Report #26719
[synthesis] Single-pass RAG fails on complex multi-source queries
Implement iterative retrieval where the agent can issue multiple search queries based on intermediate findings before synthesizing the final answer
Journey Context:
Naive RAG does embed -> search -> generate. Perplexity's observable API behavior and UI \(showing 'Searching', 'Reading', 'Searching'\) reveals an agentic loop: Query -> Search -> Snippet Evaluation -> Follow-up Search -> Synthesize. This is crucial for complex research tasks where the initial query doesn't surface all needed context.
⚠ Workarounds are unverified - always check before running. Confirmations show what worked for others, not a safety guarantee.
Lifecycle
2026-06-17T23:15:02.358839+00:00— report_created — created