Local-first code intelligence for large codebases.

Overview · Install · Usage · Docs · Benchmarks

---

Eulix is a local-first code intelligence system designed to provide deep reasoning over massive repositories without compromising privacy or speed.

It orchestrates a high-performance pipeline of Go and Rust to bridge the gap between raw static analysis and LLM-powered insights. By combining a multi-layer retrieval strategy with rigorous anti-hallucination prompting, Eulix ensures that codebase answers are grounded in your actual source code.

---

Overview

Eulix operates as three focused binaries that work in concert:

Component	Language	Role
eulix	Go	Orchestrator — Manages the CLI, config, and the RAG pipeline
eulix_parser	Rust	Static Analyzer — Extracts symbols, call graphs, and complexity
eulix_embed	Rust	Embedder — Runs transformer models via ONNX with GPU acceleration(supports both rocm and cuda)

Smart Multi-Layer Retrieval

Unlike simple RAG tools, Eulix uses a tiered retrieval pipeline to find the most relevant context before hitting an LLM:

1. Symbol Lookup: Precision matching for functions, classes, and variables.
2. Keyword Search: Traditional lexical matching for specific terms.
3. Semantic Vector Search: Deep contextual matching using local embeddings.
4. Call-Graph Expansion: Traverses relationships to pull in relevant upstream/downstream logic.

Reliable Reasoning

Eulix is built with an anti-hallucination discipline. Our prompts are architected to force the model to cite its sources and strictly adhere to the provided context, minimizing "invented" logic or APIs.

---

Features

• Symbol Indexing — Comprehensive mapping of functions, classes, and source locations.
• Advanced Call Graphs — Maps incoming and outgoing relationships across the entire project.
• Knowledge Base — Captures control-flow structures, error handling blocks, and cyclomatic complexity.
• Local-First — All parsing and embedding happens on your machine. No code leaves your infrastructure.
• High Performance — Rust-powered backend capable of parsing millions of lines in seconds.

Supported languages

Python · Go · C

Note

rust · Typecript · C++ will be supported soon

---

Installation

Prerequisites: Go 1.22+, Rust stable, a Hugging Face account (for model downloads)

Build from source

git clone https://github.com/nurysso/eulix
cd eulix && make install

# Or you can do this if you want to test each bin
# Build the CLI
go build -o eulix ./cmd/eulix

# Build the parser
cd eulix-parser && cargo build --release

# Build embedder
# Go back to root of the project and
cd eulix-embed && cargo build --release --feature rocm
# use cuda instead of rocm if you have nvidia gpu

# try make help for other usefull commands during building or installing

---

Usage

1. Initialize a project

cd my-project
eulix init

2. Analyze the codebase

eulix analyze

This triggers the parser and embedding pipeline, generating a .eulix folder which will be used as knowledge base for llm.

3. Chat with your code

eulix chat

Open's an interactive session to query your codebase using the multi-layer retrieval pipeline.

---

CLI Reference

eulix (Go)

The main entry point for orchestration.

• init : Initialize eulix in current directory
• analyze : Analyze codebase and generate knowledge base
• chat : Start interactive chat interface
• cache : Manage cache entries
• config : Manage eulix configuration
• history : View query history interactively
• version : Displays version of eulix and eulix_parser, eulix_embed
• glados : Checks for errors in knowledge base and embeddings size
• aspirine : tries to fix embedings.bin and kb MEANT TO BE USED IN TEST

eulix_parser (Rust)

Fast static analysis tool.

• -r, --root : Project root directory
• -v, --ver : parser version
• -o, --output : Output file for knowledge base [default: knowledge_base.json]
• -t, --threads : Number of threads for parallel parsing [default: 4]
• -v, --verbose : Verbose output
• -l, --languages : Languages to parse (comma-separated, or "all") [default: all]
• --no-analyze : Skip analysis phase (faster, only parse files)
• --euignore : Path to custom .euignore file (defaults to /.euignore)
• -h, --help : Print help
• -V, --version : Print version

eulix_embed (Rust)

Vector generation via ONNX. Supports sentence-transformers/all-MiniLM-L6-v2, BAAI/bge-small-en-v1.5, BAAI/bge-base-en-v1.5, and more. Native CUDA/ROCm support for high-throughput embedding. eulix_embed [COMMAND] [OPTIONS]

COMMANDS:

• embed : Generate embeddings for knowledge base (default)
• query : Generate embedding for a query string

EMBED OPTIONS:

• -k, --kb-path : Path to knowledge base JSON file
• -o, --output : Output directory for embeddings
• -m, --model : HuggingFace model name or local path

QUERY OPTIONS:

• -q, --query : Query text to embed

• -m, --model : HuggingFace model name or local path

• -f, --format : Output format: json (default) or binary

• -h, --help : Show this help message

• -v, --version : Show version

Benchmarks will be added in docs soon.

Contributing

Contributions are welcome. Please open an issue before submitting a pull request for significant changes.

---

LICENSE