Welcome to Eulix

Eulix turns a code repository into something you can ask questions of: where a function is defined, who calls it, how a feature works, what depends on what. It is built for large, mixed-language repositories, runs on your own machine, and uses a language model only when a question actually needs one.

Last verified against: ad744e9

The idea in one minute

Most questions about code do not need an AI. “Where is applyBudget defined?” has an exact answer in the code itself. Eulix therefore does the expensive work once, up front: it reads every file and records what is in it and how the pieces connect. After that:

  • Factual questions (locations, callers, entry points, dependencies, metrics, TODOs) are answered straight from that record, instantly, with no chance of invention.
  • Questions that need understanding (“why does the budget code split tokens this way?”) go through retrieval: Eulix picks the few pieces of code that matter, fits them into a model’s limited window, and asks the model to explain them, with instructions to say what it cannot see.

Parse once. Navigate many times.

The pieces

your repository
      │
      ▼
 eulix_parser ──── uses ──── libeulix          read code → knowledge base (JSON files)
      │                       (Rust)           and a call graph (PRISM)
      ▼
 eulix_embed                                   knowledge base → vectors (optional)
      │
      ▼
 .eulix/  ◄──────────────────────────────┐
      │                                  │
      ▼                                  │
 eulix  (Go)  ── chat / query / mcp ─────┘     route the question, retrieve code,
      │                                        build the prompt, call the model
      ├── vizeulize                             browse the call graph
      └── MCP server ──► AI coding agents
Piece What it is Docs
libeulix The Rust engine that parses nine languages and links calls to definitions. libeulix
eulix-parser The command-line program that runs it over a repository and writes the JSON files. Fast on large code because it saves work to disk and skips unchanged files. eulix-parser, Spill to Disk, Incremental Parsing
eulix-embed The Python tool that turns code into vectors so questions can match code by meaning. eulix-embed
eulix-cli The eulix program you use: setup, analysis, chat, one-shot queries, the MCP server. Contains the question router and the two retrieval engines. eulix-cli
vizEulize A browser page that draws the call graph. vizEulize
eulix-lsp A language server built on libeulix. Not documented here yet.

Try it

Install first: see Installation.

cd your-project
eulix init        # create .eulix/, .euignore and eulix.toml (a short wizard)
eulix analyze     # parse the code and build vectors; fast when little has changed
eulix chat        # ask questions in a terminal interface

Other ways in:

eulix query "who calls applyBudget?"   # print the answer, or the context prompt for a model
eulix mcp --repo .                      # let an AI coding agent query your code
eulix vizeulize                         # explore the call graph in a browser

What to know before you rely on it

  • The call graph is approximate. It links calls by names and scopes, not by types. Treat edges as strong hints, not proof. See PRISM v2, and Call-Graph Accuracy for how often its edges are right in each language: function calls are reliable everywhere, method calls are often missed, and Java gets no edges from PRISM v2.
  • Some languages are better covered than others. Method calls resolve best in Python, TypeScript and Rust today. See Known Issues.
  • The model is optional and replaceable. Local models (Ollama, LM Studio, vLLM) and many hosted providers work. See Llm.
  • Your code stays on your machine unless you choose a hosted model, in which case the retrieved snippets are sent to that provider.
  • Parts of this project are in beta: the MCP server, the chat screen, and retrieval tuning.

Where to go next

You are… Start with
New and curious This page, then eulix-cli Overview
Setting up a project Config and eulix-parser Overview
Connecting an AI agent MCP Overview
Wanting to understand answers Retrieval Overview
Working on the code libeulix Public API, Known Issues