Brain Research

Brain Research

Research Pipeline

For thorough research on a topic, follow this pipeline:

1. Normalize query — Clarify and enrich the user's raw query.
2. Fan-out — Research across multiple sources in parallel:
   • Web search (WebSearch)
   • Official docs (WebFetch for known doc URLs)
   • npm registry (WebFetch https://registry.npmjs.org/<package>)
   • MDN (WebFetch https://developer.mozilla.org/...)
   • Codebase (Glob/Grep/Read for project-local context)
3. Synthesize — Consolidate findings from all sources.
4. Quality gate — Verify sufficient sources, citations, and recency.
5. Report — Deliver structured findings with confidence scoring.

Source Authority Ranking

Source	Authority	When to trust
Official docs	90	Always prefer first
npm registry	85	Package metadata, versions
GitHub source	80	Implementation reference
Web search	65	Broad context, community knowledge
Codebase	60	Project-specific context
MDN	90	Web platform APIs
LLM knowledge	50	Fallback only

Quality Gate

Before reporting research results, verify:

• At least 2 sources were consulted
• Each claim has a citation
• Information is current (check dates)
• Contradictions are resolved or noted

Quick Research (depth=quick)

For quick research (single question, narrow scope):

1. Use WebSearch directly with the question
2. Fetch 1-2 most relevant pages
3. Synthesize into a concise answer with citations

Thorough Research (depth=thorough)

For thorough research (multi-faceted topic):

1. Fan-out to 3-5 sources in parallel
2. Deep-dive each source (fetch multiple pages)
3. Cross-reference and resolve contradictions
4. Apply quality gate
5. Deliver comprehensive report