| name | multi-provider-web-search |
| description | Multi-provider web search fanout across Exa, Tavily, Brave, LangSearch, and Marginalia with parallel dispatch, dedupe, and AI summary. Use ONLY on explicit user intent for a named provider (Exa, Tavily, Brave, LangSearch, Marginalia), cross-provider fanout, semantic code/paper search, indie small-web search, or when native web search returned weak results. Do NOT auto-pick on generic search queries; native WebSearch is preferred for those.
|
multi-provider-web-search
Thin shell to web-search-ai-summary CLI on PATH.
Use gate
- Use only on explicit user intent matching one of:
- Named provider request: Exa, Tavily, Brave, LangSearch, Marginalia
- Provider-fit query: code/docs/repos/papers (exa), current web/RAG (tavily),
independent mainstream (brave), free broad (langsearch), indie/small-web
long-tail (marginalia)
- Native WebSearch returned weak/empty results and user asks to try harder
- Do not use on generic "search the web" queries. Prefer native WebSearch,
web_search.
Run
web-search-ai-summary "<query>"
Options:
--provider exa|tavily|brave|langsearch|marginalia--num-results N from 1 to 20, default 10 per backend--timeout-ms N from 1000 to 300000, default 30000
Guidance
- Do not repeat the same query to compare backends.
- Search again with better keywords when results are weak.
- Fetch result URLs when summaries are insufficient.
- Use
gh for GitHub source lookup, not this tool.
- Use
--provider only for backend debugging or clear provider fit.
Runtime
- CLI binary symlinked to
~/.local/bin/web-search-ai-summary. Source of truth:
AI-configs/pi/extensions/web_search/run.ts.
