tech-search

Evidence-backed technical research reports are generated by decomposing a user question into parallel web searches, extracting deep page content, evaluating coverage, and synthesizing findings into structured documentation. This multi-phase workflow clarifies intent and captures key facts from reliable sources.

Query decomposition for deep web research breaks a complex technical question into multiple targeted sub-queries. These sub-queries run in parallel across web search and fetch tools to extract key facts, code examples, and expert quotes, ensuring comprehensive coverage evaluation before final evidence synthesis.

Automated web search and page extraction can optionally use MCP tools when available, but the workflow falls back to standard web search and fetch mechanisms without them. No explicit external dependencies are required to execute the multi-phase research pipeline and generate documentation.

Compiling a technical comparison report uses parallel web search and deep page extraction to gather evidence from multiple sources. The workflow evaluates coverage gaps, extracts structured JSON data, and synthesizes expert recommendations into a consolidated report for comparison and general deep technical topics.

The deep research workflow saves output files strictly under the docs/research/{YYYY-MM-DD}-{slug}/ directory. This guardrail ensures research outputs are safely contained as documentation and avoids writing implementation or production artifacts to unintended locations.