Web Research in Pi
Use Pi's Perplexity-backed tools:
web_searchfor source discovery.web_answerfor focused factual questions.web_researchfor broad or multi-step investigations.
Boundaries
Use this for:
- comparisons and trade-offs
- recent facts and release behavior
- standards and external best practices
- vendor docs or public evidence
Do not use web tools for private code, secrets, credentials, proprietary data, or local code exploration. Use local search first for repo-specific questions.
Workflow
- Restate the research question and decide if it is simple or broad.
- For simple factual questions, use
web_answerwith a focused query. - For source selection, use
web_search, then cite the best official or primary sources from the returned results. - For broad investigations, use
web_researchand tell the user the report will arrive asynchronously. - Compare claims against local project constraints before recommending changes.
- Report uncertainty and source gaps directly.
Failure Cases
- Web tools return no useful results: report the gap directly in the Gaps section; do not fabricate sources.
- Question requires private code or credentials to answer: refuse the web query, answer from local context only, and note the limitation.
web_researchis unavailable: fall back toweb_search+web_answerand note the fallback.
Output Contract
## Research Result
### Answer
<concise answer>
### Evidence
- <source title/url or grounded result> — <why it matters>
### Fit For This Repo
<what changes because of local constraints>
### Gaps
<any missing evidence>