🦀 Rust LLM Agent — Local Developer Assistant

This project is a local Rust-powered AI assistant that uses embedded documentation, agent tools, and a local LLM to help answer Rust programming questions — with context-aware code generation and crate discovery.

---

📦 Features

• RAG (Retrieval-Augmented Generation) with ChromaDB
• 🕵️‍♂️ Agents for:
  • Crate docs (docs.rs)
  • GitHub repo search + README parsing
  • Standard library and Rust Book content
• 🤖 Local LLM (LLaMA3 / Mistral / OpenHermes) integration
• 🧠 Chat interface with memory + tool calling
• 📁 Modular source loading from local docs (rustup, crates, scraped pages)

---

🛠️ Setup

1. Clone the project

git clone https://github.com/bradleyd/rustopedia.git
cd rustopedia

2. Set up Python (for embeddings + ChromaDB)

cd rag 
python3 -m venv venv
source venv/bin/activate
pip install -r requirements.txt

---

🧠 Load Documentation into Vector DB

You can load any Rust content (books, docs, READMEs) into ChromaDB.

Supported Sources

• Local docs via rustup (HTML extraction + chunking)
• Crate docs from docs.rs (automated fetching)
• Rust book or stdlib
• GitHub crate READMEs

✅ Quick Setup - Load All Documentation

Option 1: Load Rust core docs (Book + Stdlib)

cd rag
chmod +x rustup_docs.sh
./rustup_docs.sh

Option 2: Load popular crate documentation

cd rag
chmod +x crate_docs.sh
./crate_docs.sh all  # Fetches serde, tokio, clap, etc.

Option 3: Load specific crate docs

./crate_docs.sh fetch tokio     # Fetch specific crate
./crate_docs.sh embed           # Embed all fetched crates

Manual Document Loading

Place .md or .txt files into sample_docs/ under collection folders:

sample_docs/rust-book/
sample_docs/rust-docs/
sample_docs/crates/

Then run:

python embed_docs.py --dir ../sample_docs/your-collection --collection your-name --chunk-size 800 --overlap 100

All files will be embedded with intelligent chunking and stored in ./chroma_db.

---

🔍 Querying the Knowledge Base

You can manually query docs using:

python query_docs.py "How do I deserialize JSON into an enum?"

This returns the most relevant chunks across all embedded docs.

---

💬 Running the Assistant (Chat Mode)

Launch the LLM shell with agent orchestration:

cargo run --bin llm_shell

You can now ask things like:

• "What’s the best way to build a CLI app in Rust?"
• "How do I initialize a Vec with a capacity?"
• "Give me an example using serde to serialize structs"

The system will:

• Search ChromaDB
• Route queries to agents (GitHub, crate, docs)
• Return helpful, idiomatic answers using a local LLM

---

🧱 Agent Overview

GitHub Discovery Agent (Rust)

• Classifies queries into topics (e.g. cli, web)
• Uses GitHub API to find top crates
• Pulls README and returns as context

Docs Agent

• Fetches documentation from docs.rs or stdlib/book
• Extracts relevant examples
• Adds results to Chroma

---

📦 Rust Project Structure

rustopedia/
├── llm_shell/                    # Main chat interface with agent orchestration
├── rag_engine/                   # Core RAG functionality
├── rag_server/                   # HTTP server for RAG queries
├── agents/
│   ├── github_agent/            # GitHub repository discovery
│   ├── docs_agent/              # Documentation fetching
│   └── crate_agent/             # Crate.io integration
├── /
│   ├── embed_docs.py            # Document embedding with chunking
│   ├── query_docs.py            # Manual query interface
│   ├── rustup_docs.sh           # Automated Rust docs extraction
│   ├── crate_docs.sh            # Crate documentation manager
│   └── sample_docs/             # Document storage (gitignored)
└── chroma_db/                   # Persistent vector database

---

🧪 Testing Examples

# Load Rust core documentation
cd rag
./rustup_docs.sh

# Load popular crate docs
./crate_docs.sh all

# Query manually
python query_docs.py "How do I create a VecDeque?"

# Chat with LLM
cd ../
cargo run

---

📎 TODO (next ideas)

• Allow agent results to include multiple sources
• Stream LLM responses with token buffering
• Add RAG-based fallback when LLM confidence is low
• Use embeddings to determine best agent (vs. keyword match)

---

🔧 Advanced Usage

Crate Documentation Management

The crate_docs.sh script provides comprehensive crate documentation management:

# View all available commands
./crate_docs.sh

# Fetch specific crates
./crate_docs.sh fetch serde
./crate_docs.sh fetch tokio

# Fetch popular crates automatically
./crate_docs.sh popular

# Re-embed all fetched crates
./crate_docs.sh embed

# View collection statistics
./crate_docs.sh stats

# Clear and rebuild everything
./crate_docs.sh all

Custom Documentation Sources

Install additional Rust docs locally:

rustup component add rust-docs

The rustup_docs.sh script will automatically extract and embed them with optimal chunking.

---

License

MIT