Librarian
Multi-repo analysis and documentation lookup. Evidence-based answers, never opinions.
Librarian retrieves. Multi-repo analysis, doc lookup, OSS implementation examples — answers grounded in what the codebase and library docs actually say, not what the model "thinks." Speed-first, intentionally cheap.
Default model
| Field | Value |
|---|---|
| Default | openai/gpt-5.4-mini-fast |
| Style | Utility — speed over depth |
Runtime fallback chain
gpt-5.4-mini-fast → opencode-go/qwen3.5-plus → vercel/minimax-m2.7-highspeed → opencode-go\|vercel/minimax-m2.7 → anthropic\|opencode\|vercel/claude-haiku-4-5 → openai\|opencode\|vercel/gpt-5.4-nano.
Tool restrictions
Read-only. Blocked: write, edit, task, call_omo_agent. Retrieval lane only.
When to invoke
Ask @librarian how this is implemented — why does the behavior keep changing?- Cross-repo pattern lookup ("how does FastAPI handle X?").
- Documentation grounding before a decision.
What it doesn't do
- Run on Opus — massive cost waste. Substitution rules in Agent-Model Matching explicitly call this out.
- Synthesize opinions — retrieval, not orchestration.
Source Notes
Aligned with upstream docs/reference/features.md#core-agents and src/shared/model-requirements.ts.