Property Graph in SQLite
Parses source files into vertices and edges for modules, classes, functions, and relationships, then persists them to a single SQLite database. Smart sync runs a full build first, then incremental updates.
MIT • Python • CLI + HTTP API
Trailhead
Trailhead parses a source tree into a property graph of modules, classes, functions, and their relationships, then stores it in a single SQLite file. Add vector embeddings for semantic search over your own codebase through a CLI, a warm-model HTTP API, and a browser UI that all run locally.
14
Languages via tree-sitter
SQLite
Single-file graph & vectors
MIT
Open source license
A Queryable Graph of Your Own Code
Index once, then query it from the CLI, the HTTP API, or a browser UI. Your code stays on your machine.
Polyglot via Tree-sitter
Python works out of the box. JavaScript, TypeScript, Rust, Go, Java, C#, C, C++, Ruby, PHP, Bash, HTML, and Ansible-shaped YAML are available as optional extras. Install only the languages you need.
Semantic Similarity Search
Generate embeddings with sentence-transformers and query them for semantic similarity, not just exact text match. Falls back to SQLite BLOB storage when the vector extension isn't available on a given platform.
Background Reindexing
The server watches the source tree and reindexes changed files incrementally while keeping the embedding model warm in memory, so there is no need to rerun the indexer by hand.
Browser UI & CLI
A single th command handles indexing, serving, and querying. The same server exposes an interactive browser UI at localhost:8000 for visualizing and querying the graph directly.
HTTP API & OpenAPI Schema
Full REST API for embeddings, SQL queries, semantic search, and graph traversal, documented at /openapi.json and /docs. Built for scripting and for feeding an LLM prompt context.
Who It’s Built For
Anyone who needs a codebase's structure and meaning queryable, without shipping the code anywhere.
AI & LLM Tooling Builders
Give an agent or a RAG pipeline structured, queryable context about a codebase instead of dumping raw files into a prompt window.
Platform & Developer Tooling Teams
Explore an unfamiliar or legacy codebase's structure, including module boundaries, call relationships, and cross-file references, without reading it file by file.
Regulated & Air-Gapped Environments
Everything runs locally against a SQLite file. No code, embeddings, or queries leave the machine running Trailhead.
Get Started
pip install trailhead
th index .
th serve .Then visit http://localhost:8000 for the browser UI, or query straight from the CLI with th query similar "...".
Works Well With
Dev-tooling neighbors from the same shop.
WebUtils
Repo2Prompt packages a codebase for LLM prompt contexts entirely in the browser. Trailhead is the deeper local option, with a persistent queryable graph and semantic search over the same code.
Learn More →Adapt
Point Adapt at a folder for auth-gated APIs and MCP resource exposure. Point Trailhead at the same folder for semantic code search and a queryable property graph.
Learn More →Free to Use. Commercially Supported.
Trailhead is MIT licensed. McIndi offers commercial support contracts for organizations that need SLA-backed maintenance, custom language support, and integration work.