhigh' and 'minimal' are not accepted
by OpenAI-format providers (MiniMax, etc.). The OpenAI applier now maps
non-standard levels to the nearest valid reasoning_effort value before
writing to the request body:
xhigh → high
minimal → low
auto → medium
Allow disabling the proxy_ tool name prefix on a per-account basis.
Users who route their own Anthropic account through CPA can set
"tool_prefix_disabled": true in their OAuth auth JSON to send tool
names unchanged to Anthropic.
Default behavior is fully preserved — prefix is applied unless
explicitly disabled.
Changes:
- Add ToolPrefixDisabled() accessor to Auth (reads metadata key
"tool_prefix_disabled" or "tool-prefix-disabled")
- Gate all 6 prefix apply/strip points with the new flag
- Add unit tests for the accessor
The proxy_ prefix logic correctly skips built-in tools (those with a
non-empty "type" field) in tools[] definitions but does not skip them
in messages[].content[] tool_use blocks or tool_choice. This causes
web_search in conversation history to become proxy_web_search, which
Anthropic does not recognize.
Fix: collect built-in tool names from tools[] into a set and also
maintain a hardcoded fallback set (web_search, code_execution,
text_editor, computer) for cases where the built-in tool appears in
history but not in the current request's tools[] array. Skip prefixing
in messages and tool_choice when name matches a built-in.
- Collect built-in tool names (those with a "type" field like
web_search, code_execution) and skip prefixing tool_reference
blocks that reference them, preventing name mismatch.
- Refactor if-else if chains to switch statements in all three
prefix functions for idiomatic Go style.
tool_reference blocks can appear nested inside tool_result.content[]
arrays, not just at the top level of messages[].content[]. The prefix
logic now iterates into tool_result blocks with array content to find
and prefix/strip nested tool_reference.tool_name fields.
applyClaudeToolPrefix, stripClaudeToolPrefixFromResponse, and
stripClaudeToolPrefixFromStreamLine now handle "tool_reference" blocks
(field "tool_name") in addition to "tool_use" blocks (field "name").
Without this fix, tool_reference blocks in conversation history retain
their original unprefixed names while tool definitions carry the proxy_
prefix, causing Anthropic API 400 errors: "Tool reference 'X' not found
in available tools."
Co-authored-by: Kirill Turanskiy <kt@novamedia.ru>
Upstream changes:
- feat: add Google One login and improve Gemini CLI auto-discovery
- feat: add gpt-5.3-codex-spark model definition
- feat: add 'kimi' channel support in model definitions
- feat: pass cache usage from codex to openai chat completions
- fix: normalize Gemini credential file prefix for consistency
- refactor: replace manual pointer assignments with new()
- chore: update Go to 1.26, update Codex client version
Fork fixes:
- Bump mise Go pin to 1.26.0
- Remove stale .syso resource files (incompatible with Go 1.26 linker)
Adds explicit handling for context.Canceled errors in the reverse proxy error handler to return 499 status code without logging, which is more appropriate for client-side cancellations during polling.
Also adds a test case to verify this behavior.
- Added unit tests for combining OAuth excluded models across global and attribute-specific scopes.
- Implemented priority attribute parsing with support for different formats and trimming.
Upstream changes (v6.8.0–v6.8.11):
- feat: add Kimi (Moonshot AI) provider support with OAuth
- feat: add iFlow provider with OAuth and cookie auth
- feat: support Claude thinking type adaptive
- feat: capture cached token count in usage metadata
- fix: correct max completion tokens for opus 4.6 thinking
- fix: improve content handling for system and user messages
- fix: rewrite model name in response.model for Responses API SSE
- refactor: simplify provider lifecycle and registration logic
- refactor: migrate config-access to internal package
Fork integration:
- Migrate configaccess from sdk/access/config to internal/access/config_access
- Update reconcile.go to use internalconfig types directly
- Add Kimi and iFlow login flags to ProxyPilot CLI
- Integrate managementasset auto-updater
- Keep all ProxyPilot fork features (TUI, agent setup, service, etc.)
Add Google One personal account login to Gemini CLI OAuth flow:
- CLI --login shows mode menu (Code Assist vs Google One)
- Web management API accepts project_id=GOOGLE_ONE sentinel
- Auto-discover project via onboardUser without cloudaicompanionProject when project is unresolved
Improve robustness of auto-discovery and token handling:
- Add context-aware auto-discovery polling (30s timeout, 2s interval)
- Distinguish network errors from project-selection-required errors
- Refresh expired access tokens in readAuthFile before project lookup
- Extend project_id auto-fill to gemini auth type (was antigravity-only)
Unify credential file naming to geminicli- prefix for both CLI and web.
Add extractAccessToken unit tests (9 cases).
Sanitize tool schemas by stripping prefill, enumTitles, $id, and patternProperties to prevent Gemini INVALID_ARGUMENT 400 errors, and add unit and executor-level tests to lock in the behavior.
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
- Added support for single and array-based `content` cases.
- Enhanced `system_instruction` structure population logic.
- Improved handling of user role assignment for string-based `content`.
Silences logging for client cancellations during polling to reduce noise in logs.
Client-side cancellations are common during long-running operations and should not be treated as errors.
The `ConvertOpenAIResponsesRequestToGemini` function had code that attempted
to uppercase JSON Schema type values (e.g. "string" -> "STRING") for Gemini
compatibility. This broke nullable types because when `type` is a JSON array
like `["string", "null"]`:
1. `gjson.Result.String()` returns the raw JSON text `["string","null"]`
2. `strings.ToUpper()` produces `["STRING","NULL"]`
3. `sjson.Set()` stores it as a JSON **string** `"[\"STRING\",\"NULL\"]"`
instead of a JSON array
4. The downstream `CleanJSONSchemaForGemini()` / `flattenTypeArrays()`
cannot detect it (since `IsArray()` returns false on a string)
5. Gemini/Antigravity API rejects it with:
`400 Invalid value at '...type' (Type), "["STRING","NULL"]"`
This was confirmed and tested with Droid Factory (Antigravity) Gemini models
where Claude Code sends tool schemas with nullable parameters.
The fix removes the uppercasing logic entirely and passes the raw schema
through to `parametersJsonSchema`. This is safe because:
- Antigravity executor already runs `CleanJSONSchemaForGemini()` which
properly handles type arrays, nullable fields, and all schema cleanup
- Gemini/Vertex executors use `parametersJsonSchema` which accepts raw
JSON Schema directly (no uppercasing needed)
- The uppercasing code also only iterated top-level properties, missing
nested schemas entirely
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
The ResponseRewriter's modelFieldPaths was missing 'response.model',
causing the mapped model name to leak through SSE streaming events
(response.created, response.in_progress, response.completed) in the
OpenAI Responses API (/v1/responses).
This caused Amp CLI to report 'Unknown OpenAI model' errors when
model mapping was active (e.g., gpt-5.2-codex -> gpt-5.3-codex),
because the mapped name reached Amp's backend via telemetry.
Also sorted modelFieldPaths alphabetically per review feedback
and added regression tests for all rewrite paths.
Fixes#1463
fix(translator): restructure message content handling to support multiple content types
- Consolidated `input_text` and `output_text` handling into a single case.
- Added support for processing `input_image` content with associated URLs.
- Updated `KimiAPIBaseURL` to remove versioning from the root path.
- Integrated `ClaudeExecutor` fallback in `KimiExecutor` methods for compatibility with Claude requests.
- Simplified token counting by delegating to `ClaudeExecutor`.
- Standardized the handling of `stop_reason` and `finish_reason` across Codex and Gemini responses.
- Restricted pass-through of specific reasons (`max_tokens`, `stop`) for consistency.
- Enhanced fallback logic for undefined reasons.
- Added support to extract and include `cachedContentTokenCount` in `usage.prompt_tokens_details`.
- Logged warnings for failures to set cached token count for better debugging.
- Introduced `RequestKimiToken` API for Kimi authentication flow.
- Integrated device ID management throughout Kimi-related components.
- Enhanced header management for Kimi API requests with device ID context.
Changes (v6.7.49 - v6.7.53):
- feat(registry): add Claude 4.6 (Opus) model
- feat(registry): add GPT 5.3 Codex model
- fix(registry): correct Claude Opus 4.6 model metadata
- fix: enable extended thinking support for Claude Haiku 4.5
- refactor(executor): centralize Codex client version and user agent
- refactor(translator, executor): remove redundant bytes.Clone calls
- refactor(util): optimize JSON schema processing
Upstream: https://github.com/router-for-me/CLIProxyAPI/releases/tag/v6.7.53
- OAuth2 device authorization grant flow (RFC 8628) for authentication
- Streaming and non-streaming chat completions via OpenAI-compatible API
- Models: kimi-k2, kimi-k2-thinking, kimi-k2.5
- CLI `--kimi-login` command for device flow auth
- Token management with automatic refresh
- Thinking/reasoning effort support for thinking-enabled models
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
- Replaced all instances of `bytes.Clone` with direct references to enhance efficiency.
- Simplified payload handling across executors and translators by eliminating unnecessary data duplication.
- Consolidated path-finding logic into a new `findPathsByFields` helper function.
- Refactored repetitive loop structures to improve readability and performance.
- Added depth-based sorting for deletion paths to ensure proper removal order.
Changes:
- feat(registry): add Kimi-K2.5 model to static data
- refactor(util, executor): optimize payload handling and schema processing
- fix(thinking): support user-defined includeThoughts setting (camelCase/snake_case)
- fix(thinking): ensure includeThoughts is false for ModeNone
- fix(gemini): support snake_case thinking config fields from Python SDK
- feat(gemini-cli): support image content in Claude request conversion
- fix(test): upstream fixed the builtin tools test we also fixed
Upstream: https://github.com/router-for-me/CLIProxyAPI/releases/tag/v6.7.48
Claude Haiku 4.5 (claude-haiku-4-5-20251001) supports extended thinking
according to Anthropic's official documentation:
https://docs.anthropic.com/en/docs/build-with-claude/extended-thinking
The model was incorrectly marked as not supporting thinking in the static
model definitions. This fix adds ThinkingSupport with the same parameters
as other Claude 4.5 models (Sonnet, Opus):
- Min: 1024 tokens
- Max: 128000 tokens
- ZeroAllowed: true
- DynamicAllowed: false
- Replaced repetitive string operations with a centralized `escapeGJSONPathKey` function.
- Streamlined handling of JSON schema cleaning for Gemini and Antigravity requests.
- Improved payload management by transitioning from byte slices to strings for processing.
- Removed unnecessary cloning of byte slices in several places.
Changes from v6.7.43-v6.7.47:
- fix(logging): expand tilde in auth-dir and log resolution errors
- feat(pprof): add configurable pprof HTTP debug server
- feat(watcher): log auth field changes on reload
- fix(cliproxy): update auth before model registration
- docs: README updates, AICodeMirror sponsorship
Fixed upstream test bug: TestOpenAIResponsesToOpenAI now correctly expects
built-in tools to be dropped (upstream intentionally changed this behavior).
Upstream: https://github.com/router-for-me/CLIProxyAPI/releases/tag/v6.7.47
Google official Gemini Python SDK sends thinking_level, thinking_budget,
and include_thoughts (snake_case) instead of thinkingLevel, thinkingBudget,
and includeThoughts (camelCase). This caused thinking configuration to be
ignored when using Python SDK.
Changes:
- Extract layer: extractGeminiConfig now reads snake_case as fallback
- Apply layer: Gemini/CLI/Antigravity appliers clean up snake_case fields
- Translator layer: Gemini->OpenAI/Claude/Codex translators support fallback
- Tests: Added 4 test cases for snake_case field coverage
Fixes#1426
- Add logic to handle `image` content type during request translation.
- Map Claude base64 image data to Gemini's `inlineData` structure.
- Support automatic extraction of `media_type` and `data` for image parts.
- Introduced a new `pprof` server to enable/debug HTTP profiling.
- Added configuration options for enabling/disabling and specifying the server address.
- Integrated pprof server lifecycle management with `Service`.
#1287
This merge brings in upstream changes from CLIProxyAPI v6.7.42 while
preserving ProxyPilot-specific customizations:
- Updated API server and executor modules
- New OAuth model alias functionality
- Enhanced thinking provider for iflow
- Updated model definitions and registry
- Improved docker build scripts
Conflicts were resolved by keeping ProxyPilot branding and merging
upstream improvements for translator and runtime components.
- Use util.ResolveAuthDir to properly expand ~ to user home directory
- Fixes issue where logs were created in literal "~/.cli-proxy-api" folder
- Add warning log when auth-dir resolution fails for debugging
Bug introduced in 62e2b67 (refactor(logging): centralize log directory
resolution logic), where strings.TrimSpace was used instead of
util.ResolveAuthDir to process auth-dir path.
MEDIUM: Return 400 on body read error instead of passing empty body
- Previously, on io.ReadAll error, body was consumed but c.Next()
was called, causing downstream handlers to see empty body
- Now returns http.StatusBadRequest on read errors
MEDIUM: Recompute token analysis after scaffold injection
- Token analysis was computed before scaffold state was injected
- If scaffold pushed request over token threshold, trimming was skipped
- Now recomputes analyzeTokenBudget() after injection
LOW: Always strip X-CLIProxyAPI-Todo header
- Header was only stripped when memory store was available
- If memory disabled, internal header leaked upstream
- Now stripped immediately regardless of memory store state
Co-Authored-By: Claude (gemini-claude-opus-4-5-thinking) <noreply@anthropic.com>
The trimming loop condition `for keep >= minKeep` was never true because
`keep` started at 20 while `minKeep` defaults to 25. Added clamp to ensure
`keep >= minKeep` at initialization, fixing prompt budget enforcement.
Affected functions:
- trimOpenAIChatCompletions
- trimOpenAIResponses
- trimOpenAIChatCompletionsWithMemory
- trimOpenAIResponsesWithMemory
- trimClaudeMessagesWithMemory
Co-Authored-By: Claude (gemini-claude-opus-4-5-thinking) <noreply@anthropic.com>
Split the 3,891 line kiro_executor.go into focused modules:
- kiro_executor.go (966 lines): Main executor, public interface
- kiro_auth.go (462 lines): Authentication/credential logic
- kiro_request.go (422 lines): Request transformation and retry logic
- kiro_response.go (1,765 lines): Response handling/streaming
Fix: Corrected isRetryableError to use errors.As for EventStreamError
type assertion instead of errors.Is on string constants.
Co-Authored-By: Claude (gemini-claude-opus-4-5-thinking) <noreply@anthropic.com>
Split the 2,280-line conductor.go into three focused files:
- conductor.go (1,416 lines): Core types, Manager struct, execution
- conductor_oauth.go (383 lines): OAuth/API key model alias resolution
- conductor_token.go (501 lines): Token refresh and HTTP request handling
This is a pure refactor with no logic changes.
Co-Authored-By: Claude (gemini-claude-opus-4-5-thinking) <noreply@anthropic.com>
Replace `for i, part := range` with `for _, part := range` since the
index variable was captured but never used (just discarded with `_ = i`).
Co-Authored-By: Claude (gemini-claude-opus-4-5-thinking) <noreply@anthropic.com>
Remove the V1 RouteModule interface which was marked as deprecated.
All modules now use RouteModuleV2 exclusively. The RegisterModule
function signature is simplified to accept RouteModuleV2 directly
instead of interface{}.
Co-Authored-By: Claude (gemini-claude-opus-4-5-thinking) <noreply@anthropic.com>
Remove TODO marker and update comment to indicate the stub is
intentional for API compatibility, not incomplete work.
Co-Authored-By: Claude (gemini-claude-opus-4-5-thinking) <noreply@anthropic.com>
Update FileVersion and ProductVersion in versioninfo.json files
from 0.1.6 to 0.3.0 for both proxypilot-tray and server commands.
Co-Authored-By: Claude (gemini-claude-opus-4-5-thinking) <noreply@anthropic.com>
Cherry-picked from upstream 6d8609e4
Allows configuring rules to remove specific JSON paths from request/response payloads.
Co-Authored-By: Claude (gemini-claude-opus-4-5-thinking) <noreply@anthropic.com>
Cherry-picked from upstream 7ff3936e
Adds multi-turn cache control tests.
Co-Authored-By: Claude (gemini-claude-opus-4-5-thinking) <noreply@anthropic.com>
Cherry-picked from upstream 6b6d030e
Adds uTLS-based HTTP transport to mimic browser TLS fingerprints for Claude API.
Co-Authored-By: Claude (gemini-claude-opus-4-5-thinking) <noreply@anthropic.com>
Cherry-picked from upstream ca796510
Uses post-order traversal and DeleteBytes for better performance.
Co-Authored-By: Claude (gemini-claude-opus-4-5-thinking) <noreply@anthropic.com>
Cherry-picked from upstream 2666708c
When Claude API sends an assistant message with empty text content,
the translator was creating an empty part, causing Gemini API to return
"required oneof field 'data' must have one initialized field" error.
Co-Authored-By: Claude (gemini-claude-opus-4-5-thinking) <noreply@anthropic.com>
Introduce `Filter` rules in the payload configuration to remove specified JSON paths from the payload. Update related helper functions and add examples to `config.example.yaml`.
- Add ErrorLogsMaxFiles config field with default value 10
- Support hot-reload via config file changes
- Add Management API: GET/PUT/PATCH /v0/management/error-logs-max-files
- Maintain SDK backward compatibility with NewFileRequestLogger (3 params)
- Add NewFileRequestLoggerWithOptions for custom error log retention
When request logging is disabled, forced error logs are retained up to
the configured limit. Set to 0 to disable cleanup.
Introduce a custom HTTP client utilizing utls with Firefox TLS fingerprinting to bypass Cloudflare fingerprinting on Anthropic domains. Includes support for proxy configuration and enhanced connection management for HTTP/2.
Tool outputs like "[1,2,3]\n\n[Process exited with code 0]" were being
partially parsed by gjson.Parse() which would extract just "[1,2,3]" and
lose the trailing text. This caused invalid JSON in the Gemini request.
Fix: Use gjson.Valid() to verify the ENTIRE string is valid JSON before
treating it as raw JSON. If validation fails, treat as a string with
proper escaping via sjson.Set().
Co-Authored-By: Claude (gemini-claude-opus-4-5-thinking) <noreply@anthropic.com>
Tests were failing with "auth_not_found: no auth available" because the
conductor.pickNextMixed() function checks ClientSupportsModel() against
the global registry. Test auths weren't registered with their models.
Changes:
- sdk/cliproxy/auth/conductor_test.go: Add registry.RegisterClient calls
- internal/translator/gemini/.../gemini_openai-responses_request.go:
Skip empty text parts to avoid invalid Gemini content
Note: Some pre-existing flaky tests remain in test/integration and
internal/translator/gemini (NonJSON tool output test).
Co-Authored-By: Claude (gemini-claude-opus-4-5-thinking) <noreply@anthropic.com>
- Add SupportedEndpoints field to ModelInfo
- Add GetMiniMaxModels, GetZhipuModels stub functions
- Add GetGitHubCopilotModels, GetKiroModels, GetAmazonQModels
- Update model registry with new providers
Cherry-picked and merged from plus remote.
Co-Authored-By: Claude (gemini-claude-opus-4-5-thinking) <noreply@anthropic.com>
When all oauth-model-alias entries are deleted, ensure an empty map
is persisted to config file rather than being omitted entirely.
Fixes#1305
Cherry-picked from upstream: a709e5a1
Co-Authored-By: Claude (gemini-claude-opus-4-5-thinking) <noreply@anthropic.com>
- Add firstChunkTimestamp field to ResponseWriterWrapper for sync capture
- Capture TTFB in Write() and WriteString() before async channel send
- Add SetFirstChunkTimestamp() to StreamingLogWriter interface
- Make requestTimestamp/apiResponseTimestamp required in LogRequest()
- Remove timestamp capture from WriteAPIResponse() (now via setter)
- Fix Gemini handler to set API_RESPONSE_TIMESTAMP before writing response
This ensures accurate TTFB measurement for all streaming API formats
(OpenAI, Gemini, Claude) by capturing timestamp synchronously when
the first response chunk arrives, not when the stream finalizes.
Cherry-picked from upstream: 295f34d7
Co-Authored-By: Claude (gemini-claude-opus-4-5-thinking) <noreply@anthropic.com>
Previously:
- REQUEST INFO timestamp was captured at log write time (not request arrival)
- API RESPONSE had NO timestamp at all
This fix:
- Captures REQUEST INFO timestamp when request first arrives
- Adds API RESPONSE timestamp when upstream response arrives
- Add Timestamp field to RequestInfo, set at middleware initialization
- Set API_RESPONSE_TIMESTAMP in appendAPIResponse() and gemini handler
- Pass timestamps through logging chain to writeNonStreamingLog()
- Add timestamp output to API RESPONSE section
Cherry-picked from upstream: c41ce77e
Co-Authored-By: Claude (gemini-claude-opus-4-5-thinking) <noreply@anthropic.com>
When using Gemini API format with Antigravity backend, the executor
renames usageMetadata to cpaUsageMetadata in non-terminal chunks.
The Gemini translator was returning this internal field name directly
to clients instead of the standard usageMetadata field.
Add restoreUsageMetadata() to rename cpaUsageMetadata back to
usageMetadata before returning responses to clients.
Cherry-picked from upstream: 9dc0e6d0
Co-Authored-By: Claude (gemini-claude-opus-4-5-thinking) <noreply@anthropic.com>
Add support for Gemini's code_execution and url_context tools in the
request translators, enabling:
- Agentic Vision: Image analysis with Python code execution for
bounding boxes, annotations, and visual reasoning
- URL Context: Live web page content fetching and analysis
Tools are passed through using the same pattern as google_search:
- code_execution: {} -> codeExecution: {}
- url_context: {} -> urlContext: {}
Tested with Gemini 3 Flash Preview agentic vision successfully.
Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>
Removes x-* extension fields from JSON schemas to ensure compatibility with the Gemini API.
These fields, while valid in OpenAPI/JSON Schema, are not recognized by the Gemini API and can cause issues.
The change recursively walks the schema, identifies these extension fields, and removes them, except when they define properties.
Amp-Thread-ID: https://ampcode.com/threads/T-019c0cd1-9e59-722b-83f0-e0582aba6914
Co-authored-by: Amp <amp@ampcode.com>
- Add ensureCacheControl() to auto-inject cache breakpoints
- Cache tools (last tool), system (last element), and messages (2nd-to-last user turn)
- Add prompt-caching-2024-07-31 beta header
- Return original payload on sjson error to prevent corruption
- Include verification test for caching logic
Enables up to 90% cost reduction on cached tokens.
Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>
- Add firstChunkTimestamp field to ResponseWriterWrapper for sync capture
- Capture TTFB in Write() and WriteString() before async channel send
- Add SetFirstChunkTimestamp() to StreamingLogWriter interface
- Make requestTimestamp/apiResponseTimestamp required in LogRequest()
- Remove timestamp capture from WriteAPIResponse() (now via setter)
- Fix Gemini handler to set API_RESPONSE_TIMESTAMP before writing response
This ensures accurate TTFB measurement for all streaming API formats
(OpenAI, Gemini, Claude) by capturing timestamp synchronously when
the first response chunk arrives, not when the stream finalizes.
Add migration logic for converting oauth-model-mappings to oauth-model-alias
with comprehensive test coverage.
Co-Authored-By: Claude (gemini-claude-opus-4-5-thinking) <noreply@anthropic.com>
Add empty auth provider for handling requests that don't require authentication.
Co-Authored-By: Claude (gemini-claude-opus-4-5-thinking) <noreply@anthropic.com>
Add complete Amp module for routing requests through the Amp CLI:
- Core module with lifecycle management
- Gemini bridge for model translation
- Model mapping configuration
- Proxy handler for request forwarding
- Response rewriter for format conversion
- Secret management for API keys
- Route definitions and fallback handlers
- Comprehensive test coverage
Co-Authored-By: Claude (gemini-claude-opus-4-5-thinking) <noreply@anthropic.com>
Previously:
- REQUEST INFO timestamp was captured at log write time (not request arrival)
- API RESPONSE had NO timestamp at all
This fix:
- Captures REQUEST INFO timestamp when request first arrives
- Adds API RESPONSE timestamp when upstream response arrives
Changes:
- Add Timestamp field to RequestInfo, set at middleware initialization
- Set API_RESPONSE_TIMESTAMP in appendAPIResponse() and gemini handler
- Pass timestamps through logging chain to writeNonStreamingLog()
- Add timestamp output to API RESPONSE section
This enables accurate measurement of backend response latency in error logs.
Cherry-picked changes from CLIProxyAPI v6.7.28:
- feat(handlers): add test to verify no retries after partial stream response (4eb1e60)
- fix(config): prune oauth-model-alias when preserving config (d0bada7)
Note: fix(api): update amp module only on config changes (8510fc3) was skipped
as ProxyPilot does not have the Amp feature.
Co-Authored-By: Claude (gemini-claude-opus-4-5-thinking) <noreply@anthropic.com>
Introduce `TestExecuteStreamWithAuthManager_DoesNotRetryAfterFirstByte` to validate that stream executions do not retry after receiving partial responses. Implement `payloadThenErrorStreamExecutor` for test coverage of this behavior.
When using Gemini API format with Antigravity backend, the executor
renames usageMetadata to cpaUsageMetadata in non-terminal chunks.
The Gemini translator was returning this internal field name directly
to clients instead of the standard usageMetadata field.
Add restoreUsageMetadata() to rename cpaUsageMetadata back to
usageMetadata before returning responses to clients.
When Claude API sends an assistant message with empty text content like:
{"role":"assistant","content":[{"type":"text","text":""}]}
The translator was creating a part object {} with no data field,
causing Gemini API to return error:
"required oneof field 'data' must have one initialized field"
This fix:
1. Skips empty text parts (text="") during translation
2. Skips entire messages when their parts array becomes empty
This ensures compatibility when clients send empty assistant messages
in their conversation history.
Optimize channel operations by introducing reusable context-aware send functions (`send` and `sendErr`) across `wsrelay`, `handlers`, and `cliproxy`. Ensure graceful handling of canceled contexts during stream operations.
Cherry-picked changes from CLIProxyAPI:
- fix(gemini): force type to string for enum fields (fixes Antigravity Gemini API error)
- feat(auth): add request_retry and disable_cooling per-auth overrides
- feat(auth): add skip persistence context key for file watcher events
Manually added RequestRetryOverride/DisableCoolingOverride methods to
types.go to preserve ProxyPilot's conductor.go compatibility while
gaining new functionality.
Implement `request_retry` and `disable_cooling` metadata overrides for authentication management. Update retry and cooling logic accordingly across `Manager`, Antigravity executor, and file synthesizer. Add tests to validate new behaviors.
- Antigravity retry logic with improved error handling and capacity management
- Added request_id to error logs for better debugging
- Export OAuth constants from auth packages (antigravity, claude, codex, gemini)
- Added Antigravity AuthService in internal/auth/antigravity
- Updated Claude executor test expectations for system message handling
- Skip built-in tools in OAuth tool prefix
- Preserve displayName and description in Gemini models list
- Added Embed stubs for ProviderExecutor interface compliance
Introduce `WithSkipPersist` to disable persistence during Manager Update/Register calls, preventing write-back loops caused by redundant file writes. Add corresponding tests and integrate with existing file store and conductor logic.
- Updated all executors with context cancellation handling
- Added payload rules using requested model
- Filter out Top_P when Temperature is set on Claude
- Remove empty system messages in translators
- Handle quota cooldown in retry logic
- Include requested model in executor metadata
- Normalize plan type filenames to lowercase
- Added Embed stubs for ProviderExecutor interface compliance
- Added payloadRequestedModel helper function
Fixes#1082
When all Antigravity accounts are unavailable, the error response now returns
HTTP 502 (Bad Gateway) instead of HTTP 400 (Bad Request). This ensures that
NewAPI and other clients will retry the request on a different channel,
improving overall reliability.
Previously GeminiModels handler unconditionally overwrote displayName
and description with the model name, losing the original values defined
in model definitions (e.g., 'Gemini 3 Pro Preview').
Now only set these fields as fallback when they are missing or empty.
Major changes from upstream v6.7.16:
- Remove sessionID from signature caching API entirely
- CacheSignature(modelName, text, sig) instead of 4 params
- GetCachedSignature(modelName, text) instead of 3 params
- HasValidSignature(modelName, sig) instead of 1 param
- Add Gemini fallback: returns "skip_thought_signature_validator" on cache miss
- Add sliding expiration (TTL refresh on cache access)
- Add GetModelGroup() for grouping cache keys by model family
- Simplify antigravity executor by removing project switch logic
- Fix OpenAI Responses request handler to skip empty assistant text
- Fix OpenAI Responses request to validate full JSON before parsing
Local fixes for compatibility:
- Add Embed stub method to AntigravityExecutor for interface compliance
- Update test to clear cache for isolation (ThinkingBlockWithoutSignature)
- Update Gemini test to expect thinking blocks to be annotated not removed
- Fix gjson.Valid check before treating tool output as raw JSON
Port features from upstream CLIProxyAPI:
v6.7.13:
- Raw JSON payload rules (DefaultRaw, OverrideRaw)
- OpenCode instructions support with user agent detection
v6.7.14:
- Improve signature handling by associating with model name
v6.7.15:
- Unify model group references with GetModelGroup helper function
Features ported from upstream:
- feat(claude): add native request cloaking support
- feat(translator): add candidate count (n param) support for Gemini
- fix(translator): consolidate multiple system messages into single message (#1064)
- feat(codex): add config toggle for codex instructions injection
- feat(auth): add hashed account ID to credential filenames for team plans
- feat(config): add support for raw JSON payload rules (default-raw, override-raw)
- feat(codex): add OpenCode instructions based on user agent
Also includes minor whitespace cleanup and code organization.
The checkSystemInstructions was being called on every request but the
cloaking/system prompt handling is now done elsewhere. This removes
redundant processing.
Cherry-picked from CLIProxyAPI: 7831cba9
Extract session ID from metadata.user_id field when available, using
the "session_" prefix marker. Falls back to first user message hash.
Cherry-picked from CLIProxyAPI: 6184c433 (Fixed: #1109)
Update tests to reflect the new signature handling behavior where only
cached signatures are used, not client-provided ones. This prevents
"Corrupted thought signature" errors when switching between Claude and
Gemini models.
Tests now properly cache signatures before testing conversion functions.
Add support for Imagen 3.0 and 4.0 image generation models in Vertex AI:
- Add 5 Imagen model definitions (4.0, 4.0-ultra, 4.0-fast, 3.0, 3.0-fast)
- Implement :predict action routing for Imagen models
- Convert Imagen request/response format to match Gemini structure
- Transform prompts to Imagen's instances/parameters format
- Convert base64 image responses to Gemini-compatible inline data
Cherry-picked from CLIProxyAPI: 8734d4cb
Add InfoByProvider map to ModelRegistration to store provider-specific
ModelInfo, enabling different capabilities for the same model across
providers (e.g., direct API vs OpenRouter).
Changes:
- Add InfoByProvider field to ModelRegistration struct
- Update GetModelInfo to accept provider parameter for provider-specific lookup
- Update LookupModelInfo to accept variadic provider parameter
- Add LookupStaticModelInfo for fallback to static model definitions
- Update ApplyThinking to pass providerKey for registry lookups
- Update all GetModelInfo callers to use new signature
- Remove ensureMaxTokensForThinking test (function moved to thinking package)
Ported from CLIProxyAPI e641fde2
Move ensureMaxTokensForThinking logic from claude_executor to
thinking/provider/claude/apply.go. This centralizes Claude-specific
thinking budget constraints in the thinking package.
Cherry-picked from c421d653.
Add 6th parameter 'original' to applyPayloadConfigWithRoot to support
checking defaults against the original payload. This enables proper
default rule validation when translating between formats.
Based on upstream 2a663d5c.
- Comment out client signature extraction (incompatible across model families)
- Use only cached signatures for thinking blocks
- Switch to thinking.GetThinkingText for text extraction
- Add explanatory comment about cross-model signature issues
Cherry-picked from d4bb4e66.
- Remove static Name mapping from AntigravityModelConfig
- Add new model entries: claude-sonnet-4-5, gpt-oss-120b-medium, tab_flash_lite_preview
- Add model skip list for unlisted models
- Update thinking limits (Max: 128000, ZeroAllowed: true)
Cherry-picked from c175821c.
Add support for Imagen 3.0 and 4.0 image generation models in Vertex AI:
- Add 5 Imagen model definitions (4.0, 4.0-ultra, 4.0-fast, 3.0, 3.0-fast)
- Implement :predict action routing for Imagen models
- Convert Imagen request/response format to match Gemini structure like gemini-3-pro-image
- Transform prompts to Imagen's instances/parameters format
- Convert base64 image responses to Gemini-compatible inline data
Add new PATCH /v0/management/auth-files/status endpoint that allows
toggling the disabled state of auth files without deleting them.
This enables users to temporarily disable credentials from the
management UI.
Rename include_thoughts to includeThoughts across all translator files
for consistency with the Gemini API field naming convention.
Cherry-picked key change from d5ef4a6d.
Port the internal/thinking package from CLIProxyAPI for unified thinking
configuration processing across all AI providers.
Features:
- Core types: ThinkingConfig, ThinkingMode, ThinkingLevel, SuffixResult
- Suffix parsing: model-name(16384) or model-name(high) syntax
- Validation: auto-conversion between budget/level formats, clamping
- Provider appliers: claude, gemini, antigravity, openai, codex
- Config extraction: extract thinking config from various request formats
- Stripping: remove thinking config for non-supporting models
Registry updates:
- Add UserDefined field to ModelInfo for passthrough behavior
- Add LookupModelInfo function for model capability lookup
- Add ResolveLogDirectory() for consistent log path resolution
- Add SetLogLevel() for string-based log level configuration
- Update ConfigureLogOutput() to take *config.Config parameter
- Update all callers to use simplified API
Cherry-picked from CLIProxyAPI with adaptations.
Gemini API requires enum values to be strings. This adds NormalizeEnumValues
to convert numeric or boolean enum values to their string representations.
Cherry-picked from CLIProxyAPI.
- Introduced `ResolveLogDirectory` function in `logging` package to standardize log directory determination across components.
- Replaced redundant logic in `server`, `global_logger`, and `handlers` with the new utility function.
Update ApplyThinking signature to accept fromFormat and toFormat parameters
instead of a single provider string. This enables:
- Proper level-to-budget conversion when source is level-based (openai/codex)
and target is budget-based (gemini/claude)
- Strict budget range validation when source and target formats match
- Level clamping to nearest supported level for cross-format requests
- Format alias resolution in SDK translator registry for codex/openai-response
Also adds ErrBudgetOutOfRange error code and improves iflow config extraction
to fall back to openai format when iflow-specific config is not present.
Refactor thinking configuration tests by separating model name suffix-based
scenarios from request body parameter-based scenarios into distinct test
functions with independent case numbering.
Architectural improvements:
- Extract thinkingTestCase struct to package level for shared usage
- Add getTestModels() helper returning complete model fixture set
- Introduce runThinkingTests() runner with protocol-specific field detection
- Register level-subset-model fixture with constrained low/high level support
- Extend iflow protocol handling for glm-test and minimax-test models
- Add same-protocol strict boundary validation cases (80-89)
- Replace error responses with clamped values for boundary-exceeding budgets
Gemini API requires all enum values in function declarations to be
strings. Some MCP tools (e.g., roxybrowser) define schemas with numeric
enums like `"enum": [0, 1, 2]`, causing INVALID_ARGUMENT errors.
Add convertEnumValuesToStrings() to automatically convert numeric and
boolean enum values to their string representations during schema
transformation.
- Add Auto-Updates and Rollback Support features to Features list
- Add Auto-Updates section with configuration options
- Add Rollback Support section explaining crash detection and recovery
- Document tray menu options for updates and rollback
Tray app integration:
- Health check on startup with auto-rollback trigger
- Background update polling with configurable interval
- One-click update menu item (shown when available)
- Manual rollback menu item in Settings
- Mark healthy after 30s of stable operation
- Graceful cleanup on exit (stop engine, stop poller)
New rollback module for version recovery:
- Version-aware backup metadata tracking
- Platform-specific restore (Windows batch, Unix swap)
- Crash loop detection via startup health monitoring
- Auto-rollback after 3 rapid failures in 30s window
- MarkHealthy() clears failure counter after stable run
Install module updated to save rollback info on update.
New UpdateBanner component for dashboard:
- Auto-checks for updates on mount with 2s delay
- Shows dismissible banner when update available
- Session-based dismissal persistence
- Download progress bar during update
- Integrated into App.tsx main layout
New poller module for periodic update checks:
- Configurable polling interval
- Channel support (stable/prerelease)
- Callback-based notification system
- Global singleton for tray integration
- Graceful start/stop with context cancellation
Cherry-pick from CLIProxyAPI v6.7.7 (f8f3ad8).
Fixes issue where only the last system/developer message was preserved
when multiple were provided. Added systemPartIndex counter to accumulate
parts across all system messages instead of overwriting at index 0.
Also fixed path inconsistency in gemini_openai_request.go where the
array branch incorrectly used request.systemInstruction instead of
system_instruction.
Fixes: router-for-me/CLIProxyAPI#1064
feat(translator): improve system message handling and content indexing across translators
- Updated logic for processing system messages in `claude`, `gemini`, `gemini-cli`, and `antigravity` translators.
- Introduced indexing for `systemInstruction.parts` to ensure proper ordering and handling of multi-part content.
- Added safeguards for accurate content transformation and serialization.
- Introduced `payloadModelAliases` and `payloadModelCandidates` functions to support model aliases for improved flexibility.
- Updated rule matching logic to handle multiple model candidates.
- Refactored variable naming in executor to improve code clarity and consistency.
- Enhanced ID matching in `cliproxy` by adding additional conditions to better handle ID equality cases.
- Updated `gemini` handlers to include `displayName` and `description` in normalized models for enriched metadata.
- Added conditional logic for Codex instruction injection based on configuration.
- Updated role terminology from "user" to "developer" for better alignment with context.
- Added logic to transform `inputResults` into structured JSON for improved processing.
- Removed redundant `safety_identifier` field in executor payload to streamline requests.
Add check to ensure 'instructions' field exists in Execute, ExecuteStream,
and CountTokens functions before sending request to Codex API.
Cherry-picked from:
- CLIProxyAPI v6.7.4 (902bea2)
Change conditional check from gemini-3-pro-preview (alias) to
gemini-3-pro-high (internal name) for system instruction injection.
Cherry-picked from:
- CLIProxyAPI v6.7.3 (c3ef46f)
- Introduced `default-raw` and `override-raw` rules to handle raw JSON values.
- Enhanced `PayloadConfig` to validate and sanitize raw JSON payload rules.
- Updated executor logic to apply `default-raw` and `override-raw` rules.
- Extended example YAML to include usage of raw JSON rules.
This change removes the translation logic for several non-standard, proprietary extensions used to configure thinking/reasoning. Specifically, support for `extra_body.google.thinking_config` and the Anthropic-style `thinking` object has been dropped from the OpenAI request translators.
This simplification streamlines the translators, focusing them on the standard `reasoning_effort` parameter. It also removes the need to look up model information from the registry within these components.
BREAKING CHANGE: Support for non-standard thinking configurations via `extra_body.google.thinking_config` and the Anthropic-style `thinking` object has been removed. Clients should now use the standard `reasoning_effort` parameter to control reasoning.
- Added `metadataEqualIgnoringTimestamps` to compare metadata while ignoring volatile fields.
- Prevented redundant writes caused by changes in timestamp-related fields.
- Improved efficiency in filestore operations by skipping unnecessary updates.
feat(auth): fetch and update Antigravity project ID from metadata during filestore operations
- Added support to retrieve and update `project_id` using the access token if missing in metadata.
- Integrated HTTP client to fetch project ID dynamically.
- Enhanced metadata persistence logic.
- Updated `ideType` to `ANTIGRAVITY` in request payload.
- Introduced tier-selection logic to determine default tier for onboarding.
- Added `antigravityOnboardUser` function for project ID retrieval via polling.
- Enhanced error handling and response decoding for onboarding flow.
This change introduces environment variable interpolation for volume paths, allowing users to customize where configuration, authentication, and log data are stored.
Why: Makes the project easier to deploy on various hosting environments that require decoupled data management without needing to modify the core docker-compose.yml..
Key points:
Defaults to existing paths (./config.yaml, ./auths, ./logs) to ensure zero breaking changes for current users.
Follows the existing naming convention used in the project.
Enhances portability for CI/CD and cloud-native deployments.
feat(translator): add function name to response output item serialization
- Included `item.name` in the serialized response output to enhance output item handling.
feat(oauth): add support for customizable OAuth callback ports
- Introduced `oauth-callback-port` flag to override default callback ports.
- Updated SDK and login flows for `iflow`, `gemini`, `antigravity`, `codex`, `claude`, and `openai` to respect configurable callback ports.
- Refactored internal OAuth servers to dynamically assign ports based on the provided options.
- Revised tests and documentation to reflect the new flag and behavior.
When switching from Claude models (e.g., Opus 4.5) to Gemini models
(e.g., Flash) mid-conversation via Antigravity OAuth, the client-provided
thinking signatures from Claude would cause "Corrupted thought signature"
errors since they are incompatible with Gemini API.
Changes:
- Remove fallback to client-provided signatures in thinking block handling
- Only use cached signatures (from same-session Gemini responses)
- Skip thinking blocks without valid cached signatures
- tool_use blocks continue to use skip_thought_signature_validator when
no valid signature is available
This ensures cross-model switching works correctly while preserving
signature validation for same-model conversations.
Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>
- Updated `SDKConfig` to use `nonstream-keepalive-interval` (seconds) instead of the boolean `nonstream-keepalive`.
- Refactored handlers and logic to incorporate the new interval-based configuration.
- Updated config diff, tests, and example YAML to reflect the changes.
- Introduced `StartNonStreamingKeepAlive` to emit periodic blank lines during non-streaming responses.
- Added `nonstream-keepalive` configuration option in `SDKConfig`.
- Updated handlers to utilize `StartNonStreamingKeepAlive` and ensure proper cleanup.
- Extended config diff and tests to include `nonstream-keepalive` changes.
feat(cliproxy): support multiple aliases for OAuth model mappings
- Updated mapping logic to allow multiple aliases per upstream model name.
- Adjusted `SanitizeOAuthModelMappings` to ensure aliases remain unique within channels.
- Added test cases to validate multi-alias scenarios.
- Updated example config to clarify multi-alias support.
- Add Claude 4.5 model variants (opus, sonnet, haiku)
- Add Gemini 2.0/2.5 Flash and Pro pricing
- Add OpenAI o3/o4-mini pricing
- Update date reference from Dec 2024 to Jan 2026
- GET /v0/management/rate-limits - detailed per-credential status
- GET /v0/management/rate-limits/summary - availability overview
- Shows available, cooling_down, disabled counts and next recovery time
- Bubbletea-based TUI with 5 tabs: Status, Accounts, Rate Limits, Usage, Logs
- Industrial control room aesthetic with orange/green color scheme
- Double-line borders (╔═╗), active tab indicator 【】
- 3-column stat boxes on Status, 4-column on Usage
- Visual gauge bar for rate limit availability
- Config-optional mode allows TUI to run without config.yaml
- Full-width frame with properly aligned borders
New Accounts submenu:
- Copy Account List (-list-accounts)
- Cleanup Expired (-cleanup-expired)
- Export Accounts... (-export-accounts)
- Import Accounts... (-import-accounts)
All menu items harness CLI binary for single source of truth.
Add menu items to copy CLI output to clipboard:
- Copy Account Status (account health summary)
- Copy Usage Stats (token usage by provider)
- Copy Model List (available models)
- Copy Recent Logs (last 100 log entries)
Uses stdout capture to get formatted CLI output.
- Add -usage flag (token usage stats per provider)
- Add -logs flag (view recent logs, -n for line count)
- Add -refresh flag (force token refresh)
- Add -list-models flag (show available models)
- Add -export-accounts/-import-accounts (backup/restore)
- Add --include-tokens and --force flags for export/import
- install.ts: read version from package.json (auto-sync)
- release.yml: manual dispatch only (no auto-trigger on tags)
- proxypilot-release.yml: add checksums.txt generation
- Add 26 tests for SetLogLevel() in logging package
- Add 6 tests for AppError in errors package
- Add 10 tests for SDKConfig in config package
- Add 3 tests for EmailRequiredError in sdk/auth
- Remove dead code: GeminiCLIToken type alias, LoadGeminiCLIToken,
LoadGeminiCLITokenFromPath (no callers found)
- Add tests for gemini auth (CredentialFileName, SaveTokenToFile)
Cherry-picked from e8e3bc86. Adds HttpRequest method to ProviderExecutor
interface for injecting credentials into arbitrary HTTP requests.
- Added HttpRequest to conductor.go interface
- Added HttpRequest stubs to kiro, minimax, zhipu executors
- Added http-request example
- Updated mock executors in tests
Also kept our Embed method and CountTokens implementations.
- feat(executor): enhance Antigravity payload with user role and dynamic system instructions
- feat(executor): add model-specific support for "gemini-3-pro" in execution
- feat(executor): update system instruction handling for Claude and Gemini-3-Pro models
- fix(executor): update gemini model identifier to gemini-3-pro-preview
- feat(executor): centralize systemInstruction handling
Cherry-picked: 67985d82, bda04eed, 1b2f9076, b6a0f7a0, 59a448b6
- fix(logging): improve request/response capture
- fix(handlers): match raw error text before JSON body for duplicate detection
- feat(logging): disambiguate OAuth credential selection in debug logs
- Various logging refinements for auth selection
Cherry-picked from upstream: a091d12f, 3ca5fb10, 0e4148b2, 79fbcb3e, 48f6d7ab, 48f19aab, cb3bdffb, a95428f2
- Add internal/auth/antigravity package with token loading from ~/.gemini/oauth_creds.json
- Add ImportFromGeminiCLI method to AntigravityAuthenticator
- Add DoAntigravityImport CLI command
- Add --antigravity-import flag to server binary
Usage: proxypilot --antigravity-import
Keep the upstream payload prefixed for OAuth while passing the unprefixed request body into response translators. This avoids proxy_ leaking into OpenAI Responses echoed tool metadata while preserving the Claude OAuth workaround.
Prefix tool names with proxy_ for Claude OAuth requests and strip the prefix from streaming and non-streaming responses to restore client-facing names.
Updates the Claude executor to:
- add prefixing for tools, tool_choice, and tool_use messages when using OAuth tokens
- strip the prefix from tool_use events in SSE and non-streaming payloads
- add focused unit tests for prefix/strip helpers
Update the model name check in `buildRequest` to target "gemini-3-pro-preview" instead of "gemini-3-pro" when applying specific system instruction handling.
Routes requests to the account with lowest daily output tokens.
Spreads load across accounts to avoid hitting quota limits.
Config: routing.strategy = "usage-aware" or "least-used"
Initialize response/prompt caches in Service.Run from config.
Add cache lookup to executor payload helpers.
Document cache config and management API in README.
ProviderLogins table now shows:
- Priority column with up/down reorder buttons
- Expires column with color-coded status (green/yellow/red)
API exposes token_expires_at for WebUI consumption.
When streaming responses with tool calls, the finish_reason was being
overwritten. The upstream sends functionCall in chunk 1, then
finishReason: STOP in chunk 2. The old code would set finish_reason
from every chunk, causing "tool_calls" to be overwritten by "stop".
This broke clients like Claude Code that rely on finish_reason to
detect when tool calls are complete.
Changes:
- Add SawToolCall bool to track tool calls across entire stream
- Add UpstreamFinishReason to cache the finish reason
- Only emit finish_reason on final chunk (has both finishReason + usage)
- Priority: tool_calls > max_tokens > stop
Includes 5 unit tests covering:
- Tool calls not overwritten by subsequent STOP
- Normal text gets "stop"
- MAX_TOKENS without tool calls gets "max_tokens"
- Tool calls take priority over MAX_TOKENS
- Intermediate chunks have no finish_reason
Fixes streaming tool call detection for Claude Code + Gemini models.
🤖 Generated with [Claude Code](https://claude.com/claude-code)
Co-Authored-By: Claude <noreply@anthropic.com>
Previously, metadataEqualIgnoringTimestamps() ignored access_token for all
providers, which prevented refreshed tokens from being persisted to disk/database.
This caused tokens to be lost on server restart for providers like iFlow.
This change makes the behavior provider-specific:
- Providers like gemini/gemini-cli that issue new tokens on every refresh and
can re-fetch when needed will continue to ignore access_token (optimization)
- Other providers like iFlow will now persist access_token changes
🤖 Generated with [Claude Code](https://claude.com/claude-code)
Co-Authored-By: Claude <noreply@anthropic.com>
integrate claude-cloak functionality to disguise api requests:
- add CloakConfig with mode (auto/always/never) and strict-mode options
- generate fake user_id in claude code format (user_[hex]_account__session_[uuid])
- inject claude code system prompt (configurable strict mode)
- obfuscate sensitive words with zero-width characters
- auto-detect claude code clients via user-agent
🤖 Generated with [Claude Code](https://claude.com/claude-code)
Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>
Enhanced node structure by including `thoughtSignature` for inline data parts in Gemini OpenAI, Gemini CLI, and Antigravity request handlers to improve traceability of thought processes.
Fixes issue where free tier users cannot access Gemini 3 preview models
due to frontend/backend project ID mapping.
## Problem
Google's Gemini API uses a frontend/backend project mapping system for
free tier users:
- Frontend projects (e.g., gen-lang-client-*) are user-visible
- Backend projects (e.g., mystical-victor-*) host actual API access
- Only backend projects have access to preview models (gemini-3-*)
Previously, CLIProxyAPI ignored the backend project ID returned by
Google's onboarding API and kept using the frontend ID, preventing
access to preview models.
## Solution
### CLI (internal/cmd/login.go)
- Detect free tier users (gen-lang-client-* projects or FREE/LEGACY tier)
- Show interactive prompt allowing users to choose frontend or backend
- Default to backend (recommended for preview model access)
- Pro users: maintain original behavior (keep frontend ID)
### Web UI (internal/api/handlers/management/auth_files.go)
- Detect free tier users using same logic
- Automatically use backend project ID (recommended choice)
- Pro users: maintain original behavior (keep frontend ID)
### Deduplication (internal/cmd/login.go)
- Add deduplication when user selects ALL projects
- Prevents redundant API calls when multiple frontend projects map to
same backend
- Skips duplicate project IDs in activation loop
## Impact
- Free tier users: Can now access gemini-3-pro-preview and
gemini-3-flash-preview models
- Pro users: No change in behavior (backward compatible)
- Only affects Gemini CLI OAuth (not antigravity or API key auth)
## Testing
- Tested with free tier account selecting single project
- Tested with free tier account selecting ALL projects
- Verified deduplication prevents redundant onboarding calls
- Confirmed pro user behavior unchanged
- Add Priority field to Auth for per-provider account ordering
- Add UsageStats struct tracking total/daily tokens and requests
- Track usage in MarkResult with daily reset logic
- Add TokenExpiresAt for proactive refresh (5min buffer)
- Sort available auths by priority in selector
Remove os.RemoveAll() call in syncAuthFromBucket() that was causing
a race condition with the file watcher.
Problem:
1. syncAuthFromBucket() wipes local auth directory with RemoveAll
2. File watcher detects deletions and propagates them to remote store
3. syncAuthFromBucket() then pulls from remote, but files are now gone
Solution:
Use incremental sync instead of delete-then-pull. Just ensure the
directory exists and overwrite files as they're downloaded.
This prevents the watcher from seeing spurious delete events.
Fixed incorrect boundary logic for `message_delta` emission, ensuring proper handling of usage updates and `emitMessageStopIfNeeded` within the response loop.
Implemented `Fork` flag in `ModelNameMapping` to allow aliases as additional models while preserving the original model ID. Updated the `applyOAuthModelMappings` logic, added tests for `Fork` behavior, and updated documentation and examples accordingly.
Introduced `tokenRefreshTimeout` constant for token refresh operations and enhanced context propagation for `refreshToken` by embedding roundtrip information if available. Adjusted `refreshAuth` to ensure default context initialization and handle cancellation errors appropriately.
Replaced file-based auth entry counting with `TokenStore`-backed implementation, enhancing flexibility and context-aware token management. Updated related logic to reflect this change.
Added support for external hooks to observe model registry events using the `ModelRegistryHook` interface. Implemented thread-safe, non-blocking execution of hooks with panic recovery. Comprehensive tests added to verify hook behavior during registration, unregistration, blocking, and panic scenarios.
Refactored `applyPayloadConfig` to `applyPayloadConfigWithRoot`, adding support for default rule validation against the original payload when available. Updated all executors to use `applyPayloadConfigWithRoot` and incorporate an optional original request payload for translations.
This commit reverts all modifications within internal/translator. A separate issue
will be created for the maintenance team to integrate SanitizeFunctionName into
the translators.
Co-authored-by: factory-droid[bot] <138933559+factory-droid[bot]@users.noreply.github.com>
This commit centralizes tool name sanitization in SanitizeFunctionName,
applying character compliance, starting character rules, and length limits.
It also fixes a regression in gemini_schema tests and preserves MCP-specific
shortening logic while ensuring compliance.
Co-authored-by: factory-droid[bot] <138933559+factory-droid[bot]@users.noreply.github.com>
Implemented SanitizeFunctionName utility to ensure Claude tool names meet
Gemini/Upstream strict naming conventions (alphanumeric, starts with letter/underscore, max 64 chars).
Applied sanitization to tool definitions and usage in all relevant translators.
Co-authored-by: factory-droid[bot] <138933559+factory-droid[bot]@users.noreply.github.com>
When receiving HTTP 429 (Too Many Requests) responses, parse the retry
delay from the response body using parseRetryDelay and populate the
statusErr.retryAfter field. This allows upstream callers to respect
the server's requested retry timing.
Applied to all error paths in Execute, executeClaudeNonStream,
ExecuteStream, CountTokens, and refreshToken functions.
COMPLETED:
- Added ViewWithSize(width, height) pattern to all submodels
- login.go: responsive provider rows with adaptive widths
- switch.go: responsive agent rows without card borders
- mappings.go: responsive table columns with calculateColumnWidths()
- usage.go: responsive stat cards (2x2 grid layout)
- logs.go: responsive log entries with message truncation
- Fixed main View() layout calculation in tui.go
- Properly accounts for border (2) + padding (2) = 4 chars per panel
- sidebarWidth now represents content width (20/14 for normal/narrow)
- contentWidth = m.width - sidebarWidth - 9 (accounts for both panels)
- Added new screen files:
- integrations.go: setup/integration screen
- logs.go: real-time log viewer with filtering
- mappings.go: model alias mappings table
- usage.go: usage statistics with charts
PENDING TEST:
- Verify sidebar displays correctly alongside content (was overflowing)
- Test all screens at various terminal widths (60, 80, 100+ cols)
KNOWN ISSUE BEING FIXED:
- Sidebar was showing only "[8] Setup" at top due to width overflow
- Content panel was too wide, causing horizontal wrap
Cherry-picked fixes:
- fix(antigravity): inject required placeholder when properties exist without required
- fix(translators): correct key path for system_instruction.parts in Claude request logic
- Add package.json with bun postinstall script
- Add install.ts for binary download
- Add bin wrapper script
- Add README with usage instructions
- Version 0.1.0
Add 'Shell' button per agent that opens modal displaying environment
variables and shell profile configuration snippets. Includes copy-to-
clipboard functionality for easy shell integration.
Add live/history toggle, date range filtering, model/provider/status
filters, pagination, export/import/clear buttons, and statistics display.
Fetches from persistent /request-history endpoint.
Add standalone ppswitch binary for fast agent configuration switching
without server logging overhead. Usage:
- ppswitch Show all agent statuses
- ppswitch <agent> Show specific agent status
- ppswitch <agent> <mode> Switch agent to proxy/native mode
Add --switch and --mode flags plus subcommand-style 'proxypilot switch'
syntax for quick agent configuration switching. Supports both flag style
(--switch claude --mode proxy) and subcommand style (proxypilot switch
claude proxy).
Add DoSwitch() and DoSwitchStatusAll() for switching agent configs
between proxy mode (routed through ProxyPilot) and native mode (direct
API access). Supports claude, gemini, codex, opencode, droid, cursor,
kilo, and roocode agents with backup/restore functionality.
Add GET/PUT /thinking-budget endpoint for configuring extended
thinking token budgets. Supports preset levels (Low 2048, Medium 8192,
High 32768) and custom values with enable/disable toggle.
Add management endpoints for request history:
- GET /request-history - list entries with pagination
- GET /request-history/stats - aggregate statistics
- DELETE /request-history - clear all entries
- POST /request-history/export - export to JSON
- POST /request-history/import - import from JSON
- POST /request-history/save - force save to disk
Integrate RequestHistoryManager with request logging middleware.
Add RequestHistoryManager with file-based persistence at
~/.cli-proxy-api/request-history.json. Features:
- Max 500 entries with automatic pruning
- Auto-save every 30 seconds
- Thread-safe with mutex protection
- In-memory ring buffer for performance
Add GetShellConfig() to AgentHandler interface returning ShellConfigInfo
with environment variables and shell profile snippets. Implement for all
agents: Claude Code, Codex, Gemini CLI, Cursor, OpenCode, Factory Droid,
and Amazon Q CLI.
Major overhaul of agent integration system:
- Add support for additional agent types
- Improve configuration detection and management
- Better error handling and validation
- Add --minimax-login, --zhipu-login for API key providers
- Add --amazonq-import for Amazon Q CLI token import
- Add --kiro-import for Kiro IDE token import
- Add --kiro-aws-login for AWS Builder ID flow
- Improve agent detection logic for multiple CLI tools
- Add support for detecting Amazon Q CLI, OpenCode
- Better path resolution and state management
Add React-based UI for the custom installer:
- Clean modern installation wizard interface
- Progress tracking and status display
- TypeScript + Vite build
Document the migration from iFlow to MiniMax and Zhipu AI providers.
Includes:
- Summary of changes
- Files deleted/created/modified
- New model definitions
- API details for new providers
- Auth file format examples
- Migration notes for existing users
Remove the iFlow OAuth and cookie-based authentication implementation.
This provider is being replaced with simpler API key-based providers.
Deleted files:
- internal/auth/iflow/cookie_helpers.go
- internal/auth/iflow/iflow_auth.go
- internal/auth/iflow/iflow_token.go
- internal/auth/iflow/oauth_server.go
When multiple OAuth providers share an account email, the existing "Use OAuth" debug lines are ambiguous and hard to correlate with management usage stats. Include provider, auth file, and auth index in the selection log, and only compute these fields when debug logging is enabled to avoid impacting normal request performance.
Before:
[debug] Use OAuth user@example.com for model gemini-3-flash-preview
[debug] Use OAuth user@example.com (project-1234) for model gemini-3-flash-preview
After:
[debug] Use OAuth provider=antigravity auth_file=antigravity-user_example_com.json auth_index=1a2b3c4d5e6f7788 account="user@example.com" for model gemini-3-flash-preview
[debug] Use OAuth provider=gemini-cli auth_file=gemini-user@example.com-project-1234.json auth_index=99aabbccddeeff00 account="user@example.com (project-1234)" for model gemini-3-flash-preview
Includes:
- Vision base64 parsing fix for OpenAI-to-Claude translation
- Memory API localhost authentication with API keys
- Global model mapping SDK hook for cross-provider aliases
- Comprehensive unit tests for auth, config, memory, and runtime
Adds GlobalModelMapperFunc type and hook to SDKConfig, allowing
SDK-level handlers to access Config-level global model mappings.
This enables cross-provider model aliasing (e.g., gpt-4o → claude-sonnet).
Anthropic API does not allow extended thinking when tool_choice is set
to "any" or a specific tool. This was causing 400 errors when using
features like Amp's /handoff command which forces tool_choice.
Added disableThinkingIfToolChoiceForced() that removes thinking config
when incompatible tool_choice is detected, applied to both streaming
and non-streaming paths.
Fixesrouter-for-me/CLIProxyAPI#630
This release removes the need for a separate proxypilot-engine.exe binary.
The ProxyPilot.exe tray app now runs the proxy service in-process using
the cliproxy SDK builder pattern.
Changes:
- Add internal/embedded package for in-process server management
- Update desktopctl to support embedded mode alongside subprocess mode
- Clean up vestigial -exe flag and engine binary references
- Update package scripts to build only single binary
- Bump version to 0.1.2
The embedded engine in cmd/cliproxytray was already implemented; this
commit cleans up the remaining references to the old two-binary architecture.
Prepare release v0.1.1 with new features:
- Model-specific pricing for cost estimation
- Auto-update with download, verify, and install flow
- Cost savings display in Analytics
- Logo branding assets
Update Header with logo display, improve AgentSetup and ProviderLogins
components. Update App.tsx and index.html for better structure.
Adjust Vite config for development.
Add complete update workflow with stages: checking, available, downloading,
ready, installing. Show download progress bar, version info, and release
notes. Add Install & Restart button for applying updates.
Add Cost Savings hero card showing proxy cost vs direct API cost with
visual comparison bars. Add Cost column to Model Breakdown table.
Display savings percentage and total amount saved.
Add handlers for update status, download, verify, and install operations.
Register new routes: GET /updates/status, POST /updates/download,
POST /updates/verify, POST /updates/install. Track download progress
with background goroutine and mutex-protected state.
Add install.go with archive extraction (zip, tar.gz) and binary replacement
logic. Windows uses batch script for safe replacement of running executable.
Unix directly replaces binary with backup. Includes cleanup for old versions.
Add DownloadUpdate function to download platform-specific binaries from
GitHub releases with progress tracking. Add verify.go with SHA256 checksum
verification for downloaded files. Fix GitHub repo URL and add 404 handling
for repositories without releases.
Add comprehensive pricing table for Claude, GPT, Gemini, Qwen, and DeepSeek
models with per-million-token rates. Update cost calculation to use actual
model pricing instead of hardcoded estimates. Add new cost fields to
UsageStats for tracking actual cost, direct API cost, and savings.
- GLM-4.7: Uses extra_body={"thinking": {"type": "enabled"}, "clear_thinking": false}
- MiniMax-M2.1: Uses reasoning_split=true for OpenAI-style reasoning separation
- Added preserveReasoningContentInMessages() to support re-injection of reasoning
content in assistant message history for multi-turn conversations
- Added ThinkingSupport to MiniMax-M2.1 model definition
- Add agent detection and configuration bindings
- Add update check and download bindings
- Add usage and request monitoring bindings
- Update useProxyContext with new desktop API hooks
- Refactor EngineControl with new status indicators
- Update Integrations with provider cards
- Enhance LogsViewer with better filtering and display
- Improve MemoryManager with modern UI patterns
- Update ModelMappings with collapsible sections
- Refactor ProviderLogins with OAuth flow improvements
- Update SemanticMemory styling
- Add versioninfo.json for goversioninfo in tray and server
- Generate resource.syso files with embedded icon
- Update ico.go to embed actual icon.ico for tray
- All exes now show ProxyPilot logo in File Explorer
- Embed React dashboard assets in tray binary
- Open dashboard in WebView2 window from tray
- Remove need for separate proxypilotui.exe
- Simplifies distribution to just ProxyPilot.exe + proxypilot-engine.exe
🤖 Generated with [Claude Code](https://claude.com/claude-code)
Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>
- Remove logo and test files from tracking
- Keep files locally but exclude from repo
🤖 Generated with [Claude Code](https://claude.com/claude-code)
Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>
Credits Factory.ai and Anthropic for their research on context
compression and building effective agents.
🤖 Generated with [Claude Code](https://claude.com/claude-code)
Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>
- Cleaner centered branding with star badge
- Bold tagline with concise value prop
- GitHub tip callout for latest models
- Emoji bullet feature list
- Simplified sections
- Setup guide links
🤖 Generated with [Claude Code](https://claude.com/claude-code)
Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>
Removed dead code (commented-out ErrTokenExpired error constant) from
both claude/errors.go and codex/errors.go.
🤖 Generated with [Claude Code](https://claude.com/claude-code)
Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>
Removed redundant `_ "embed"` import as `embed` is already imported
and used for embed.FS.
🤖 Generated with [Claude Code](https://claude.com/claude-code)
Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>
Remove desktop-ui-spec.md and windows-ui-spec.md which reference
deleted cliproxyctl CLI that no longer exists.
🤖 Generated with [Claude Code](https://claude.com/claude-code)
Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>
Type was defined but never used anywhere in codebase.
Keep EmailRequiredError which is actively used.
🤖 Generated with [Claude Code](https://claude.com/claude-code)
Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>
Document Factory.ai context compression and Anthropic agentic harness
implementations with architecture diagrams and configuration reference.
🤖 Generated with [Claude Code](https://claude.com/claude-code)
Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>
Test context compression adapter and harness middleware work together
without conflicts.
🤖 Generated with [Claude Code](https://claude.com/claude-code)
Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>
Removed feature not present in original Factory.ai or Anthropic sources.
Keep implementation pure to research references.
🤖 Generated with [Claude Code](https://claude.com/claude-code)
Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>
Enable Anthropic-style INITIALIZER/CODING mode prompts for long-running
agentic clients.
🤖 Generated with [Claude Code](https://claude.com/claude-code)
Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>
Add coreManagerWrapper to adapt auth manager types and initialize
LLM-based summarization after server creation.
🤖 Generated with [Claude Code](https://claude.com/claude-code)
Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>
Enables Factory.ai-style LLM summarization by wiring the summarizer
executor to the core auth manager.
🤖 Generated with [Claude Code](https://claude.com/claude-code)
Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>
Bridge type differences between coreauth.Manager and memory.AuthExecutor
to enable LLM-based context compression through the auth pipeline.
🤖 Generated with [Claude Code](https://claude.com/claude-code)
Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>
- Centered logo and badges
- Catchy tagline "Stop juggling API keys"
- Emoji-styled feature list
- Latest models callout section
- Cleaner installation/usage sections
🤖 Generated with [Claude Code](https://claude.com/claude-code)
Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>
- Add downloadBlob helper to reduce duplication
- Refactor export functions to use helper
- Fix model mapping edit to use controlled React state
instead of DOM getElementById manipulation
🤖 Generated with [Claude Code](https://claude.com/claude-code)
Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>
Delete input.tsx, progress.tsx, tabs.tsx - none were imported
or used anywhere in the codebase.
🤖 Generated with [Claude Code](https://claude.com/claude-code)
Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>
Complete rewrite with:
- ProxyPilot branding and badges
- Feature list and architecture overview
- Quick start and configuration guide
- API endpoints documentation
🤖 Generated with [Claude Code](https://claude.com/claude-code)
Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>
- Remove internal/letta/ package (unused memory bridge)
- Remove Letta references from run.go and start-cliproxy.ps1
- Move docker-build scripts to scripts/
- Move TROUBLESHOOTING_ANTIGRAVITY.md and WARP.md to docs/
🤖 Generated with [Claude Code](https://claude.com/claude-code)
Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>
- Update droid.go to use ~/.factory/settings.json (runtime config)
- Add /healthz endpoint for Droid health checks
- Add root-level API routes (/responses, /messages, etc.)
- Fix detection in proxypilotui to check settings.json
- Use thinking models with provider: "openai" for Responses API
🤖 Generated with [Claude Code](https://claude.com/claude-code)
Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>
Add WebView2 JavaScript bindings:
- pp_detect_agents: List installed CLI agents
- pp_configure_agent: Configure agent for ProxyPilot
- pp_get_model_mappings: Get global mappings
- pp_save_model_mappings: Save global mappings
🤖 Generated with [Claude Code](https://claude.com/claude-code)
Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>
Remove bloated React bundle (~500KB) in favor of
simple embedded HTML (~15KB). Update embed.go to
only include index.html and vite.svg.
🤖 Generated with [Claude Code](https://claude.com/claude-code)
Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>
Generate config files to point CLI agents at ProxyPilot:
- Claude Code: ~/.claude/settings.json
- Codex CLI: ~/.codex/config.toml
- Factory Droid: ~/.factory/config.json
🤖 Generated with [Claude Code](https://claude.com/claude-code)
Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>
Detect installed CLI agents (Claude Code, Codex, Droid, Gemini)
by checking binaries and config files.
🤖 Generated with [Claude Code](https://claude.com/claude-code)
Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>
Add global-model-mappings section to config.example.yaml
showing smart/rush/think aliases.
🤖 Generated with [Claude Code](https://claude.com/claude-code)
Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>
Add API handler for managing global model mappings
stored in ~/.proxypilot/model-mappings.json.
🤖 Generated with [Claude Code](https://claude.com/claude-code)
Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>
Add GlobalModelMapping struct and LookupGlobalModelMapping()
for routing friendly model names to actual models.
🤖 Generated with [Claude Code](https://claude.com/claude-code)
Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>
Add CRUD operations for harness files:
- GET /v0/management/harness/files - list files
- GET /v0/management/harness/file - read file
- PUT /v0/management/harness/file - write file
- GET /v0/management/harness/export - export all
- POST /v0/management/harness/import - import files
🤖 Generated with [Claude Code](https://claude.com/claude-code)
Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>
Improve INITIALIZER and CODING prompts based on Anthropic guidance:
- Structured feature_list.json format
- GET BEARINGS workflow for coding mode
- Puppeteer MCP detection for browser testing
- Session-aware state detection
🤖 Generated with [Claude Code](https://claude.com/claude-code)
Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>
Add utilities for estimating token counts using tiktoken
to support token-aware compression decisions.
🤖 Generated with [Claude Code](https://claude.com/claude-code)
Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>
Implement proactive compression based on token count rather than bytes:
- Uses tiktoken for accurate token estimation
- Compresses at 75% threshold before hitting limits
- Reserves tokens for output generation
🤖 Generated with [Claude Code](https://claude.com/claude-code)
Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>
Update file_store.go with BuildAnchoredSummaryWithLLM() for
generating context-aware summaries during compression.
🤖 Generated with [Claude Code](https://claude.com/claude-code)
Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>
Add structured summary generation for context compression:
- structured_summary.go: Types and parsing for summaries
- summarizer.go: LLM summarization logic
- summarizer_executor.go: Routes to LLM pipeline
🤖 Generated with [Claude Code](https://claude.com/claude-code)
Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>
Add Kiro OAuth handling in management API:
- Fix ExpiresAt field name in KiroTokenStorage
🤖 Generated with [Claude Code](https://claude.com/claude-code)
Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>
Add KiroAuthenticator for AWS Builder ID device flow:
- Client registration with AWS SSO OIDC
- Device authorization with user code display
- Token polling with timeout handling
🤖 Generated with [Claude Code](https://claude.com/claude-code)
Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>
Add Kiro support in SnapshotCoreAuths():
- Synthesize auth entries from KiroKey config
- Support token-file and direct access-token modes
- Handle profile-arn, region, agent-task-type
🤖 Generated with [Claude Code](https://claude.com/claude-code)
Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>
Add KiroExecutor for AWS CodeWhisperer API:
- AWS Event Stream binary format parsing
- Kiro payload building with origin field
- Token refresh with mutex protection
- Quota fallback: AI_EDITOR -> CLI on 429
- Auto-refresh on 401/403 errors
🤖 Generated with [Claude Code](https://claude.com/claude-code)
Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>
Remove dead code that was never used:
- toolCallIDToName map: built but never read from
- seenToolCallIDs: declared but never populated, only suppressed with _
Fix three issues in Kiro OpenAI translator that caused "Improperly formed request"
errors when processing LiteLLM-translated requests with tool_use/tool_result:
1. Skip merging tool role messages in MergeAdjacentMessages() to preserve
individual tool_call_id fields
2. Track pendingToolResults and attach to the next user message instead of
only the last message. Create synthetic user message when conversation
ends with tool results.
3. Insert synthetic user message with tool results before assistant messages
to maintain proper alternating user/assistant structure. This fixes the case
where LiteLLM translates Anthropic user messages containing only tool_result
blocks into tool role messages followed by assistant.
Adds unit tests covering all tool result handling scenarios.
Improved consistency across OpenAI, Claude, and Gemini handlers by replacing initial `select` statement with a `for` loop for better readability and error-handling robustness.
Refined header assignment to use `x-api-key` for Anthropic API requests, ensuring correct authorization behavior based on request attributes and URL validation.
- Handle errors in promptInput instead of ignoring them
- Improve promptSelect to provide feedback on invalid input and re-prompt
- Use sentinel errors (ErrAuthorizationPending, ErrSlowDown) instead of
string-based error checking with strings.Contains
- Move hardcoded x-amz-user-agent header to idcAmzUserAgent constant
Addresses code review feedback from Gemini Code Assist.
Add persistRefreshedAuth function to write refreshed tokens back to the
auth file after inline token refresh. This prevents repeated token
refreshes on every request when the token expires.
Changes:
- Add persistRefreshedAuth() to kiro_executor.go
- Call persist after all token refresh paths (401, 403, pre-request)
- Remove unused log import from sdk/auth/kiro.go
Updated Antigravity, Gemini, and Gemini-CLI translators to process `systemResult` of type `string` for system instructions. Ensures properly formatted JSON with dynamic content assignment.
fix(antigravity): validate function arguments before serialization
Ensure `function.arguments` is a valid JSON before setting raw bytes, fallback to setting as parameterized content if invalid.
feat: handle array input for system instructions in translators
Enhanced Gemini, Gemini-CLI, and Antigravity translators to process array content for system instructions. Adds support for assigning roles and handling multiple content parts dynamically.
Added comprehensive tests for `FillFirstSelector` and `RoundRobinSelector` to ensure proper behavior, including deterministic, cyclical, and concurrent scenarios. Introduced dynamic routing strategy updates in `service.go`, normalizing strategies and seamlessly switching between `fill-first` and `round-robin`. Updated `Manager` to support selector changes via the new `SetSelector` method.
Optimized the handling of JSON serialization and deserialization by replacing redundant `json.Marshal` and `json.Unmarshal` calls with `sjson` and `gjson`. Introduced a `marshalJSONValue` utility for compact JSON encoding, improving performance and code simplicity. Removed unused `encoding/json` imports.
This update enhances the `FileRequestLogger` by introducing support for spooling large request and response bodies to temporary files, reducing memory consumption. It adds atomic requestLogID generation for sequential log naming and new methods for non-streaming/streaming log assembly. Also includes better error handling during logging and temp file cleanups.
Improve parsing of tool call inputs and Antigravity compatibility to avoid invalid thinking/tool_use errors.
- Parse tool call inputs robustly by accepting both object and JSON-string formats and only produce a functionCall part when valid args exist, reducing spurious or malformed parts.
- Preserve the skip_thought_signature_validator approach for calls without a valid thinking signature but stop toggling/tracking a separate "disable thinking" flag; this prevents unnecessary removal of thinkingConfig.
- Sanitize tool input schemas before attaching them to the Antigravity request to improve compatibility.
- Append the interleaved-thinking hint as a new parts entry instead of overwriting/setting text directly, preserving structure.
- Remove unused tracking logic and related comments to simplify flow.
These changes reduce errors related to missing/invalid thinking signatures, improve schema compatibility, and make hint injection safer and more consistent.
Removes the addition of the "anthropic-beta: interleaved-thinking-2025-05-14" header for Claude thinking models when building HTTP requests.
This prevents sending an experimental/feature flag header that is no longer required and avoids potential compatibility or routing issues with downstream services. Keeps request headers simpler and more standard.
Prefer cached signatures and avoid injecting dummy thinking blocks; instead remove unsigned thinking blocks and add a skip sentinel for tool calls without a valid signature. Generate stable session IDs from the first user message, apply schema cleaning only for Claude models, and reorder thinking parts so thinking appears first. For Gemini, remove thinking blocks and attach a skip sentinel to function calls. Simplify response handling by passing raw function args through (remove special Bash conversion). Update and add tests to reflect the new behavior.
These changes prevent rejected dummy signatures, improve compatibility with Antigravity’s signature validation, provide more stable session IDs for conversation grouping, and make request/response translation more robust.
Remove the opencode-antigravity-auth submodule reference from the repository.
Cleans up the project by eliminating an external submodule pointer that is no longer needed or maintained, reducing repository complexity and avoiding dangling submodule state.
Removes two obsolete git submodules to clean up repository state and reduce maintenance overhead.
This eliminates external references that are no longer needed, simplifying dependency management and repository maintenance going forward.
fix: unify response field naming across translators
Standardize `text` to `delta` and add missing `output` field in all response payloads for consistency across OpenAI, Claude, and Gemini translators.
This commit introduces several improvements to the AMP (Advanced Model Proxy) module:
- **Model Mapping Logic:** The `FallbackHandler` now uses a more robust approach for model mapping. It includes the extraction and preservation of dynamic "thinking suffixes" (e.g., `(xhigh)`) during mapping, ensuring that these configurations are correctly applied to the mapped model. A new `resolveMappedModel` function centralizes this logic for cleaner code.
- **ModelMapper Verification:** The `ModelMapper` in `model_mapping.go` now verifies that the target model of a mapping has available providers *after* normalizing it. This prevents mappings to non-existent or unresolvable models.
- **Gemini Thinking Configuration Cleanup:** In `gemini_thinking.go`, unnecessary `generationConfig.thinkingConfig.include_thoughts` and `generationConfig.thinkingConfig.thinkingBudget` fields are now deleted from the request body when applying Gemini thinking levels. This prevents potential conflicts or redundant configurations.
- **Testing:** A new test case `TestModelMapper_MapModel_TargetWithThinkingSuffix` has been added to `model_mapping_test.go` to specifically cover the preservation of thinking suffixes during model mapping.
This commit introduces a new configuration option `logs-max-total-size-mb` that allows users to set a maximum total size (in MB) for log files in the logs directory. When this limit is exceeded, the oldest log files will be automatically deleted to stay within the specified size. Setting this value to 0 (the default) disables this feature. This change enhances log management by preventing excessive disk space usage.
Add applyPayloadConfig calls to all Antigravity executor paths (Execute,
executeClaudeNonStream, ExecuteStream) to enable config.yaml payload
overrides for Antigravity/Gemini-Claude models.
This allows users to configure thinking budget and other parameters via
payload.override in config.yaml for models like gemini-claude-opus-4-5*.
Implements a new middleware to handle agentic headers and dynamically inject system prompts for 'Initializer' and 'Coding' agents. Also registers this middleware in the server setup.
Fixes a critical bug where rate limit errors were only updating the parent Auth object and not the specific ModelState. This caused the selector to incorrectly view the model as active and retry the same account immediately. Also hardens the selector to respect timestamps even if availability flags are desynchronized.
Currently, the executor ignores upstream retry-after headers, causing the manager to default to a 1s backoff. This change parses standard HTTP Retry-After headers (seconds or date) and passes them to the error handler.
- Add claude-cli detection to isAgenticCLI and mustKeepTools checks
- Add /v1/messages Claude Messages API format support
- Implement trimClaudeMessagesWithMemory with tool_use/tool_result adjacency preservation
- Add extractLastUserTextFromClaude, extractTextFromClaudeMessage, truncateClaudeMessageContent helpers
- Add claude case to prependToLastUserText, appendToLastUserText, extractPinnedContext
- Add response headers for localhost diagnostics:
- X-ProxyPilot-Session: session ID
- X-ProxyPilot-Request-Shape: detected format (chat/responses/claude)
- X-ProxyPilot-Memory-Stored: dropped message count
- X-ProxyPilot-Memory-Retrieved: injected snippet count
Refactors the reasoning effort conversion logic for Gemini models.
The update specifically addresses how `reasoning_effort` is translated into Gemini 3 specific thinking configurations (`thinkingLevel`, `includeThoughts`) and ensures that numeric budgets are not incorrectly applied to level-based models.
Changes include:
- Differentiating conversion logic for Gemini 3 models versus other models.
- Handling `none`, `auto`, and validated thinking levels for Gemini 3.
- Maintaining existing conversion for models not using discrete thinking levels.
The prompt for the gpt-5.2 codex model has been updated with more comprehensive instructions. This includes detailed guidelines on general usage, editing constraints, the plan tool, sandboxing configurations, handling special user requests, frontend task considerations, and final message presentation. The updates aim to improve the model's understanding and execution of complex coding tasks by providing clearer directives and constraints.
Improve the cache eviction routine to sort entries by timestamp using the standard library sort routine (stable, clearer and faster than the prior manual selection/bubble logic), and remove a redundant request-derived session ID helper in favor of the centralized session ID function. Also drop now-unused crypto/encoding imports.
This yields clearer, more maintainable eviction logic and removes duplicated/unused code and imports to reduce surface area and potential inconsistencies.
Improve robustness when handling "thinking" content by using a dedicated helper to extract the thinking text. This ensures wrapped or nested thinking objects are handled correctly instead of relying on a direct string extraction, reducing parsing errors for complex payloads.
This change introduces specific logic to load and use instructions for the 'gpt-5.2-codex' model variant by recognizing the 'gpt-5.2-codex_prompt.md' filename. This ensures the correct prompts are used when the '5.2-codex' model is identified, complementing the recent addition of its definition.
Normalize Bash tool arguments by converting a "command" key into "cmd" using JSON-aware parsing, avoiding brittle string replacements that could corrupt values. Apply this conversion in both streaming and non-streaming response paths so bash-style tool calls are emitted with the expected "cmd" field.
Add support for accumulating thinking text and carrying session identifiers to enable signature caching/restore for unsigned thinking blocks, improving handling of thinking-state continuity across requests/responses.
Also perform small cleanups: import logging, tidy comments and test descriptions. These changes make tool-argument handling more robust and enable reliable signature restoration for thinking blocks.
Ensures compatibility with the Amp client by suppressing
"thinking" blocks when "tool_use" blocks are also present in
the response.
The Amp client has issues rendering both types of blocks
simultaneously. This change filters out "thinking" blocks in
such cases, preventing rendering problems.
Gemini API does not support the JSON Schema `propertyNames` keyword,
causing 400 errors when Claude tool schemas containing this field are
proxied through the Antigravity provider.
Add `propertyNames` to the list of unsupported keywords removed by
CleanJSONSchemaForGemini(), alongside existing removals like $ref,
definitions, and additionalProperties.
Introduce a centralized OAuth session store with TTL-based expiration
to replace the previous simple map-based status tracking. Add a new
/api/oauth/callback endpoint that allows remote clients to relay OAuth
callback data back to the CLI proxy, enabling OAuth flows when the
callback cannot reach the local machine directly.
- Add oauth_sessions.go with thread-safe session store and validation
- Add oauth_callback.go with POST handler for remote callback relay
- Refactor auth_files.go to use new session management APIs
- Register new callback route in server.go
Add metadataEqualIgnoringTimestamps() function to compare metadata JSON
without timestamp/expired/expires_in/last_refresh/access_token fields.
This prevents unnecessary file writes when only these fields change
during refresh, breaking the fsnotify event → Watcher callback → refresh loop.
Key insight: Google OAuth returns a new access_token on each refresh,
which was causing file writes and triggering the refresh loop.
Fixes antigravity channel excessive log generation issue.
🤖 Generated with [Claude Code](https://claude.com/claude-code)
Co-Authored-By: Claude <noreply@anthropic.com>
The previous commit added thinkingLevel support but didn't apply it
when the reasoning effort came from model name suffix (e.g., model(minimal)).
This was because ResolveThinkingConfigFromMetadata returns nil for
level-based models, bypassing the metadata application.
Changes:
- Add ApplyGemini3ThinkingLevelFromMetadata for standard Gemini API
- Add ApplyGemini3ThinkingLevelFromMetadataCLI for CLI API format
- Update gemini_cli_executor to apply Gemini 3 thinkingLevel from metadata
- Update antigravity_executor to apply Gemini 3 thinkingLevel from metadata
- Update aistudio_executor to apply Gemini 3 thinkingLevel from metadata
- Add comprehensive test coverage for Gemini 3 thinkingLevel functions
Per Google's official documentation, Gemini 3 models should use
thinkingLevel (string) instead of thinkingBudget (number) for
optimal performance.
From Google's Gemini Thinking docs:
> Use the thinkingLevel parameter with Gemini 3 models. While
> thinkingBudget is accepted for backwards compatibility, using
> it with Gemini 3 Pro may result in suboptimal performance.
Changes:
- Add model family detection functions (IsGemini3Model, IsGemini25Model,
IsGemini3ProModel, IsGemini3FlashModel)
- Add ApplyGeminiThinkingLevel and ApplyGeminiCLIThinkingLevel functions
for applying thinkingLevel config
- Add ValidateGemini3ThinkingLevel for model-specific level validation
- Add ThinkingBudgetToGemini3Level for backward compatibility conversion
- Update NormalizeGeminiThinkingBudget to convert budget to level for
Gemini 3 models
- Update ApplyDefaultThinkingIfNeeded to not set a default level for
Gemini 3 (lets API use its dynamic default "high")
- Update ConvertThinkingLevelToBudget to preserve thinkingLevel for
Gemini 3 models
- Add Levels field to all Gemini 3 model definitions:
- Gemini 3 Pro: ["low", "high"]
- Gemini 3 Flash: ["minimal", "low", "medium", "high"]
Backward compatibility:
- Gemini 2.5 models continue to use thinkingBudget as before
- If thinkingBudget is provided for Gemini 3, it's converted to the
appropriate thinkingLevel
- Existing configurations continue to work
Introduces the capability to count tokens for Antigravity-backed requests. This implementation leverages the `countTokens` endpoint of the Antigravity API, replacing the prior unsupported stub.
Key aspects of this update include:
- **API Integration**: Direct integration with the Antigravity `countTokens` API, including necessary request payload translation and authentication.
- **Resilient Infrastructure**: A fallback mechanism has been established, allowing the system to attempt connections across multiple Antigravity base URLs to ensure request success even in the event of temporary service interruptions.
- **Model Aliasing**: Added mappings for `gemini-3-flash` and `gemini-3-flash-preview` to ensure compatibility with the latest model variants.
- **Robust Error Handling**: Comprehensive error handling and logging are in place to manage failures during API interactions.
Updates schema flattening logic to handle multiple non-null types, providing a more descriptive "Accepts" hint.
Removes redundant tracking of the current tool name in `Params` as it's no longer needed for streaming limits, simplifying the structure.
Enhances compatibility with the Gemini API by implementing a schema cleaning process.
This includes:
- Centralizing schema cleaning logic for Gemini in a dedicated utility function.
- Converting unsupported schema keywords to hints within the description field.
- Flattening complex schema structures like `anyOf`, `oneOf`, and type arrays to simplify the schema.
- Handling streaming responses with empty tool names, which can occur in subsequent chunks after the initial tool use.
Enhance the configuration diff logic to include detection and reporting of `prefix` changes for all model types. Update related struct naming for consistency across the watcher module.
- Document Windows Claude Code settings for using CLIProxyAPI (base URL, auth token, default model mapping)
- Add Claude alias remapping for opus/sonnet/haiku to Gemini-backed models
- Improve auth unavailability reporting (429 cooldown vs 503 blocked) and fallback provider inference
Normalize action handling by accommodating wildcard patterns in route definitions for Gemini endpoints. Adjust `request.Action` parsing logic to correctly process routes with prefixed actions.
Refactor model management to include an optional `prefix` field for model credentials, enabling better namespace handling. Update affected configuration files, APIs, and handlers to support prefix normalization and routing. Remove unused OpenAI compatibility provider logic to simplify processing.
Introduce formatProxyURL helper to sanitize proxy addresses before
logging, stripping credentials and path components while preserving
host information. Rework model hash computation to sort and deduplicate
name/alias pairs with case normalization, ensuring consistent output
regardless of input ordering. Add signature-based identification for
anonymous OpenAI-compatible provider entries to maintain stable keys
across configuration reloads. Replace direct stdout prints with
structured logger calls for file change notifications.
Break out config diffing, hashing, and OpenAI compatibility utilities into a dedicated diff package, update watcher to consume them, and add comprehensive tests for diff logic and watcher behavior.
Unify thinking budget-to-effort conversion in a shared helper, handle disabled/default thinking cases in translators, adjust zero-budget mapping, and drop the old OpenAI-specific helper with updated tests.
Some OpenAI-compatible providers (like GitHub Copilot) may send tool_calls
in the first streaming chunk without including the role field. The previous
implementation only emitted message_start when the first chunk contained
role="assistant", causing Anthropic protocol violations when tool calls
arrived first.
This fix ensures message_start is always emitted on the very first chunk,
preventing 'content_block_start before message_start' errors in clients
that strictly validate Anthropic SSE event ordering.
Introduce `panel-github-repository` in the configuration to allow specifying a custom repository for management panel assets. Update dependency versions and enhance asset URL resolution logic to support overrides.
- Pass through non-function tool definitions like web_search
- Translate tool_choice for built-in tools and function tools
- Add regression tests for built-in tool passthrough
The response writer wrapper has been refactored to more reliably capture response bodies for logging, fixing several edge cases.
- Implements `WriteString` to capture writes from `io.StringWriter`, which were previously missed by the `Write` method override.
- A new `shouldBufferResponseBody` helper centralizes the logic to ensure the body is buffered only when logging is active or for errors when `logOnErrorOnly` is enabled.
- Streaming detection is now more robust. It correctly handles non-streaming error responses (e.g., `application/json`) that are generated for a request that was intended to be streaming.
BREAKING CHANGE: The public methods `Status()`, `Size()`, and `Written()` have been removed from the `ResponseWriterWrapper` as they are no longer required by the new implementation.
The WriteErrorResponse function now caches the error response body in the gin context.
The deferred request logger checks for this cached response. If an error response is found, it bypasses the standard response logging. This prevents scenarios where an error is logged twice or an empty payload log overwrites the original, more detailed error log.
The API error handling is updated to return a structured JSON payload
instead of a plain text message. This provides more context and allows
clients to programmatically handle different error types.
The new error response has the following structure:
{
"error": {
"message": "...",
"type": "..."
}
}
The `type` field is determined by the HTTP status code, such as
`authentication_error`, `rate_limit_error`, or `server_error`.
If the underlying error message from an upstream service is already a
valid JSON string, it will be preserved and returned directly.
BREAKING CHANGE: API error responses are now in a structured JSON
format instead of plain text. Clients expecting plain text error
messages will need to be updated to parse the new JSON body.
All Amp management endpoints (e.g., /api/user, /threads) are now protected by the standard API key authentication middleware. This ensures that all management operations require a valid API key, significantly improving security.
As a result of this change:
- The `restrict-management-to-localhost` setting now defaults to `false`. API key authentication provides a stronger and more flexible security control than IP-based restrictions, improving usability in containerized environments.
- The reverse proxy logic now strips the client's `Authorization` header after authenticating the initial request. It then injects the configured `upstream-api-key` for the request to the upstream Amp service.
BREAKING CHANGE: Amp management endpoints now require a valid API key for authentication. Requests without a valid API key in the `Authorization` header will be rejected with a 401 Unauthorized error.
Expose thinking/effort normalization helpers from the executor package
so conversion tests use production code and stay aligned with runtime
validation behavior.
Route OpenAI reasoning effort through ThinkingEffortToBudget for Claude
translators, preserve "minimal" when translating OpenAI Responses, and
treat blank/unknown efforts as no-ops for Gemini thinking configs.
Also map budget -1 to "auto" and expand cross-protocol thinking tests.
When budget 0 maps to "none" for models that use thinking levels
but don't support that effort level, strip thinking fields instead
of setting an invalid reasoning_effort value.
Tests now expect removal for this edge case.
Move OpenAI `reasoning_effort` -> Gemini `thinkingConfig` budget logic into
shared helpers used by Gemini, Gemini CLI, and antigravity translators.
Normalize Claude thinking handling by preferring positive budgets, applying
budget token normalization, and gating by model support.
Always convert Gemini `thinkingBudget` back to OpenAI `reasoning_effort` to
support allowCompat models, and update tests for normalization behavior.
Only map OpenAI reasoning effort to Claude thinking for models that support
thinking and use budget tokens (not level-based thinking).
Also add "xhigh" effort mapping and adjust minimal/low budgets, with new
raw-payload conversion tests across protocols and models.
- **Thinking Support**:
- Enabled thinking support for all Kiro Claude models, including Haiku 4.5 and agentic variants.
- Updated `model_definitions.go` with thinking configuration (Min: 1024, Max: 32000, ZeroAllowed: true).
- Fixed `extended_thinking` field names in `model_registry.go` (from `min_budget`/`max_budget` to `min`/`max`) to comply with Claude API specs, enabling thinking control in clients like Claude Code.
- **Kiro Executor Fixes**:
- Fixed `budget_tokens` handling: explicitly disable thinking when budget is 0 or negative.
- Removed aggressive duplicate content filtering logic that caused truncation/data loss.
- Enhanced thinking tag parsing with `extractThinkingFromContent` to correctly handle interleaved thinking/text blocks.
- Added EOF handling to flush pending thinking tag characters, preventing data loss at stream end.
- **Performance**:
- Optimized Claude stream handler (v6.2) with reduced buffer size (4KB) and faster flush interval (50ms) to minimize latency and prevent timeouts.
Ensure thinking settings translate correctly across providers:
- Only apply reasoning_effort to level-based models and derive it from numeric
budget suffixes when present
- Strip effort string fields for budget-based models and skip Claude/Gemini
budget resolution for level-based or unsupported models
- Default Gemini include_thoughts when a nonzero budget override is set
- Add cross-protocol conversion and budget range tests
When using OpenAI-compatible providers with model aliases (e.g., glm-4.6-zai -> glm-4.6),
the alias resolution was correctly applied but then immediately overwritten by
ResolveOriginalModel, causing 'Unknown Model' errors from upstream APIs.
This fix skips the ResolveOriginalModel override when a model alias has already
been resolved, ensuring the correct model name is sent to the upstream provider.
Co-authored-by: Amp <amp@ampcode.com>
- Write each SSE chunk directly to c.Writer and flush immediately
- Remove buffered writer and ticker-based flushing that caused delayed output
- Add 500ms timeout case for consistency with OpenAI/Gemini handlers
- Clean up unused bufio import
This fixes the 'not streaming' issue where small responses were held
in the buffer until timeout/threshold was reached.
Amp-Thread-ID: https://ampcode.com/threads/T-019b1186-164e-740c-96ab-856f64ee6bee
Co-authored-by: Amp <amp@ampcode.com>
Add package and constructor documentation for AI Studio, Antigravity,
Gemini CLI, Gemini API, and Vertex executors to describe their roles and
inputs.
Introduce a shared stream scanner buffer constant in the Gemini API
executor and reuse it in Gemini CLI and Vertex streaming code so stream
handling uses a consistent configuration.
Update Refresh implementations for AI Studio, Gemini CLI, Gemini API
(API key), and Vertex executors to short‑circuit and simply return the
incoming auth object, while keeping Antigravity token renewal as the
only executor that performs OAuth refresh.
Remove OAuth2-based token refresh logic and related dependencies from
the Gemini API executor, since it now operates strictly with API key
credentials.
Align thinking suffix handling on a single bracket-style marker.
NormalizeThinkingModel strips a terminal `[value]` segment from
model identifiers and turns it into either a thinking budget (for
numeric values) or a reasoning effort hint (for strings). Emission
of `ThinkingIncludeThoughtsMetadataKey` is removed.
Executor helpers and the example config are updated so their
comments reference the new `[value]` suffix format instead of the
legacy dash variants.
BREAKING CHANGE: dash-based thinking suffixes (`-thinking`,
`-thinking-N`, `-reasoning`, `-nothinking`) are no longer parsed
for thinking metadata; only `[value]` annotations are recognized.
NormalizeThinkingModel now checks ModelSupportsThinking before removing
"-thinking" or "-thinking-<ver>", avoiding accidental parsing of model
names where the suffix is part of the official id (e.g., kimi-k2-thinking,
qwen3-235b-a22b-thinking-2507).
The registry adds ThinkingSupport metadata for several models and
propagates it via ModelInfo (e.g., kimi-k2-thinking, deepseek-r1,
qwen3-235b-a22b-thinking-2507, minimax-m2), enabling accurate detection
of thinking-capable models and correcting base model inference.
Add fallback parsing for quota reset delay when RetryInfo is not present:
- Try ErrorInfo.metadata.quotaResetDelay (e.g., "373.801628ms")
- Parse from error.message "Your quota will reset after Xs."
This ensures proper cooldown timing for rate-limited requests.
🤖 Generated with [Claude Code](https://claude.com/claude-code)
Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>
## Changes Overview
This commit includes multiple improvements to the Kiro executor for better
stability, API compatibility, and code quality.
## Detailed Changes
### 1. Output Token Calculation Improvement (lines 317-330)
- Replace simple len(content)/4 estimation with tiktoken-based calculation
- Add fallback to character count estimation if tiktoken fails
- Improves token counting accuracy for usage tracking
### 2. Stream Handler Panic Recovery (lines 528-533)
- Add defer/recover block in streamToChannel goroutine
- Prevents single request crashes from affecting the entire service
### 3. Struct Field Reordering (lines 670-673)
- Reorder kiroToolResult struct fields: Content, Status, ToolUseID
- Ensures consistency with API expectations
### 4. Message Merging Function (lines 778-780, 2356-2483)
- Add mergeAdjacentMessages() to combine consecutive messages with same role
- Add helper functions: mergeMessageContent(), blockToMap(), createMergedMessage()
- Required by Kiro API which doesn't allow adjacent messages from same role
### 5. Empty Content Handling (lines 791-800)
- Add default content for empty history messages
- User messages with tool results: "Tool results provided."
- User messages without tool results: "Continue"
### 6. Assistant Last Message Handling (lines 811-830)
- Detect when last message is from assistant
- Create synthetic "Continue" user message to satisfy Kiro API requirements
- Kiro API requires currentMessage to be userInputMessage type
### 7. Duplicate Content Event Detection (lines 1650-1660)
- Track lastContentEvent to detect duplicate streaming events
- Skip redundant events to prevent duplicate content in responses
- Based on AIClient-2-API implementation for Kiro
### 8. Streaming Token Calculation Enhancement (lines 1785-1817)
- Add accumulatedContent buffer for streaming token calculation
- Use tiktoken for accurate output token counting during streaming
- Add fallback to character count estimation with proper logging
### 9. JSON Repair Enhancement (lines 2665-2818)
- Implement conservative JSON repair strategy
- First try to parse JSON directly - if valid, return unchanged
- Add bracket balancing detection and repair
- Only repair when necessary to avoid corrupting valid JSON
- Validate repaired JSON before returning
### 10. HELIOS_CHK Filtering Removal (lines 2500-2504, 3004-3039)
- Remove filterHeliosDebugInfo function
- Remove heliosDebugPattern regex
- HELIOS_CHK fields now handled by client-side processing
### 11. Comment Translation
- Translate Chinese comments to English for code consistency
- Affected areas: token calculation, buffer handling, message processing
- Added support for parsing and normalizing dynamic thinking model suffixes.
- Centralized budget resolution across executors and payload helpers.
- Retired legacy Gemini-specific thinking handlers in favor of unified logic.
- Updated executors to use metadata-based thinking configuration.
- Added `ResolveOriginalModel` utility for resolving normalized upstream models using request metadata.
- Updated executors (Gemini, Codex, iFlow, OpenAI, Qwen) to incorporate upstream model resolution and substitute model values in payloads and request URLs.
- Ensured fallbacks handle cases with missing or malformed metadata to derive models robustly.
- Refactored upstream model resolution to dynamically incorporate metadata for selecting and normalizing models.
- Improved handling of thinking configurations and model overrides in executors.
- Removed hardcoded thinking model entries and migrated logic to metadata-based resolution.
- Updated payload mutations to always include the resolved model.
* feat(api): add comprehensive ampcode management endpoints
Add new REST API endpoints under /v0/management/ampcode for managing
ampcode configuration including upstream URL, API key, localhost
restriction, model mappings, and force model mappings settings.
- Move force-model-mappings from config_basic to config_lists
- Add GET/PUT/PATCH/DELETE endpoints for all ampcode settings
- Support model mapping CRUD with upsert (PATCH) capability
- Add comprehensive test coverage for all ampcode endpoints
* refactor(api): simplify request body parsing in ampcode handlers
* feat(logging): add upstream API request/response capture to streaming logs
* style(logging): remove redundant separator line from response section
* feat(antigravity): enforce thinking budget limits for Claude models
* refactor(logging): remove unused variable in `ensureAttempt` and redundant function call
---------
Co-authored-by: hkfires <10558748+hkfires@users.noreply.github.com>
Add new REST API endpoints under /v0/management/ampcode for managing
ampcode configuration including upstream URL, API key, localhost
restriction, model mappings, and force model mappings settings.
- Move force-model-mappings from config_basic to config_lists
- Add GET/PUT/PATCH/DELETE endpoints for all ampcode settings
- Support model mapping CRUD with upsert (PATCH) capability
- Add comprehensive test coverage for all ampcode endpoints
Add a configuration option to control whether model mappings take
precedence over local API keys for Amp CLI requests.
- Add PrioritizeModelMappings field to AmpCode config struct
- When false (default): Local API keys take precedence (original behavior)
- When true: Model mappings take precedence over local API keys
- Add management API endpoints GET/PUT /prioritize-model-mappings
This allows users who want mapping priority to enable it explicitly
while preserving backward compatibility.
Config example:
ampcode:
model-mappings:
- from: claude-opus-4-5-20251101
to: gemini-claude-opus-4-5-thinking
prioritize-model-mappings: true
Skip text content blocks that are empty or contain only whitespace
when translating Claude messages to OpenAI format. This fixes GLM-4.6
and other strict OpenAI-compatible providers that reject empty text
with error 'text cannot be empty'.
Simplifies routing logic by delegating all provider/mapping/proxy
decisions to FallbackHandler. Previously, the route checked for
provider/mapping availability before calling the handler, then
FallbackHandler performed the same checks again.
Changes:
- Remove model extraction and provider checking from route (lines 182-201)
- Route now only checks if request is POST with /models/ path
- FallbackHandler handles provider -> mapping -> proxy fallback
- Remove unused internal/util import
Benefits:
- Eliminates duplicate checks (addresses PR review feedback #2)
- Centralizes all provider/mapping logic in FallbackHandler
- Reduces routing code by ~20 lines
- Aligns with how other /api/provider routes work
Performance: No impact (checks still happen once in FallbackHandler)
- Change createGeminiBridgeHandler to accept gin.HandlerFunc instead of *gemini.GeminiAPIHandler
This allows tests to inject mock handlers instead of duplicating bridge logic
- Replace magic number 8 with len(modelsPrefix) for better maintainability
- Remove redundant test case that doesn't test edge case in production
- Update routes.go to pass geminiHandlers.GeminiHandler directly
Addresses PR review feedback on test architecture and code clarity.
Amp-Thread-ID: https://ampcode.com/threads/T-1ae2c691-e434-4b99-a49a-10cabd3544db
Gemini handler extracts model from URL path, not JSON body, so
rewriting the request body alone wasn't sufficient for model mapping.
- Add MappedModelContextKey constant for context passing
- Update routes.go to use NewFallbackHandlerWithMapper
- Add check for valid mapping before routing to local handler
- Add tests for gemini bridge model mapping
- Remove custom stringContains and findSubstring helper functions
- Use standard library strings.Contains for better maintainability
- No functional change, just cleaner code
Addresses Gemini Code Assist review feedback
- Added setKiroIncognitoMode() helper function to handle Kiro auth incognito mode setting
- Replaced 3 duplicate code blocks (21 lines) with single function calls (3 lines)
- Kiro auth defaults to incognito mode for multi-account support
- Users can override with --incognito or --no-incognito flags
This addresses the code duplication noted in PR #1 review.
- Add --kiro-aws-login flag for AWS Builder ID device code flow
- Add DoKiroAWSLogin function for AWS SSO OIDC authentication
- Complete Kiro integration with AWS, Google OAuth, and social auth
- Add kiro executor, translator, and SDK components
- Update browser support for Kiro authentication flows
- Added `gpt_5_1_prompt.md` and `gpt_5_codex_prompt.md` to document Codex instruction guidelines.
- These detail the behavior, constraints, and execution policies for GPT-5-based Codex agents in the CLI environment.
* feat(config): add pruning of stale YAML mapping keys during config save
* Revert watcher.go in "fix: enable hot reload for amp-model-mappings config"
- Renamed `generationConfig.responseSchema` to `generationConfig.responseJsonSchema` in Gemini request transformation to align with updated schema expectations.
- Removed `vertex-compat` executor and related configuration.
- Consolidated Vertex compatibility checks into `vertex` handling with `apikey`-based model resolution.
- Streamlined model generation logic for Vertex API key entries.
- Store ampModule in Server struct to access it during config updates
- Call ampModule.OnConfigUpdated() in UpdateClients() for hot reload
- Watch config directory instead of file to handle atomic saves (vim, VSCode, etc.)
- Improve config file event detection with basename matching
- Add diagnostic logging for config reload tracing
Extract applyThinkingMetadata and applyThinkingMetadataCLI helpers to
payload_helpers.go and use them across all four Gemini-based executors:
- gemini_executor.go (Execute, ExecuteStream, CountTokens)
- gemini_cli_executor.go (Execute, ExecuteStream, CountTokens)
- aistudio_executor.go (translateRequest)
- antigravity_executor.go (Execute, ExecuteStream)
This eliminates code duplication introduced in the -reasoning suffix PR
and centralizes the thinking config application logic.
Net reduction: 28 lines of code.
- Updated README files to reflect the new 'Plus' variant with detailed third-party provider support information.
- Adjusted Dockerfile, `docker-compose.yml`, and build configurations to align with the `CLIProxyAPIPlus` naming convention.
- Added information about GitHub Copilot OAuth integration contributed by the community.
- Updated Docker and release workflows to use `cli-proxy-api-plus` as the Docker repository name and adjusted tag generation logic.
- Renamed artifacts in Goreleaser configuration to align with the new 'plus' variant naming convention.
- Updated release and Docker workflows to ensure the `-plus` suffix is added to build versions when missing.
- Adjusted Goreleaser configuration to include `-plus` suffix in `main.Version` during build process.
Adds support for the `-reasoning` model name suffix which enables
thinking/reasoning mode with dynamic budget. This allows clients to
request reasoning-enabled inference using model names like
`gemini-2.5-flash-reasoning` without explicit configuration.
The suffix is normalized to the base model (e.g., gemini-2.5-flash)
with thinkingBudget=-1 (dynamic) and include_thoughts=true.
Follows the existing pattern established by -nothinking and
-thinking-N suffixes.
- Updated `usage_helpers.go` to call `EnsureIndex()` for proper index assignment in reporter initialization.
- Adjusted `auth/manager.go` to assign auth indices inside a locked section when they are unassigned, ensuring thread safety and consistency.
AMP CLI requests /threads.rss at the root level, but the AMP module
only registered routes under /api/*. This caused a 404 error during
AMP CLI startup.
Add the missing root-level route with the same security middleware
(noCORS, optional localhost restriction) as other management routes.
- Add model mapping feature to README.md Amp CLI section
- Add detailed Model Mapping Configuration section to amp-cli-integration.md
- Update architecture diagram to show model mapping flow
- Update Model Fallback Behavior to include mapping step
- Add Table of Contents entry for model mapping
- Add AmpModelMapping config to route models like 'claude-opus-4.5' to 'claude-sonnet-4'
- Add ModelMapper interface and DefaultModelMapper implementation with hot-reload support
- Enhance FallbackHandler to apply model mappings before falling back to ampcode.com
- Add structured logging for routing decisions (local provider, mapping, amp credits)
- Update config.example.yaml with amp-model-mappings documentation
- Add Unwrap() to AuthenticationError for proper error chain handling with errors.Is/As
- Extract hardcoded header values to constants for maintainability
- Replace verbose status code checks with isHTTPSuccess() helper
- Remove unused ExtractBearerToken() and BuildModelsURL() functions
- Make buildChatCompletionURL() private (only used internally)
- Remove unused 'strings' import
Add complete GitHub Copilot support including:
- Device flow OAuth authentication via GitHub's official client ID
- Token management with automatic caching (25 min TTL)
- OpenAI-compatible API executor for api.githubcopilot.com
- 16 model definitions (GPT-5 variants, Claude variants, Gemini, Grok, Raptor)
- CLI login command via -github-copilot-login flag
- SDK authenticator and refresh registry integration
Enables users to authenticate with their GitHub Copilot subscription and
use it as a backend provider alongside existing providers.
- Updated `antigravity_openai_request.go` to process non-JSON outputs gracefully by verifying and distinguishing between JSON and plain string formats.
- Ensured proper assignment of parsed or raw response to `functionResponse`.
- Added `ContextLength` field with a value of 200,000 to all applicable Claude model definitions.
- Standardized `MaxCompletionTokens` values across models for consistency and alignment.
**fix(translator): add support for "xhigh" reasoning effort in OpenAI responses**
- Updated handling in `openai_openai-responses_request.go` to include the new "xhigh" reasoning effort level.
Fixes an issue where Claude thinking models would return 400 errors when
the thinking.budget_tokens was greater than or equal to max_tokens.
Changes:
- Add MaxCompletionTokens: 128000 to all Claude thinking model definitions
- Add ensureMaxTokensForThinking() function in claude_executor.go that:
- Checks if thinking is enabled with a budget_tokens value
- Looks up the model's MaxCompletionTokens from the registry
- Ensures max_tokens is set to at least the model's MaxCompletionTokens
- Falls back to budget_tokens + 4000 buffer if registry lookup fails
This ensures Anthropic API constraint (max_tokens > thinking.budget_tokens)
is always satisfied when using extended thinking features.
Fixes: #339🤖 Generated with [Claude Code](https://claude.com/claude-code)
Co-Authored-By: Claude <noreply@anthropic.com>
- Implemented logic to pair consecutive function calls and their outputs, ensuring proper sequencing for processing.
- Adjusted `gemini_openai-responses_request.go` to normalize message structures and maintain expected flow.
- Updated handling of `output` in `gemini_openai-responses_request.go` to use `.Str` instead of `.Raw` when parsing non-JSON string outputs.
- Added checks to distinguish between JSON and non-JSON `output` types for accurate `functionResponse` construction.
- Updated handling of `output` in `gemini_openai-responses_request.go` to use `.Raw` instead of `.String` for preserving original JSON encoding.
- Ensured proper setting of raw JSON output when constructing `functionResponse`.
- Updated handling of `thoughtSignature` across all translator modules to retain other content payloads if present.
- Adjusted logic for `thought_signature` and `inline_data` keys for consistent processing.
- Added new `Gemini 3 Pro Image Preview` model with detailed metadata and configuration.
- Removed outdated `Claude Sonnet 4.5 Thinking` model definition for cleanup and relevance.
**feat(handlers, executor): add Gemini 3 Pro Preview support and refine Claude system instructions**
- Added support for the new "Gemini 3 Pro Preview" action in Gemini handlers, including detailed metadata and configuration.
- Removed redundant `cache_control` field from Claude system instructions for cleaner payload structure.
**fix(executor): replace redundant commented code with `checkSystemInstructions` helper**
- Replaced commented-out `sjson.SetRawBytes` lines with the new `checkSystemInstructions` function.
- Centralized system instruction handling for better code clarity and reuse.
- Ensured consistent logic for managing `system` field across Claude executor flows.
- Commented out multiple instances of `sjson.SetRawBytes` for setting `system` key to Claude instructions as they are redundant.
- Code cleanup to improve clarity and maintainability without affecting functionality.
- Add Claude thinking model definitions (sonnet-4-5-thinking, opus-4-5-thinking variants)
- Add Thinking support for antigravity models with -thinking suffix
- Add injectThinkingConfig() for automatic thinking budget based on model suffix
- Add resolveUpstreamModel() mappings for thinking variants to actual Claude models
- Add extractAndRemoveBetas() to convert betas array to anthropic-beta header
- Update applyClaudeHeaders() to merge custom betas from request body
Closes#324
- Introduced `appendAPIResponse` helper to preserve and append data to existing API responses.
- Ensured newline inclusion when appending, if necessary.
- Improved `nil` and data type checks for response handling.
- Updated middleware to skip request logging for `GET` requests.
- Added additional metadata fields (`Name`, `Description`, `DisplayName`, `Version`) to `ModelInfo` struct initialization for better model representation.
- Removed unnecessary whitespace in the code.
- Ensured `functionDeclarations` key renaming only occurs if the key exists in Gemini tools processing.
- Prevented unnecessary JSON reassignment when the target key is absent.
- Fixed parameter key renaming to correctly handle `functionDeclarations` and `parametersJsonSchema` in Gemini tools.
- Resolved potential overwriting issue by reassigning JSON strings after each key rename.
- Introduced `TLSConfig` to support HTTPS configurations, including enabling TLS, specifying certificate and key files.
- Updated HTTP server logic to handle HTTPS mode when TLS is enabled.
- Enhanced `config.example.yaml` with TLS settings example.
- Adjusted internal URL generation to respect protocol based on TLS state.
- Fixed improper handling of `indexAssigned` and `Index` during auth reassignment.
- Ensured `EnsureIndex` is invoked after validating existing auth entries.
**feat(retry): add configurable retry logic with cooldown support**
- Introduced `max-retry-interval` configuration for cooldown durations between retries.
- Added `SetRetryConfig` in `Manager` to handle retry attempts and cooldown intervals.
- Enhanced provider execution logic to include retry attempts, cooldown management, and dynamic wait periods.
- Updated API endpoints and YAML configuration to support `max-retry-interval`.
- Introduced `logOnErrorOnly` mode to enable logging only for error responses when request logging is disabled.
- Added endpoints to list and download error logs (`/request-error-logs`).
- Implemented error log file cleanup to retain only the newest 10 logs.
- Refactored `ResponseWriterWrapper` to support forced logging for error responses.
- Enhanced middleware to capture data for upstream error persistence.
- Improved log file naming and error log filename generation.
- Restored `thoughtSignature` validator bypass for model-specific parts in Gemini content processing.
- Removed redundant logic from the `executor` for cleaner handling.
Remove generationConfig.maxOutputTokens, generationConfig.responseMimeType and generationConfig.responseJsonSchema from the Gemini payload in translateRequest so we no longer send unsupported or conflicting response configuration fields. This lets the backend or caller control response formatting and output limits and helps prevent potential API errors caused by these keys.
- Updated `defaultAntigravityAgent` to version `1.11.5`.
- Adjusted const value formatting for improved readability.
**feat(executor): introduce fallback mechanism for Antigravity base URLs**
- Added retry logic with fallback order for Antigravity base URLs to handle request errors and rate limits.
- Refactored base URL handling with `antigravityBaseURLFallbackOrder` and related utilities.
- Enhanced error handling in non-streaming and streaming requests with retry support and improved metadata reporting.
- Updated `buildRequest` to support dynamic base URL assignment.
- Improved Antigravity executor to handle `thinkingConfig` adjustments and default `thinkingBudget` when `thinkingLevel` is removed.
- Updated translator response handling to set default values for output token counts when specific token data is missing.
- Introduced `modelName2Alias` and `alias2ModelName` functions for mapping between model names and aliases.
- Improved Antigravity payload transformation to include alias-to-model name conversion.
- Enhanced processing for Claude Sonnet models to adjust template parameters based on schema presence.
- Added `authFileUnchanged` to skip reloads for unchanged files based on SHA256 hash comparisons.
- Introduced `isKnownAuthFile` to verify known files before handling removal events.
- Improved event processing in `handleEvent` to reduce unnecessary reloads and enhance performance.
- Refactored response translation logic to support mixed content types (`input_text`, `output_text`, `input_image`) with better role assignments and part handling.
- Added image processing logic for embedding inline data with MIME type and base64 encoded content.
- Updated Antigravity request conversion to replace Gemini CLI references for consistency.
**feat(executor): enhance WebSocket error handling and metadata logging**
- Added handling for stream closure before start with appropriate error recording.
- Improved metadata logging for non-OK HTTP status codes in WebSocket responses.
- Consolidated event processing logic with `processEvent` for better error handling and payload management.
- Refactored stream initialization to include the first event handling for smoother execution flow.
- Added new endpoint `/antigravity-auth-url` to initiate Antigravity authentication.
- Implemented `RequestAntigravityToken` to manage the OAuth flow, including token exchange and user info retrieval.
- Introduced `.oauth-antigravity` temporary file handling for state and code management.
- Added `sanitizeAntigravityFileName` utility for safe token file names based on user email.
- Registered `/antigravity/callback` endpoint for OAuth redirects.
- Introduced request and response translation functions to enable compatibility between OpenAI Chat Completions API and Antigravity.
- Registered translation utilities for both streaming and non-streaming scenarios.
- Added support for reasoning content, tool calls, and metadata handling.
- Established request normalization and embedding for Antigravity-compatible payloads.
- Added new fields to `Params` struct for better tracking of finish reasons, usage metadata, and tool usage.
- Refactored handling of response transitions, final events, and state-driven logic in `ConvertAntigravityResponseToClaude`.
- Introduced `appendFinalEvents` and `resolveStopReason` helper functions for cleaner separation of concerns.
- Added `TotalTokenCount` field to `Params` struct for enhanced token tracking.
- Updated token count calculations to fallback on `TotalTokenCount` when specific counts are missing.
- Introduced `hasNonZeroUsageMetadata` function to validate presence of token data in `usage_metadata`.
- Replaced `http.DefaultClient` with a configurable `http.Client` instance for Antigravity OAuth flow methods.
- Updated `exchangeAntigravityCode` and `fetchAntigravityUserInfo` to accept `httpClient` as a parameter.
- Added `util.SetProxy` usage to initialize the `httpClient` with proxy support.
- Implemented OAuth login flow for the Antigravity provider in `auth/antigravity.go`.
- Added `AntigravityExecutor` for handling requests and streaming via Antigravity APIs.
- Created `antigravity_login.go` command for triggering Antigravity authentication.
- Introduced OpenAI-to-Antigravity translation logic in `translator/antigravity/openai/chat-completions`.
**refactor(translator, executor): update Gemini CLI response translation and add Antigravity payload customization**
- Renamed Gemini CLI translation methods to align with response handling (`ConvertGeminiCliResponseToGemini` and `ConvertGeminiCliResponseToGeminiNonStream`).
- Updated `init.go` to reflect these method changes.
- Introduced `geminiToAntigravity` function to embed metadata (`model`, `userAgent`, `project`, etc.) into Antigravity payloads.
- Added random project, request, and session ID generators for enhanced tracking.
- Streamlined `buildRequest` to use `geminiToAntigravity` transformation before request execution.
**feat(executor): add thinking level to budget conversion utility**
- Introduced `ConvertThinkingLevelToBudget` to map thinking level ("high"/"low") to corresponding budget values.
- Applied the utility in `aistudio_executor.go` before stripping unsupported configs.
- Updated dependencies to include `tidwall/gjson` for JSON parsing.
- Added `shouldLogRequest` helper to simplify path-based request logging logic.
- Updated middleware to skip management endpoints for improved security.
- Introduced an explicit `nil` logger check for minimal overhead.
- Updated dependencies in `go.mod`.
**feat(auth): add handling for 404 response with retry logic**
- Introduced support for 404 `not_found` status with a 12-hour backoff period.
- Updated `manager.go` to align state and status messages for 404 scenarios.
**refactor(translator): comment out debug logging in Gemini responses request**
- Updated `README_CN.md` to introduce Amp CLI and IDE support.
- Added detailed integration guide in `docs/amp-cli-integration_CN.md`.
- Covered setup, configuration, OAuth, security, and usage of Amp CLI with Google/ChatGPT/Claude subscriptions.
Fix critical security vulnerability in amp-restrict-management-to-localhost
feature where attackers could bypass localhost restriction by spoofing
X-Forwarded-For headers.
Changes:
- Use RemoteAddr (actual TCP connection) instead of ClientIP() in
localhostOnlyMiddleware to prevent header spoofing attacks
- Add comprehensive test coverage for spoofing prevention (6 test cases)
- Update documentation with reverse proxy deployment guidance and
limitations of the RemoteAddr approach
The fix prevents attacks like:
curl -H "X-Forwarded-For: 127.0.0.1" https://server/api/user
Trade-off: Users behind reverse proxies will need to disable the feature
and use alternative security measures (firewall rules, proxy ACLs).
Addresses security review feedback from PR #287.
AMP CLI sends Gemini requests to non-standard paths that were being
directly proxied to ampcode.com without checking for local OAuth.
This fix adds:
- GeminiBridge handler to transform AMP CLI paths to standard format
- Enhanced model extraction from AMP's /publishers/google/models/* paths
- FallbackHandler wrapper to check for local OAuth before proxying
Flow:
- If user has local Google OAuth → use it (free tier)
- If no local OAuth → fallback to ampcode.com (charges credits)
Fixes issue where gemini-3-pro-preview requests always charged AMP
credits even when user had valid Google Cloud OAuth configured.
- Update README.md to present Amp CLI as standard feature, not fork differentiator
- Remove USING_WITH_FACTORY_AND_AMP.md (fork-specific, Factory docs live upstream)
- Add comprehensive docs/amp-cli-integration.md with setup, config, troubleshooting
- Eliminate fork justification messaging throughout documentation
- Prepare Amp CLI integration for upstream merge consideration
This positions Amp CLI support as a natural extension of CLIProxyAPI's
multi-client architecture rather than a fork-specific feature.
- Replace deprecated GPT-5 and GPT-5-Codex with GPT-5.1 family
- Add explicit reasoning effort levels (low/medium/high)
- Remove duplicate base models (use medium as default)
- GPT-5.1 Codex Mini supports medium/high only (per OpenAI docs)
- Remove older Claude Sonnet 4 (keep 4.5)
- Final config: 11 models (3 Claude + 8 GPT-5.1 variants)
Amp CLI sends 'context-1m-2025-08-07' in Anthropic-Beta header which
requires a special 1M context window subscription. After upstream rebase
to v6.3.7 (commit 38cfbac), CLIProxyAPI now respects client-provided
Anthropic-Beta headers instead of always using defaults.
When users configure local OAuth providers (Claude, etc), requests bypass
the ampcode.com proxy and use their own API subscriptions. These personal
subscriptions typically don't include the 1M context beta feature, causing
'long context beta not available' errors.
Changes:
- Add filterBetaFeatures() helper to strip specific beta features
- Filter context-1m-2025-08-07 in fallback handler when using local providers
- Preserve full headers when proxying to ampcode.com (paid users get all features)
- Add 7 test cases covering all edge cases
This fix is isolated to the Amp module and only affects the local provider
path. Users proxying through ampcode.com are unaffected and receive full
1M context support as part of their paid service.
- add fallback handler that forwards Amp provider requests to ampcode.com when the provider isn’t configured locally
- wrap AMP provider routes with the fallback so requests always have a handler
- share Gemini thinking model normalization helper between core handlers and AMP fallback
Add full Amp CLI support to enable routing AI model requests through the proxy
while maintaining Amp-specific features like thread management, user info, and
telemetry. Includes complete documentation and pull bot configuration.
Features:
- Modular architecture with RouteModule interface for clean integration
- Reverse proxy for Amp management routes (thread/user/meta/ads/telemetry)
- Provider-specific route aliases (/api/provider/{provider}/*)
- Secret management with precedence: config > env > file
- 5-minute secret caching to reduce file I/O
- Automatic gzip decompression for responses
- Proper connection cleanup to prevent leaks
- Localhost-only restriction for management routes (configurable)
- CORS protection for management endpoints
Documentation:
- Complete setup guide (USING_WITH_FACTORY_AND_AMP.md)
- OAuth setup for OpenAI (ChatGPT Plus/Pro) and Anthropic (Claude Pro/Max)
- Factory CLI config examples with all model variants
- Amp CLI/IDE configuration examples
- tmux setup for remote server deployment
- Screenshots and diagrams
Configuration:
- Pull bot disabled for this repo (manual rebase workflow)
- Config fields: AmpUpstreamURL, AmpUpstreamAPIKey, AmpRestrictManagementToLocalhost
- Compatible with upstream DisableCooling and other features
Technical details:
- internal/api/modules/amp/: Complete Amp routing module
- sdk/api/httpx/: HTTP utilities for gzip/transport
- 94.6% test coverage with 34 comprehensive test cases
- Clean integration minimizes merge conflict risk
Security:
- Management routes restricted to localhost by default
- Configurable via amp-restrict-management-to-localhost
- Prevents drive-by browser attacks on user data
This provides a production-ready foundation for Amp CLI integration while
maintaining clean separation from upstream code for easy rebasing.
Amp-Thread-ID: https://ampcode.com/threads/T-9e2befc5-f969-41c6-890c-5b779d58cf18
Fix critical bug where ExecuteStream would create a streaming channel
from a failed (non-2xx) response after exhausting all retries with no
fallback models available.
When retries were exhausted on the last model, the code would break from
the inner loop but fall through to streaming channel creation (line 401),
immediately returning at line 461. This made the error handling code at
lines 464-471 unreachable, causing clients to receive an empty/closed
stream instead of a proper error response.
Solution: Check if httpResp is non-2xx before creating the streaming
channel. If failed, continue the outer loop to reach error handling.
Identified by: codex-bot review
Ref: https://github.com/router-for-me/CLIProxyAPI/pull/280#pullrequestreview-3484560423
Fix critical bug where ExecuteStream would create a streaming channel
using a 429 error response instead of continuing to the next fallback
model after exhausting retries.
When 429 retries were exhausted and a fallback model was available,
the inner retry loop would break but immediately fall through to the
streaming channel creation, attempting to stream from the failed 429
response instead of trying the next model.
Solution: Add shouldContinueToNextModel flag to explicitly skip the
streaming logic and continue the outer model loop when appropriate.
Identified by: codex-bot review
Ref: https://github.com/router-for-me/CLIProxyAPI/pull/280#pullrequestreview-3484479106
Match official Gemini CLI behavior by always sending default
thinkingConfig when client doesn't specify reasoning parameters.
- Set thinkingBudget=-1 (dynamic) for gemini-3-pro-preview
- Set include_thoughts=true to return thinking process
- Apply to both /v1/chat/completions and /v1/responses endpoints
- See: ai-gemini-cli/packages/core/src/config/defaultModelConfigs.ts
Implement Google RetryInfo.retryDelay support for handling 429 rate
limit errors. Retries same model up to 3 times using exact delays
from Google's API before trying fallback models.
- Add parseRetryDelay() to extract Google's retry guidance
- Implement inner retry loop in Execute() and ExecuteStream()
- Context-aware waiting with cancellation support
- Cap delays at 60s maximum for safety
Add gemini-3-pro-preview model to GetGeminiCLIModels() to make it
available for OAuth-based Gemini CLI users, matching the model
already available in AI Studio provider.
Model spec:
- ID: gemini-3-pro-preview
- Version: 3.0
- Input: 1M tokens
- Output: 64K tokens
- Thinking: 128-32K tokens (dynamic)
- Introduced `gpt-5.1-codex-max` variants to model definitions (`low`, `medium`, `high`, `xhigh`).
- Updated executor logic to map effort levels for Codex Max models.
- Added `lastCodexMaxPrompt` processing for `gpt-5.1-codex-max` prompts.
- Defined instructions for `gpt-5.1-codex-max` in a new file: `codex_instructions/gpt-5.1-codex-max_prompt.md`.
Replace the "~<n>" suffix with "_<n>" when generating unique short names in codex translators (Claude, Gemini, OpenAI chat).
This avoids using a special character in identifiers, improving compatibility with downstream APIs while preserving length constraints.
Add logic to avoid exposing credentials that have been removed from disk but still persist in memory. Ensure `runtimeOnly` checks and proper handling of disabled or removed authentication states.
Introduce `migrateLegacyOpenAICompatibilityKeys` to streamline and reuse the normalization of OpenAI compatibility entries. Remove redundant loops and enhance maintainability for compatibility key handling. Add cleanup for legacy `api-keys` in YAML configuration during persistence.
Extract reasoning effort mapping into a reusable function `setReasoningEffortByAlias` to reduce redundancy and improve maintainability. Introduce support for the "gpt-5.1-none" variant in the registry and runtime executor.
Stop advertising and mapping the unsupported gpt-5.1-minimal variant in the model registry and Codex executor, and align bare gpt-5.1 requests to use medium reasoning effort like Codex CLI while preserving minimal for gpt-5.
Expand executor logic to handle GPT-5.1 Codex family and its variants, including reasoning effort configurations for minimal, low, medium, and high levels. Ensure proper mapping of models to payload parameters.
Introduce `PayloadConfig` in the configuration to define default and override rules for modifying payload parameters. Implement `applyPayloadConfig` and `applyPayloadConfigWithRoot` to apply these rules across various executors, ensuring consistent parameter handling for different models and protocols. Update all relevant executors to utilize this functionality.
feat(runtime): add support for GPT-5.1 models and variants
Introduce GPT-5.1 model family, including minimal, low, medium, high, Codex, and Codex Mini variants. Update tokenization and reasoning effort handling to accommodate new models in executor and registry.
Refine `handleEvent` to support additional file system operations (Rename, Remove) for config and auth JSON files. Improve client update/removal logic with atomic file replacement handling and incremental processing for auth changes.
Introduce a new `buildinfo` package to store version, commit, and build date metadata. Update HTTP handlers to include build metadata in response headers and modify initialization to set `buildinfo` values during runtime.
Introduce support for multi-project Gemini CLI logins, including shared and virtual credential management. Enhance runtime, metadata handling, and token updates for better project granularity and consistency across virtual and shared credentials. Extend onboarding to allow activating all available projects.
Extend `resolveUsageSource` to support Vertex projects by extracting and normalizing `project_id` or `project` from the metadata for accurate source resolution.
Extend `vertexAccessToken` to support proxy-aware HTTP clients and update calls accordingly for better configurability. Add `deleteTokenRecord` to handle token cleanup, improving management of authentication files.
- Add 'created' field to model registry for tracking model creation time
- Implement GetFirstAvailableModel() to find the first available model by newest creation timestamp
- Add ResolveAutoModel() utility function to resolve "auto" model name to actual available model
- Update request handler to resolve "auto" model before processing requests
- Ensures automatic model selection when "auto" is specified as model name
This enables dynamic model selection based on availability and creation time, improving the user experience when no specific model is requested.
feat(management): add auth ID normalization and file-based ID resolution
Introduce `authIDForPath` to standardize ID generation from file paths, improving consistency in authentication handling. Update `registerAuthFromFile` and `disableAuth` to utilize normalized IDs, incorporating relative path resolution and file name extraction where applicable.
Enhance error management for file operations and clean up temporary files. Add `NormalizeCommentIndentation` function to ensure YAML comments maintain consistent formatting.
Introduce an endpoint for importing Vertex service account JSON keys and storing them as authentication records. Add handlers for managing WebSocket authentication configuration.
Introduce `scheduleConfigReload` with debounce functionality for config reloads, ensuring efficient handling of frequent changes. Added `stopConfigReloadTimer` for stopping timers during watcher shutdown.
Introduce Vertex AI Gemini integration with support for service account-based authentication, credential storage, and import functionality. Added new executor for Vertex AI requests, including execution and streaming paths, and integrated it into the core manager. Enhanced CLI with `--vertex-import` flag for importing service account keys.
Add `ensurePublished` to guarantee request counting even when usage fields (e.g., tokens) are absent in OpenAI-compatible executor responses, particularly for streaming paths.
Added detailed operational instructions for Codex agents based on GPT-5, covering shell usage, editing constraints, sandboxing policies, and approval mechanisms. Also included comprehensive review process guidelines for flagging and communicating issues effectively.
Adds three new Codex Mini model variants (mini, mini-medium, mini-high)
that map to codex-mini-latest. Codex Mini supports medium and high
reasoning effort levels only (no low/minimal). Base model defaults to
medium reasoning effort.
<a href="https://x.com/Finessse377721"><img src="https://img.shields.io/badge/Follow-%F0%9D%95%8F%2F%40Finessse377721-1c9bf0" alt="Follow on 𝕏"></a>
<a href="https://github.com/Finesssee/ProxyPilot"><img src="https://img.shields.io/github/stars/Finesssee/ProxyPilot.svg?style=social&label=Star%20this%20repo" alt="Star this repo"></a>
</p>
It now also supports OpenAI Codex (GPT models) and Claude Code via OAuth.
<p align="center">
<strong>Stop juggling API keys.</strong> ProxyPilot is a powerful local API proxy that lets you use your existing Claude Code, Codex, Gemini, Kiro, and Qwen subscriptions with any AI coding tool – no separate API keys required.
</p>
So you can use local or multi-account CLI access with OpenAI(include Responses)/Gemini/Claude-compatible clients and SDKs.
<p align="center">
Built in Go, it handles OAuth authentication, token management, and API translation automatically. One server to route them all.
GLM CODING PLAN is a subscription service designed for AI coding, starting at just $3/month. It provides access to their flagship GLM-4.6 model across 10+ popular AI coding tools (Claude Code, Cline, Roo Code, etc.), offering developers top-tier, fast, and stable coding experiences.
---
Get 10% OFF GLM CODING PLAN:https://z.ai/subscribe?ic=8JVLJQFSKB
Browser-based tool to translate SRT subtitles using your Gemini subscription via CLIProxyAPI with automatic validation/error correction - no API keys needed
> If you developed a project based on CLIProxyAPI, please open a PR to add it to this list.
# Import providers (from existing CLI tools)
./proxypilot --kiro-import # Kiro IDE token
# API key providers (prompts for key)
./proxypilot --minimax-login # MiniMax API key
./proxypilot --zhipu-login # Zhipu AI API key
```
OAuth tokens are stored locally and auto-refreshed before expiry.
### Security Defaults (Auth + CORS)
- Proxy requests require API keys by default. To allow unauthenticated access (not recommended), set `allow-unauthenticated: true` in `config.yaml`.
- CORS is enabled for non-management endpoints by default (wildcard `*`). Management endpoints **do not** emit CORS headers unless you explicitly allow origins under `cors.management-allow-origins`.
Example:
```yaml
allow-unauthenticated:false
cors:
allow-origins:
- "http://localhost:5173"
management-allow-origins:
- "http://localhost:5173"
```
### Configure Your Tools
**Claude Code** (`~/.claude/settings.json`):
```json
{
"env":{
"ANTHROPIC_BASE_URL":"http://127.0.0.1:8317",
"ANTHROPIC_AUTH_TOKEN":"your-api-key"
}
}
```
**Codex CLI** (`~/.codex/config.toml`):
```toml
[openai]
api_base_url="http://127.0.0.1:8317"
```
**Factory Droid** (`~/.factory/settings.json`):
```json
{
"customModels":[{
"name":"ProxyPilot",
"baseUrl":"http://127.0.0.1:8317"
}]
}
```
---
## API Endpoints
```
POST /v1/chat/completions # OpenAI Chat Completions
POST /v1/responses # OpenAI Responses API
POST /v1/messages # Anthropic Messages API
GET /v1/models # List available models
GET /healthz # Health check
```
All endpoints auto-translate between formats based on the target provider.
---
## Caching
ProxyPilot includes two caching layers to reduce latency and token usage.
### Response Cache
Caches full API responses for identical requests. Useful for repeated queries during development.
**Config** (`config.yaml`):
```yaml
response-cache:
enabled:true# Default: false
max-size:1000# Max entries (default: 1000)
max-bytes:0# Optional total cache size cap in bytes
ttl-seconds:300# Cache TTL (default: 300 = 5 min)
exclude-models:# Models to skip (supports wildcards)
- "*-thinking"
- "o1-*"
```
### Prompt Cache
Synthetic prompt caching for providers without native support. Tracks repeated system prompts and estimates token savings.
**Config** (`config.yaml`):
```yaml
prompt-cache:
enabled:true# Default: false
max-size:500# Max entries (default: 500)
max-bytes:0# Optional total cache size cap in bytes
ProxyPilot automatically backs up the previous version during updates and can restore it if something goes wrong.
### Automatic Recovery
- **Crash Detection** - Tracks rapid restarts within a 30-second window
- **Auto-Rollback** - After 3 rapid failures, automatically restores previous version
- **Health Marking** - After 30 seconds of stable operation, clears the failure counter
### Manual Rollback
From the tray menu:
- **Settings → Rollback to Previous Version** - Restore the previous version
### How It Works
1. During update, current binary is saved as `.old` backup
2. Rollback metadata stored in `%APPDATA%/ProxyPilot/updates/`
3. On crash loop detection, previous version is automatically restored
4. After successful startup (30s), backup can be cleaned up
---
## CLI Tools
| Binary | Description |
|--------|-------------|
| `proxypilot` | Main CLI with server and config switching |
| `proxypilot-tray` | System tray app |
---
## Tool Integrations
Works with these AI coding tools:
- **Claude Code** - Auto-configure via settings.json
- **Codex CLI** - Auto-configure via config.toml
- **Factory Droid** - Auto-configure via settings.json
- **Cursor IDE** - Manual endpoint configuration
- **Continue** - Manual endpoint configuration
---
## Requirements
- macOS, Linux, or Windows
- Go 1.24+ (for building from source)
---
## Author
ProxyPilot is developed and maintained by [@Finesssee](https://github.com/Finesssee). Some contributors shown in the git history are from upstream [CLIProxyAPI](https://github.com/router-for-me/CLIProxyAPI) merges. Direct contributors to ProxyPilot will be listed here as the project grows.
---
## Credits
ProxyPilot builds upon excellent work from the open-source community:
- **[CLIProxyAPI](https://github.com/router-for-me/CLIProxyAPI)** - The original unified proxy server that inspired this project
- **[VibeProxy](https://github.com/automazeio/vibeproxy)** - Native macOS menu bar app showcasing clean proxy UX
Long-context features are inspired by research from the AI community:
- **[Factory.ai](https://factory.ai/news/evaluating-compression)** - Context compression techniques for long-running coding agents
- **[Anthropic](https://www.anthropic.com/engineering/effective-harnesses-for-long-running-agents)** - Effective harnesses for long-running agents
Special thanks to these teams for sharing their work and insights.
---
## License
This project is licensed under the MIT License - see the [LICENSE](LICENSE) file for details.
The following versions of ProxyPilot are currently supported with security updates:
| Version | Supported |
| ------- | ------------------ |
| latest | :white_check_mark: |
| < 1.0 | :x: |
## Reporting a Vulnerability
We take the security of ProxyPilot seriously. If you discover a security vulnerability, please report it responsibly.
### How to Report
You can report a vulnerability through one of the following methods:
1.**GitHub Security Advisory (Preferred)**: Open a private security advisory through [GitHub Security Advisories](https://github.com/smat-dev/ProxyPilot/security/advisories/new). This allows for private discussion and coordination.
2.**Email**: Send an email to the project maintainers with details of the vulnerability. Please include:
- A description of the vulnerability
- Steps to reproduce the issue
- Potential impact of the vulnerability
- Any suggested fixes (if available)
### What to Expect
After submitting a vulnerability report:
- **Initial Response**: You will receive an acknowledgment within **48 hours** confirming we have received your report.
- **Assessment**: We will investigate and assess the vulnerability within **7 days**.
- **Resolution Timeline**: We aim to release a fix for confirmed vulnerabilities within **30 days**, depending on complexity.
- **Disclosure**: Once a fix is released, we will coordinate with you on public disclosure timing.
We will keep you informed throughout the process and credit you for the discovery (unless you prefer to remain anonymous).
## Responsible Disclosure Guidelines
To protect our users, we ask that you:
- **DO NOT** disclose the vulnerability publicly until a fix has been released.
- **DO NOT** exploit the vulnerability beyond what is necessary to demonstrate the issue.
- **DO NOT** access, modify, or delete data belonging to other users.
- **DO** provide us with reasonable time to address the issue before any public disclosure.
- **DO** act in good faith to avoid privacy violations, data destruction, or service disruption.
We are committed to working with security researchers and will not take legal action against individuals who discover and report vulnerabilities responsibly.
## Security Best Practices for Users
- Always use the latest version of ProxyPilot.
- Review and understand the proxy configurations before deployment.
- Use secure connections (HTTPS) where applicable.
- Regularly review access logs for suspicious activity.
Thank you for helping keep ProxyPilot and its users safe.
This is a static, copy-paste demo showing how the middleware rewrites requests.
## Responses API (Codex CLI style)
Input:
```json
{
"model":"gpt-5-codex",
"instructions":"Follow the user's request.",
"input":[
{"role":"user","content":[{"type":"input_text","text":"Build a todo app"}]}
]
}
```
Injected (INITIALIZER mode):
```json
{
"model":"gpt-5-codex",
"instructions":"Follow the user's request.\n\n<harness_mode>INITIALIZER</harness_mode>\nYou are the **Initializer Agent** in a Long-Running Agent Harness.\nYour ONLY goal is to set up the project environment for future Coding Agents.\n\n**CRITICAL INSTRUCTIONS**:\n1. **Analyze the User's Request**: Understand what feature or app needs to be built.\n2. **Create 'feature_list.json'**:\n - Write a JSON file listing ALL features/requirements as separate items.\n - Mark them all as \"passes\": false.\n - Format:\n Request: \"Build a todo app\"\n File Content:\n [\n { \"category\": \"core\", \"description\": \"User can add a todo\", \"steps\": [\"Open app\", \"Enter text\", \"Press enter\", \"See new item\"], \"passes\": false },\n { \"category\": \"core\", \"description\": \"User can delete a todo\", \"steps\": [\"Open app\", \"Click delete\", \"Item removed\"], \"passes\": false }\n ]\n3. **Create 'claude-progress.txt'**:\n - Create this file with a header \"## Progress Log\".\n - Add an initial entry: \"- [Initializer] Environment setup started.\"\n4. **Create 'init.sh' (Optional)**:\n - If a dev server is needed, create a script to start it (e.g., 'npm run dev').\n5. **Initial Commit**:\n - Initialize git if not present.\n - Commit these foundational files.\n\n**DO NOT** start implementing features yourself. Your job is ONLY to scaffold the harness.",
"input":[
{"role":"user","content":[{"type":"input_text","text":"Build a todo app"}]}
]
}
```
## Chat Completions (Claude / OAI chat)
Input:
```json
{
"model":"claude-sonnet-4-5-thinking",
"messages":[
{"role":"system","content":"You are a helpful assistant."},
{"role":"user","content":"Set up the repo"}
]
}
```
Injected (INITIALIZER mode):
```json
{
"model":"claude-sonnet-4-5-thinking",
"messages":[
{
"role":"system",
"content":"You are a helpful assistant.\n\n<harness_mode>INITIALIZER</harness_mode>\nYou are the **Initializer Agent** in a Long-Running Agent Harness.\nYour ONLY goal is to set up the project environment for future Coding Agents.\n\n**CRITICAL INSTRUCTIONS**:\n... (same initializer prompt) ..."
},
{"role":"user","content":"Set up the repo"}
]
}
```
## CODING mode trigger
If `feature_list.json` or `claude-progress.txt` exists under `harness-root-dir`,
the middleware switches to CODING mode and injects the Coding prompt instead.
When all credentials are blocked for a model, CLIProxyAPI now reports more informative HTTP statuses:
- Quota/rate cooldown: `429` with `code: model_cooldown` and `Retry-After`
- Temporarily blocked credentials (non-quota): `503` with `code: auth_unavailable`, `Retry-After`, and a JSON error body including blocked counts and recent upstream HTTP statuses (when known)
**proxypilot** is the unified command-line interface for the ProxyPilot local AI proxy. It provides both **server** functionality (running the proxy server) and **switch** functionality (toggling AI agents between proxy mode and native mode).
## Installation
### Via bun (recommended)
```bash
bun install -g proxypilot
```
### Via npm
```bash
npm install -g proxypilot
```
### Via Go
```bash
go install github.com/Finesssee/ProxyPilot/cmd/server@latest
```
### Manual Download
Download pre-built binaries from [GitHub Releases](https://github.com/Finesssee/ProxyPilot/releases).
Available binaries:
-`proxypilot-darwin-amd64` - macOS Intel
-`proxypilot-darwin-arm64` - macOS Apple Silicon
-`proxypilot-linux-amd64` - Linux x64
-`proxypilot-linux-arm64` - Linux ARM64
-`proxypilot-windows-amd64.exe` - Windows x64
## Server Mode
Run ProxyPilot as a local proxy server:
```bash
proxypilot # Start server on default port (8317)
proxypilot -config /path/to/config # Use custom config file
```
### Server Features
- Routes AI requests through configured providers
- Automatic model aliasing and translation
- OAuth login helpers for various providers
- Agent configuration management
### Login Commands
```bash
proxypilot --login # Login to Google Account (Gemini)
proxypilot --codex-login # Login to Codex using OAuth
proxypilot --claude-login # Login to Claude using OAuth
proxypilot --qwen-login # Login to Qwen using OAuth
proxypilot --antigravity-login # Login to Antigravity using OAuth
proxypilot --kiro-login # Login to Kiro using Google OAuth
proxypilot --kiro-aws-login # Login to Kiro using AWS Builder ID
proxypilot --minimax-login # Add MiniMax API key
proxypilot --zhipu-login # Add Zhipu AI API key
```
### Login Options
```bash
--no-browser # Don't open browser automatically for OAuth
--incognito # Use incognito/private mode (useful for multiple accounts)
--no-incognito # Force disable incognito mode
```
### Import Commands
```bash
proxypilot --vertex-import <file> # Import Vertex service account key JSON
proxypilot --kiro-import # Import Kiro token from Kiro IDE
```
## Agent Configuration
ProxyPilot can automatically configure AI CLI tools to use the proxy.
### Detect Installed Agents
```bash
proxypilot --detect-agents
```
Example output:
```
Detecting installed CLI agents...
Claude Code: [+] Installed
Binary: /usr/local/bin/claude
Config: ~/.claude/settings.json
Codex CLI: [+] Installed
Binary: /usr/local/bin/codex
Config: ~/.codex
Gemini CLI: [-] Not found
```
### Setup Commands
```bash
proxypilot --setup-all # Configure all detected agents
Switch AI agents between proxy mode (through ProxyPilot) and native mode (direct API access). Think of it like `nvm` for Node versions, but for AI agent configurations.
### Usage
```bash
proxypilot --switch <agent> # Show status of specific agent
- Sends `stream: true` but typically uses `Accept: application/json`
- CLIProxyAPI uses a non-stream upstream call and **synthesizes SSE** for reliability.
- Notes:
- Droid attempts to JSON-parse each SSE `data:` line, so `data: [DONE]` must **not** be emitted for this client.
- Routing transparency:
- For localhost requests, the same headers as Codex apply for underlying routing.
- The `/v0/management/routing/preview?model=...` endpoint is the easiest way to debug which provider a `model` will map to before sending a turn from Droid/Factory.
- If body has `"stream": true` and `Accept` includes `text/event-stream`, ProxyPilot streams SSE.
- If `Accept: application/json`**without**`text/event-stream`, ProxyPilot automatically downgrades to non-streaming and returns JSON, even when `stream: true`.
- Notes:
- Responses from localhost include routing headers as above.
- For debugging, you can issue a test request and inspect:
-`X-ProxyPilot-Requested-Model` vs `X-ProxyPilot-Resolved-Model` to see alias/`auto`/thinking normalisation.
-`X-ProxyPilot-Provider-Candidates` and `X-ProxyPilot-Resolved-Provider` to see which upstream(s) and which provider actually ran.
- Returns JSON describing how a given model name will be normalised and which provider candidates will be used.
-`POST /v0/management/routing/preview`
- Accepts a full request JSON body (OpenAI/OpenAIResponses-compatible), extracts the `model` field, and returns the same routing preview for that model.
-`/v0/management/routing/recent`
- Returns a small in-memory list of the last few routed requests (timestamp, path, requested/resolved model, provider + candidates).
- Both endpoints require a management key and are intended for localhost use.
### `gemini-3-*` provider preference
- Primary: `antigravity`
- Fallback: `gemini-cli` (only used if antigravity is unavailable/cooling)
- Rotation: disabled between these two providers to keep the primary stable.
CLIProxyAPI can be used as a local OpenAI-compatible endpoint for Cursor.
## 1) Start CLIProxyAPI
Run the server and note the listening address (default: `http://localhost:8318`).
## 2) Configure Cursor
In Cursor settings:
- **API base URL**: `http://localhost:8318` (recommended)
- Alternatively: `http://localhost:8318/v1` (CLIProxyAPI also provides non-`/v1` aliases for this case)
- **API key**: use one of the keys from `config.yaml` under `api-keys`
- Example: `your-cli-proxy-api-key`
## 3) Model selection
Cursor will query `GET /v1/models` (or `/models` when base URL ends with `/v1`).
Pick any model exposed by your configured providers.
## Notes
-`POST /v1/embeddings` is implemented and returns deterministic synthetic embeddings to keep IDE features working even when an embeddings backend is not configured.
Hindsight is an **agent memory system** (retain/recall/reflect) that can run locally and expose an **MCP server**. It can help reduce “context rot” and repeated “prompt too long” failures by moving long-term memory out of the chat history and into a separate store the agent can query.
CLIProxyAPI is a **proxy + translator**. It is not the “agent brain”, so the best integration is to run Hindsight as a **sidecar** and let your CLI agent call it via MCP tools.
Repo: https://github.com/vectorize-io/hindsight
## Recommended architecture
- Keep CLIProxyAPI focused on: auth, routing, translation, compatibility fixes, and request trimming.
- Run Hindsight separately (Docker or `pip install hindsight-api`).
- Add Hindsight as an MCP server to your agentic client (Droid/Factory CLI, Claude Code, etc.).
- Teach the agent (via your system prompt) to:
-`retain` stable preferences and project facts
-`recall` before large context dumps
-`reflect` after successes/failures to improve consistency
## Run Hindsight locally
### Docker (easy)
```bash
# Provide an LLM key/model for Hindsight’s internal reasoning (separate from CLIProxyAPI)
exportHINDSIGHT_API_LLM_PROVIDER=openai
exportHINDSIGHT_API_LLM_API_KEY=sk-...
exportHINDSIGHT_API_LLM_MODEL=gpt-4o-mini
docker run --rm -it -p 8888:8888 -p 9999:9999 \
-e HINDSIGHT_API_LLM_PROVIDER \
-e HINDSIGHT_API_LLM_API_KEY \
-e HINDSIGHT_API_LLM_MODEL \
-v $HOME/.hindsight-docker:/home/hindsight/.pg0 \
ghcr.io/vectorize-io/hindsight:latest
```
- REST API: `http://localhost:8888`
- MCP server: `http://localhost:8888/mcp`
- UI: `http://localhost:9999`
### Python (no Docker)
```bash
pip install hindsight-api -U
exportHINDSIGHT_API_LLM_PROVIDER=openai
exportHINDSIGHT_API_LLM_API_KEY=sk-...
hindsight-api
```
## Connect your agent via MCP
Hindsight exposes MCP two ways:
- **HTTP MCP**: `http://localhost:8888/mcp`
- **stdio MCP**: `hindsight-local-mcp`
How you register an MCP server depends on the client. Many clients use a `.mcp.json` file. Typical patterns:
### Option A: stdio MCP (most portable)
```json
{
"mcpServers":{
"hindsight":{
"command":"hindsight-local-mcp",
"args":[],
"env":{
"HINDSIGHT_API_LLM_PROVIDER":"openai",
"HINDSIGHT_API_LLM_API_KEY":"sk-..."
}
}
}
}
```
### Option B: HTTP MCP
If your client supports HTTP MCP servers, point it at:
-`http://localhost:8888/mcp`
## Operational notes / gotchas
- **Privacy**: Hindsight stores memory content (and may call its own LLM). Treat it like a database of user data.
- **Tool-call correctness**: If you see tool sequencing errors in Claude/Vertex, check `docs/tool-calling-and-cli-compat.md`.
- **Keep it optional**: Don’t enable or inject memory tools by default from CLIProxyAPI; different clients have different tool behaviors, and tool injection can increase prompt size.
## Next step (if you want deeper integration)
If you want CLIProxyAPI to “own” the Hindsight lifecycle, we can spec a thin optional module:
-`config.yaml`:
-`hindsight.enabled`
-`hindsight.base_url`
-`hindsight.auth_header` (optional)
- A small “memory helper” doc + scripts to start/stop the Hindsight sidecar alongside CLIProxyAPI.
5. Fallback hash of `Authorization + User-Agent` (hashed; raw values are not stored)
## Retrieval / injection
- Dropped items are appended as JSONL events (`dropped_chat` / `dropped_responses`)
- Retrieval is a lightweight keyword scorer over the last ~2MB of stored events
- Up to 8 snippets (~6k chars) are injected by appending to the **last user message** (best-effort).
## Anchored summaries (heuristic)
When history is trimmed, ProxyPilot writes a lightweight anchor summary based on **heuristic extraction**
(latest user intent + file/command hits). It does not use an LLM to summarize old turns.
Key behaviors:
- No LLM-based summarization of old turns. The anchored summary is generated by heuristic extraction (latest user intent + file/command hits) from dropped messages, not a model-produced summary.
- No explicit “~140k sweet spot” threshold or scheduled compression cycle. The memory flow only kicks in when a request exceeds budget and must be trimmed.
- No explicit cache-aware anchoring beyond append-only scaffold (we preserve append-only, but we don’t do Droid’s full anchored compression pipeline).
To preserve stable prompt prefixes for long sessions, CLIProxyAPI injects session state as an **append-only scaffold** by default (it appends a new system message/input entry instead of mutating prior content).
Packed block format:
-`<proxypilot_state>`
-`<pinned>` from `pinned.md`
-`<anchor>` from `summary.md`
-`<todo>` from `todo.md`
-`<spec>` when Spec Mode is enabled
## Text normalization
Some clients (notably PowerShell `ConvertTo-Json`) can store TODO/pinned state with escaped sequences like `\\n` / `\\u2019`.
CLIProxyAPI normalizes common escape sequences when reading/writing `todo.md`, `pinned.md`, and `summary.md` so the injected state stays readable.
## AGENTS.md pinning
If `harness-root-dir` is set (or provided to the middleware), CLIProxyAPI will read `AGENTS.md` from that directory and merge it into pinned context.
## Spec Mode (optional)
Spec Mode can be enabled to require a spec/plan before any code changes:
- Header: `X-CLIProxyAPI-Spec-Mode: true`
- Env: `CLIPROXY_SPEC_MODE=1`
When enabled and not yet approved, the scaffold injects a spec prompt that asks the agent to produce a reviewable specification and wait for approval.
## Semantic memory (local embeddings)
ProxyPilot can optionally embed and retrieve semantic memory using a local Ollama embed model.
This runs fully local and is **always-on** when enabled.
Embedding of dropped history is queued in the background to keep request latency low.
Environment variables:
-`CLIPROXY_SEMANTIC_ENABLED` (default: enabled)
-`CLIPROXY_SEMANTIC_MODEL` (default: `embeddinggemma`; also supports `nomic-embed-text`)
For advanced context compression using LLM-generated structured summaries and token-aware proactive trimming, see [Context Compression](context-compression.md).
## Agentic Harness
For structured incremental development workflows (INITIALIZER/CODING modes, feature tracking, progress logging), see [Agentic Harness](agentic-harness.md).
ProxyPilot is a local AI proxy with CLI + system tray interface for agentic tools (Droid/Factory, Codex CLI, Warp).
This repo started as a fork of CLIProxyAPI, but is now branded as ProxyPilot. For compatibility, some internal names and headers still use `CLIProxyAPI` (e.g. `X-CLIProxyAPI-*`).
## What’s included
- ProxyPilot tray app (Windows) to start/stop/restart the proxy engine, open the web dashboard in browser, open logs, and toggle autostart.
- Prompt-budget middleware to reduce “prompt too long” failures in strict CLIs.
- Long-session memory with pinned state + persistent TODO (local disk).
- Optional auth debugging and cooldown reset endpoints for local troubleshooting.
- `Auto-start proxy` (starts the proxy engine when ProxyPilot launches)
On Windows this uses a per-user Run entry (no admin required).
## Local management (Dashboard / tray)
When the proxy is started by the ProxyPilot tray app, it sets a per-user `MANAGEMENT_PASSWORD` for the engine process so `/v0/management/*` endpoints are enabled for local use.
The password is stored in the ProxyPilot UI state file:
* `SetProviders` replaces the provider slice using a defensive copy.
* `Providers` retrieves a snapshot that can be iterated safely from other goroutines.
* `BuildProviders` translates `config.Config` access declarations into runnable providers. When the config omits explicit providers but defines inline API keys, the helper auto-installs the built-in `config-api-key` provider.
If the manager itself is `nil` or no providers are configured, the call returns `nil, nil`, allowing callers to treat access control as disabled.
## Authenticating Requests
```go
result, err := manager.Authenticate(ctx, req)
result, authErr := manager.Authenticate(ctx, req)
switch {
case err == nil:
case authErr == nil:
// Authentication succeeded; result describes the provider and principal.
case errors.Is(err, sdkaccess.ErrNoCredentials):
case sdkaccess.IsAuthErrorCode(authErr, sdkaccess.AuthErrorCodeNoCredentials):
// No recognizable credentials were supplied.
case errors.Is(err, sdkaccess.ErrInvalidCredential):
case sdkaccess.IsAuthErrorCode(authErr, sdkaccess.AuthErrorCodeInvalidCredential):
// Supplied credentials were present but rejected.
default:
// Transport-level failure was returned by a provider.
// Internal/transport failure was returned by a provider.
}
```
`Manager.Authenticate` walks the configured providers in order. It returns on the first success, skips providers that surface `ErrNotHandled`, and tracks whether any provider reported `ErrNoCredentials`or`ErrInvalidCredential` for downstream error reporting.
If the manager itself is `nil` or no providers are registered, the call returns `nil, nil`, allowing callers to treat access control as disabled without branching on errors.
`Manager.Authenticate` walks the configured providers in order. It returns on the first success, skips providers that return `AuthErrorCodeNotHandled`, and aggregates `AuthErrorCodeNoCredentials`/`AuthErrorCodeInvalidCredential` for a final result.
Each `Result` includes the provider identifier, the resolved principal, and optional metadata (for example, which header carried the credential).
## Configuration Layout
## Built-in `config-api-key` Provider
The manager expects access providers under the `auth.providers` key inside `config.yaml`:
The proxy includes one built-in access provider:
- `config-api-key`: Validates API keys declared under top-level `api-keys`.
- Metadata: `Result.Metadata["source"]` is set to the matched source label.
In the CLI server and `sdk/cliproxy`, this provider is registered automatically based on the loaded configuration.
```yaml
auth:
providers:
- name: inline-api
type: config-api-key
api-keys:
- sk-test-123
- sk-prod-456
api-keys:
- sk-test-123
- sk-prod-456
```
Fields map directly to `config.AccessProvider`: `name` labels the provider, `type` selects the registered factory, `sdk` can name an external module, `api-keys` seeds inline credentials, and `config` passes provider-specific options.
## Loading Providers from External Go Modules
### Loading providers from external SDK modules
To consume a provider shipped in another Go module, point the `sdk` field at the module path and import it for its registration side effect:
To consume a provider shipped in another Go module, import it for its registration side effect:
```go
import (
@@ -89,19 +80,11 @@ import (
)
```
The blank identifier import ensures `init` runs so `sdkaccess.RegisterProvider` executes before `BuildProviders`is called.
## Built-in Providers
The SDK ships with one provider out of the box:
- `config-api-key`: Validates API keys declared inline or under top-level `api-keys`. It accepts the key from `Authorization: Bearer`, `X-Goog-Api-Key`, `X-Api-Key`, or the `?key=` query string and reports `ErrInvalidCredential` when no match is found.
Additional providers can be delivered by third-party packages. When a provider package is imported, it registers itself with `sdkaccess.RegisterProvider`.
The blank identifier import ensures `init` runs so `sdkaccess.RegisterProvider` executes before you call `RegisteredProviders()`(or before `cliproxy.NewBuilder().Build()`).
### Metadata and auditing
`Result.Metadata` carries provider-specific context. The built-in `config-api-key` provider, for example, stores the credential source (`authorization`, `x-goog-api-key`, `x-api-key`, or`query-key`). Populate this map in custom providers to enrich logs and downstream auditing.
`Result.Metadata` carries provider-specific context. The built-in `config-api-key` provider, for example, stores the credential source (`authorization`, `x-goog-api-key`, `x-api-key`, `query-key`,`query-auth-token`). Populate this map in custom providers to enrich logs and downstream auditing.
A provider must implement `Identifier()` and `Authenticate()`. To expose it to configuration, call `RegisterProvider` inside `init`. Provider factories receive the specific `AccessProvider` block plus the full root configuration for contextual needs.
A provider must implement `Identifier()` and `Authenticate()`. To make it available to the access manager, call `RegisterProvider` inside `init` with an initialized provider instance.
## Error Semantics
- `ErrNoCredentials`: no credentials were present or recognized by any provider.
- `ErrInvalidCredential`: at least one provider processed the credentials but rejected them.
- `ErrNotHandled`: instructs the manager to fall through to the next provider without affecting aggregate error reporting.
- `NewNoCredentialsError()` (`AuthErrorCodeNoCredentials`): no credentials were present or recognized. (HTTP 401)
- `NewInvalidCredentialError()` (`AuthErrorCodeInvalidCredential`): credentials were present but rejected. (HTTP 401)
- `NewNotHandledError()` (`AuthErrorCodeNotHandled`): fall through to the next provider.
Return custom errors to surface transport failures; they propagate immediately to the caller instead of being masked.
Errors propagate immediately to the caller unless they are classified as `not_handled` / `no_credentials` / `invalid_credential` and can be aggregated by the manager.
## Integration with cliproxy Service
`sdk/cliproxy` wires `@sdk/access` automatically when you build a CLI service via `cliproxy.NewBuilder`. Supplying a preconfigured manager allows you to extend or override the default providers:
`sdk/cliproxy` wires `@sdk/access` automatically when you build a CLI service via `cliproxy.NewBuilder`. Supplying a manager lets you reuse the same instance in your host process:
```go
coreCfg, _ := config.LoadConfig("config.yaml")
providers, _ := sdkaccess.BuildProviders(coreCfg)
manager := sdkaccess.NewManager()
manager.SetProviders(providers)
accessManager := sdkaccess.NewManager()
svc, _ := cliproxy.NewBuilder().
WithConfig(coreCfg).
WithAccessManager(manager).
WithConfigPath("config.yaml").
WithRequestAccessManager(accessManager).
Build()
```
The service reuses the manager for every inbound request, ensuring consistent authentication across embedded deployments and the canonical CLI binary.
Register any custom providers (typically via blank imports) before calling `Build()` so they are present in the global registry snapshot.
### Hot reloading providers
### Hot reloading
When configuration changes, rebuild providers and swap them into the manager:
When configuration changes, refresh any config-backed providers and then reset the manager's provider chain:
This mirrors the behaviour in `cliproxy.Service.refreshAccessProviders` and `api.Server.applyAccessConfig`, enabling runtime updates without restarting the process.
This mirrors the behaviour in `internal/access.ApplyAccessProviders`, enabling runtime updates without restarting the process.
Merged upstream CLIProxyAPI releases `v6.8.40` through `v6.8.46` into ProxyPilot on branch `sync-upstream-v6.8.40-to-v6.8.46` and prepared the next dev release tag `v0.3.0-dev-0.40`.
## Upstream Releases Included
- `v6.8.40` - adaptive thinking updates across Claude, Codex, OpenAI, and Antigravity flows
- `v6.8.45` - GPT-5.4 description cleanup and conditional `service_tier` handling
- `v6.8.46` - Gemini 3.1 Flash Lite Preview registry support and Codex websocket `response.create` cleanup
## Conflict Resolution Notes
- Kept ProxyPilot branding in `README.md` during the `v6.8.41` merge.
- Preserved deletion of `README_CN.md` when upstream tried to reintroduce it.
- Accepted upstream behavior in translator, watcher, executor, and registry paths unless the fork already had a stronger ProxyPilot-specific customization.
## Verification
- Clean `origin/main` baseline still fails `go test ./sdk/cliproxy/auth` in `TestFileStore_Delete_RemovesFile`; treated as a pre-existing failure.
- Targeted subsystem checks passed after each upstream merge for thinking, translators, watcher, registry, executor, and OpenAI websocket paths.
- Final full-suite verification should still record the pre-existing `sdk/cliproxy/auth` failure explicitly if it remains on the final release commit.
Merged upstream CLIProxyAPI releases `v6.8.47` through `v6.8.49` into ProxyPilot on branch `sync-upstream-v6.8.47-to-v6.8.49`, fixed fork-specific release lookup drift, and prepared the next dev release tag `v0.3.0-dev-0.41`.
- Kept ProxyPilot branding and structure in `README.md` during the `v6.8.49` merge.
- Preserved deletion of `README_CN.md` when upstream tried to reintroduce it.
- Accepted upstream behavior in watcher, registry, auth scheduler, translator, and executor paths unless the fork already had a stronger ProxyPilot-specific customization.
## Fork-Specific Follow-Ups
- Updated `UPSTREAM_VERSION` to `v6.8.49`.
- Updated `features.json` upstream marker to `CLIProxyAPI v6.8.49`.
- Fixed `.github/workflows/check-upstream.yml` so it compares the tracked `UPSTREAM_VERSION` tag against the latest upstream release tag instead of comparing upstream HEAD to itself.
- Switched fork-facing release lookups in `internal/api/handlers/management/config_basic.go` and `internal/proxypilotupdate/github.go` to `Finesssee/ProxyPilot` releases.
Merged upstream CLIProxyAPI releases `v6.8.50` and `v6.8.51` directly into ProxyPilot `main`, resolved fork-specific conflicts in startup and Docker publishing paths, prepared the next dev release tag `v0.3.0-dev-0.42`, then switched release publishing to a local/manual flow and uploaded the release assets outside GitHub Actions.
## Upstream Releases Included
- `v6.8.50` - preserve string-form system prompts, cap websocket body log growth, and extend scheduler / executor coverage
- `v6.8.51` - refresh model catalog from network on startup, register Codex models by plan tier, and refresh release / workflow paths
## Conflict Resolution Notes
- Kept ProxyPilot Docker publishing identity in `.github/workflows/docker-image.yml` and added the upstream models catalog refresh step before image builds.
- Kept ProxyPilot's management-password startup flow in `cmd/server/main.go` and added `registry.StartModelsUpdater(...)` before service startup.
- Accepted the upstream move from generated static registry data to embedded `internal/registry/models/models.json`, including deletion of `internal/registry/model_definitions_static_data.go`.
## Fork-Specific Follow-Ups
- Updated `internal/cmd/list_models.go` to use the new Codex plan-tier model getters instead of the removed `registry.GetOpenAIModels`.
- Fixed `sdk/auth/filestore.go` delete-path resolution so nested relative auth IDs delete under the configured base dir, absolute paths still work for the management API, and traversal outside the base dir is rejected.
- Updated `UPSTREAM_VERSION` to `v6.8.51`.
- Updated `features.json` upstream marker to `CLIProxyAPI v6.8.51`.
- Changed `.github/workflows/proxypilot-release.yml` and `.github/workflows/docker-image.yml` from tag-triggered runs to manual dispatch so releases can be built locally and uploaded manually instead of consuming GitHub Actions minutes.
- Built the `v0.3.0-dev-0.42` release assets locally from the tag snapshot and uploaded:
- `go test ./internal/runtime/executor/...` - PASS
- `go test ./internal/registry/...` - PASS
- `go test ./sdk/api/handlers/openai/...` - PASS
- `go test ./sdk/cliproxy/auth -run 'TestFileStore_Delete_(RemovesFile|NotFound|AbsolutePath|RejectsTraversalOutsideBaseDir)' -v` - PASS
- `go test ./sdk/cliproxy/auth` - PASS
- `go test ./sdk/cliproxy/...` - PASS
- `go test ./...` - PASS
## Release Notes
- Next dev tag: `v0.3.0-dev-0.42`
- Latest upstream target at merge time: `CLIProxyAPI v6.8.51`
## Maintenance Review
- Completed parallel findings-first review of `internal/auth/*`, `internal/translator/*`, and `internal/api/handlers/*` on 2026-03-12 12:01:40 +07.
- Auth review findings:
- `internal/auth/kiro/token_repository.go` only queues `auth_method` values `idc` and `builder-id` for background refresh even though `internal/auth/kiro/oauth.go` persists social logins as `auth_method: "social"` and the refresher has a default OAuth refresh path for non-SSO methods. Result: Google/GitHub Kiro tokens are skipped by auto-refresh.
- `internal/auth/kiro/token_repository.go` walks nested token files but rewrites refreshed tokens back to `filepath.Join(baseDir, token.ID)` after truncating IDs to `filepath.Base(path)`. Result: nested Kiro token files would be rewritten into the auth root instead of their original directory.
- Translator review findings:
- `internal/translator/kiro/openai` can emit streamed tool-call argument deltas with the wrong OpenAI tool index.
- `internal/translator/kiro/openai` does not actually enforce `tool_choice`; it still forwards tools when callers request `none` or pin a specific function.
- `internal/translator/openai/openai/responses` downgrades `developer` messages to `user`.
- `internal/translator/gemini/openai/chat-completions` reads/mapping finish reasons incorrectly for some response shapes.
- `internal/translator/gemini/openai/responses` can send `functionResponse.name = "unknown"` on follow-up tool-output requests.
- API handler review findings:
- `internal/api/handlers/management/memory.go` uses raw `session` query values to build filesystem paths for delete/export/import instead of the memory store's sanitized session path helper, allowing traversal outside the memory store root.
- `internal/api/handlers/management/generic_import.go` derives auth filenames from the first characters of imported API keys, exposing secret prefixes in filenames and management responses.
- `internal/api/handlers/management/auth_files.go` writes uploaded auth files to disk before validation and leaves invalid files behind on registration failure.
- `internal/api/handlers/management/config_basic.go` returns a zero-value config from `GetConfig` because it serializes `new(*h.cfg)` instead of the current config.
- `internal/api/handlers/management/auth_priority.go` and `internal/api/handlers/management/auth_files.go` update priority in places the runtime selector does not read, so management-set auth priority does not affect routing and does not persist reliably for file-backed auths.
- Config mutation helpers in `internal/api/handlers/management/handler.go` mutate `h.cfg` before persistence succeeds, so failed writes can leave runtime config diverged from disk until restart.
- Review validation:
- `go test ./internal/auth/... ./internal/translator/... ./internal/api/handlers/...` - PASS
## Maintenance Fixes
- Hardened management memory delete/export/import so session names are validated before resolving filesystem paths, blocking traversal attempts outside the memory store root.
- Changed bulk API-key import filenames to opaque `provider-import-...` names so credential prefixes are no longer exposed in filenames or management responses.
- Changed auth upload handling to validate uploaded JSON before committing it to disk and to roll back failed registrations instead of leaving broken auth files behind.
- Fixed `GetConfig` to return the live in-memory config and changed the simple config mutation helpers to persist cloned config state instead of mutating runtime config before the file write succeeds.
- Fixed file-backed auth persistence so `priority`, `prefix`, and `proxy_url` survive saves, reloads, and management edits, and so runtime selection prefers `Auth.Priority` with a legacy attribute fallback.
- Changed token writers across codex, claude, gemini, kimi, qwen, iflow, copilot, and vertex to use atomic `0600` JSON writes.
- Fixed Kiro background refresh so blank refresh responses do not erase the existing refresh token, nested token paths refresh in place, and social auth token files participate in the refresh scan.
- Fixed Kiro OAuth callback servers so shutdown waits on a dedicated completion signal instead of racing the login flow for the only callback result.
- Fixed Kiro/OpenAI tool streaming to keep a stable mapping from content-block indexes to OpenAI tool ordinals.
- Fixed Kiro/OpenAI request translation so `tool_choice=none|required|function` is enforced instead of treated as a prompt-only hint.
- Fixed OpenAI Responses to Chat Completions conversion to preserve `developer` role messages.
- Fixed Gemini/OpenAI finish-reason mapping so OpenAI receives valid enums while the original Gemini reason is preserved in `native_finish_reason`.
- Added a bounded Gemini function-call name cache so `previous_response_id` follow-up tool outputs resolve the original function name and fail closed when they cannot.
## Maintenance Fix Verification
- `go test ./internal/auth/... ./sdk/auth ./sdk/cliproxy/auth ./internal/translator/... ./internal/api/handlers/...` - PASS
- `go test ./internal/api/handlers/management` - PASS
Merged upstream CLIProxyAPI release `v6.8.55` into ProxyPilot `main`, carrying forward the fork's workflow and branding decisions while absorbing upstream changes from `v6.8.52` through `v6.8.55`.
Built the release assets locally and published `v0.3.0-dev-0.43` on GitHub as the latest ProxyPilot release.
## Upstream Releases Included
- `v6.8.52` - Codex websocket header defaults and related compatibility updates
- `v6.8.53` - Antigravity model refresh and proxy helper improvements
- `v6.8.54` - auth file metadata exposure and management API test coverage
- `v6.8.55` - auth list priority/note support, Gemini responses tool-output fixes, and proxy utility additions
## Conflict Resolution Notes
- Kept the fork versions of `.github/workflows/docker-image.yml`, `.github/workflows/release.yaml`, `.github/workflows/pr-test-build.yml`, and `.goreleaser.yml` so this sync does not alter CI or release automation.
- Kept ProxyPilot README branding and deleted the upstream `README_CN.md` revival because this fork intentionally does not ship that file.
- Merged `internal/api/handlers/management/auth_files.go` semantically so ProxyPilot keeps its `Auth.Priority` persistence while also adding upstream `note` support and priority/note exposure in the auth-file API.
- Kept the stricter Gemini Responses JSON validation in `internal/translator/gemini/openai/responses/gemini_openai-responses_request.go` so partial JSON strings are not embedded as structured results.
## Repo State Updates
- Updated `UPSTREAM_VERSION` to `v6.8.55`.
- Updated `features.json` upstream marker to `CLIProxyAPI v6.8.55`.
WelcomeLabel2=This will install [name/ver] on your computer.%n%nProxyPilot is a local AI proxy that routes requests to Claude, OpenAI, and other providers with automatic credential management.%n%nIt is recommended that you close all other applications before continuing.
Some files were not shown because too many files have changed in this diff
Show More
Reference in New Issue
Block a user
Blocking a user prevents them from interacting with repositories, such as opening or commenting on pull requests or issues. Learn more about blocking a user.