Beyond Keyword Matching
Traditional file search finds files by name or by text content. Terraphim's KG-boosted file search finds files by their semantic relationship to your domain concepts and role-specific vocabulary.
When you search for "requirements validation", Terraphim does not just grep for those words. It traverses your knowledge graph to find documents connected to those concepts through synonyms, co-occurrence edges, and domain relationships. A document titled "acceptance criteria review" surfaces because the knowledge graph knows these concepts are related in a systems engineering context.
Multi-Source Search
Terraphim searches across multiple sources simultaneously:
- Local filesystem -- Markdown files, code, documentation
- Knowledge repositories -- Obsidian vaults, Logseq graphs
- Shared servers -- Atomic Server for team knowledge
- Code search -- GitHub, StackOverflow, and other configured haystacks
Results from all sources are unified, deduplicated, and ranked using a single knowledge graph.
Intelligent Deduplication
When searching across multiple haystacks, the same document often appears in different sources. Terraphim handles duplicates intelligently, merging results and preserving the highest-ranked version rather than showing the same content multiple times.
Role-Specific Relevance
The same search query produces different rankings for different roles:
- A systems engineer searching "testing" sees verification and validation documents first
- A project manager searching "testing" sees schedule and resource allocation documents
- A quality analyst searching "testing" sees compliance and audit documents
Each role's knowledge graph contains domain-specific concepts and synonyms that reshape how results are scored and ordered.
Context Collections
Terraphim organises searchable content into context collections -- curated sets of documents, taxonomies, and vocabulary that define a domain. Collections can be shared across a team or customised per user, providing consistent search relevance without centralised configuration.
Integration with AI Assistants
Through the MCP server, KG-boosted file search is available directly in your AI coding assistant. When Claude Code searches for files, it uses your knowledge graph to find semantically relevant results, not just filename matches.