- •
~/.hermes/config.yaml — provider: manifest, type: openai, base_url, key_env
- •
~/.hermes/.env — CUSTOM_API_KEY=your-key
- • Provider name must be manifest, not custom
- • Agent restart required to pick up changes
- • Reads config.yaml + .env at startup
- • Provider router selects "manifest"/openai slot
- • Resolves base_url and reads key from key_env
- • Attaches key as Authorization: Bearer header
- • Must implement OpenAI Chat Completions API shape
- • GET /v1/models → model list (optional; fallback = free-text)
- • POST /v1/chat/completions → SSE token stream
- • Keyless: set CUSTOM_API_KEY=none in .env