Your knowledge, your device.
Semantic search that understands.

Terraphim AI is a local-first assistant powered by knowledge graphs and Aho-Corasick automata. No cloud. No telemetry. Just fast, private, semantic search.

Install Now See How It Works

Built for privacy, designed for speed

A modular Rust workspace with 29 crates, each handling a single concern. Extend without breaking.

Knowledge Graphs

Custom graph-based semantic search using Aho-Corasick automata for fast text matching and concept extraction.

Zero Cloud Dependency

Every computation runs on your machine. No data leaves your device, ever. Privacy is not a feature -- it is the foundation.

Multiple Relevance Functions

Switch between TitleScorer, BM25, BM25F, BM25Plus, or TerraphimGraph per role and per query.

Haystack Integrations

Local files, Confluence, Jira, Discourse, JMAP email, Quickwit -- unified search across all your sources.

WASM Ready

The autocomplete engine compiles to WebAssembly for browser-based search. Under 200KB compressed.

Firecracker Isolation

Run untrusted code in microVMs with sub-2-second boot times and complete sandboxing.

How Terraphim works

1

Configure your roles

Define knowledge domains, haystacks, and relevance functions. Each role gets its own personalised search context.

2

Build knowledge graphs

Thesaurus files are compiled into Aho-Corasick automata for fast, semantic text matching across your documents.

3

Index your sources

Connect local folders, wikis, emails, and task managers. The middleware layer indexes everything into a unified search space.

4

Search semantically

Queries are expanded through the knowledge graph, matched against automata, and ranked by your chosen relevance function.

$ terraphim-agent search "knowledge graph architecture"
# Searching across 3 haystacks with BM25 relevance...
Found 12 results in 23ms

Start searching privately

Terraphim AI is free, open source, and built in Rust. Clone the repository, build with Cargo, and own your search.

View on GitHub Read the Docs