Skip to content

CLAUDE.md

Latest commit

 

History

History
74 lines (57 loc) · 2.68 KB

CLAUDE.md

File metadata and controls

74 lines (57 loc) · 2.68 KB
Raw

CLAUDE.md

This file provides guidance to Claude Code (claude.ai/code) when working with code in this repository.

Running the Application

# Install dependencies
pip install -r requirements.txt

# Run the proxy server
python opencode.py

Server starts on:

Configuration

Configuration is managed through .env file (create from .env.example):

cp .env.example .env

Key environment variables:

  • OPENCODE_PROXY - Proxy server URL
  • OPENCODE_API_KEY - API key for OpenCode service
  • OPUS_MAP_MODEL - Model for opus route (default: kimi-k2.6)
  • SONNET_MAP_MODEL - Model for sonnet route (default: glm-5.1)
  • HAIKU_MAP_MODEL - Model for haiku route (default: minimax-m2.5)

Optional server overrides:

  • OPENCODE_HOST - Bind address (default: 0.0.0.0)
  • OPENCODE_PORT - API port (default: 4000)
  • OPENCODE_WEB_PORT - Web UI port (default: 8082)

Architecture

Entry Point

  • opencode.py - Main FastAPI server with request routing and token tracking

Configuration Module (config/)

  • __init__.py - Exports: API_BASE_OPENAI, API_BASE_ANTHROPIC, HOST, PORT, WEB_PORT, MODELS, ROUTES, get_model_config, API_KEY, PROXY
  • settings.py - Loads environment variables from .env, defines:
    • PROXY and API_KEY (secrets)
    • MODELS (model endpoints and protocols)
    • ROUTES (model name mappings, built from env vars)
    • get_model_config() - Returns merged config for a model

Dashboard Module (dashboard/)

  • __init__.py - Exports: register_dashboard, log, RichLogHandler, build_display, start_input_thread
  • api.py - Dashboard API endpoints: stats, logs, history, static file serving
  • display.py - Rich terminal display: token usage table, log panel, keyboard input (j/k/g/G/arrows)

Request Flow

  1. Request comes to /v1/messages or /anthropic/v1/messages
  2. _route_for() maps model name to route config using ROUTES
  3. get_model_config() gets endpoint and protocol for the model
  4. Request is forwarded to the appropriate endpoint (OpenAI or Anthropic protocol)
  5. Response is converted between Anthropic and OpenAI formats
  6. /v1/messages/count_tokens estimates token count without forwarding

Token Tracking

  • Non-streaming: Uses actual usage from API response
  • Streaming: Estimates input tokens, reads actual usage from SSE events when available
  • Stored in SQLite database (logs/requests.db)
  • Terminal display via Rich (dashboard/display.py)
  • Web dashboard via API endpoints (dashboard/api.py)

Web Dashboard

  • Static files in static/ directory
  • Token usage stats and request history via API endpoints