Skip to content

configuration.md

Latest commit

 

History

History
32 lines (25 loc) · 1.35 KB

configuration.md

File metadata and controls

32 lines (25 loc) · 1.35 KB
Raw

Configuration

All configuration is via environment variables or a .env file in the project root.

Environment variables

Variable Required Default Description
OPENAI_API_KEY Yes — LLM API key
OPENAI_BASE_URL No https://api.deepseek.com/v1 LLM endpoint base URL
MODEL No deepseek-v4-pro Model name
CONCURRENCY No 30 Max concurrent LLM calls
TEMP_DIR No /tmp/standalone-trace-analyzer Temp storage for run outputs
BLACKBOX_API_URL No http://localhost:8000 URL the CLI connects to

Example .env

OPENAI_API_KEY=sk-...
OPENAI_BASE_URL=https://api.deepseek.com/v1
MODEL=deepseek-v4-pro
CONCURRENCY=30
TEMP_DIR=/tmp/standalone-trace-analyzer
BLACKBOX_API_URL=http://localhost:8000

Notes

  • OPENAI_BASE_URL and OPENAI_API_KEY use OpenAI-compatible naming for compatibility with the client library, but any OpenAI-compatible provider works.
  • TEMP_DIR must be writable. Stage outputs are saved as JSON files under {TEMP_DIR}/runs/{run_id}/.
  • CONCURRENCY controls the asyncio.Semaphore limit on concurrent LLM requests. Lower this if you hit rate limits.
  • BLACKBOX_API_URL is read by the CLI (src/cli/client.py). If the API is running on a different host or port, set this before launching the TUI.