docs(config): add model mappings example

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>
This commit is contained in:
Finessse
2025-12-26 05:09:11 +07:00
co-authored by Claude Opus 4.5
parent 5cf7518f26
commit f52bc59b2e
+45 -2
View File
@@ -26,8 +26,15 @@ remote-management:
# Authentication directory (supports ~ for home directory)
auth-dir: "~/.cli-proxy-api"
# Long-running agent harness files root directory (feature_list.json, claude-progress.txt).
# Defaults to the server working directory when empty.
# The harness guides AI agents through structured, incremental development workflows.
# - INITIALIZER mode: Creates feature list, progress log, and init.sh for new sessions
# - CODING mode: Forces one-feature-at-a-time workflow with verification
# Leave empty to use session-based memory storage (recommended).
# Set to a specific directory to share harness files across sessions.
# harness-root-dir: ""
#
# Environment variables:
# CLIPROXY_HARNESS_ENABLED=true/false # Enable/disable harness injection (default: true)
# API keys for authentication
api-keys:
@@ -71,13 +78,49 @@ quota-exceeded:
routing:
strategy: "round-robin" # round-robin (default), fill-first
# Global model mappings - route friendly names to actual models across all providers.
# These mappings are checked before per-credential model mappings.
# global-model-mappings:
# - from: "smart" # Client requests "smart"
# to: "claude-opus-4-5-20251101" # Routes to Claude Opus
# - from: "rush"
# to: "claude-haiku-4-5-20251001" # Routes to Claude Haiku (fast)
# - from: "think"
# to: "claude-opus-4-5-thinking" # Routes to thinking model
# - from: "gemini-pro"
# to: "gemini-2.5-pro-preview"
# provider: "gemini" # Only applies to Gemini provider
# - from: "cheap"
# to: "gpt-4o-mini"
# provider: "codex"
# enabled: false # Disabled mapping (kept for reference)
# When true, enable authentication for the WebSocket API (/v1/ws).
ws-auth: false
# Streaming behavior (SSE keep-alives + safe bootstrap retries).
# Streaming behavior (SSE keep-alives, bootstrap retries, chunk sizing).
# streaming:
# keepalive-seconds: 15 # Default: 0 (disabled). <= 0 disables keep-alives.
# bootstrap-retries: 1 # Default: 0 (disabled). Retries before first byte is sent.
# max-chunk-size: 65536 # Default: 65536 (64KB). Max bytes per response chunk. 0 disables limiting.
# Context compression behavior (Factory.ai-style structured summarization).
# Uses LLM to generate intelligent summaries with structured sections when context
# exceeds the threshold. Summaries include: session intent, file modifications,
# decisions made, next steps, and technical details.
# compression:
# enabled: true # Default: true. Enable LLM-based structured summarization.
# threshold-percent: 0.75 # Default: 0.75. Trigger compression at 75% of context window.
# max-summary-tokens: 2000 # Default: 2000. Maximum tokens for generated summaries.
# summarization-timeout-seconds: 30 # Default: 30. Timeout for LLM summarization calls.
# fallback-to-regex: true # Default: true. Use regex-based summary when LLM fails.
# Token-aware compression (proactive trimming based on model context window).
# Prevents "prompt too long" errors by trimming before hitting model limits.
# Environment variables:
# CLIPROXY_TOKEN_AWARE_ENABLED=true # Default: true. Enable token-based compression.
# CLIPROXY_COMPRESSION_THRESHOLD=0.75 # Default: 0.75. Trim at 75% of context window.
# CLIPROXY_RESERVE_TOKENS=8192 # Default: 8192. Tokens reserved for model output.
# Gemini API keys
# gemini-api-key: