intelligence/hermes-search

Cross-vault and cross-repo search with provenance is performed by routing queries to vaults, GitHub MCP, and web sources in parallel, then returning ranked results with explicit source citations for every item retrieved.

Contradiction detection during multi-source retrieval identifies conflicting information across vaults, repos, and the web, explicitly surfacing contradictions with source citations without resolving them, ensuring users see conflicting data points.

Cross-repo and cross-vault search operates in read-only mode, never writing to vaults or external systems, so users can safely retrieve and deduplicate ranked results with full provenance across all connected sources.

Deduplication during cross-source retrieval removes duplicate results by comparing items fetched from vaults, repos, and the web, then ranks the remaining unique results by relevance, recency, and credibility with provenance intact.

Cross-vault search integrates with GitHub MCP for cross-repo retrieval and supports multi-agent collaboration, routing queries to correct sources and executing parallel searches across vaults, repositories, and optional web research.

Read-only cross-source search is not suitable when you need to write, update, or resolve contradictions in your vaults or repositories, as it strictly retrieves, deduplicates, and surfaces conflicting information without modifying source data.