Personal RAG Assistant (Rafcio Assistant)

A modular Python-based agentic assistant designed for RAG (Retrieval-Augmented Generation), tool calling, and LLM evaluation. This project serves as a hands-on technical demonstration of building local-first AI agents.

🚀 Project Goals

• Local-First AI: Powered by LM Studio for full privacy and control.
• Advanced RAG: Implementation of hybrid search (Vector + BM25) and reciprocal rank fusion.
• Agentic Workflows: Multi-step reasoning using LangGraph and MCP (Model Context Protocol).
• Quality Focused: Automated evaluation using DeepEval for answer relevancy and faithfulness.

🏗️ Technology Stack

• Runtime: Python 3.11+ with uv
• Orchestration: LangChain & LangGraph
• Vector Store: ChromaDB (Embedded)
• LLM Provider: LM Studio (OpenAI-compatible API)
• Evaluation: DeepEval
• Linters: Ruff, MyPy, PyLint, Black (via Pre-commit)

🛠️ Setup & Configuration

1. Prerequisites

• Install uv
• Install LM Studio

2. Model Setup

Download and load the following models in LM Studio:

• LLM: gpt-oss:20b (or similar for chat/reasoning)
• Embeddings: text-embedding-bge-m3 (or gemma-embedding)

Ensure the local server is running at http://localhost:1234.

3. Installation

# Clone the repository
# (Assuming you are already in the project directory)

# Install dependencies
uv sync

# Install pre-commit hooks
uv run pre-commit install

4. Configuration

Create a .env file (see template in technical_specification.md or copy an existing one):

API_URL=http://127.0.0.1:1234
LLM_MODEL=openai/gpt-oss-20b
EMB_MODEL=text-embedding-bge-m3

# Feature Flags
ENABLE_HYBRID_SEARCH=true
ENABLE_MCP_TOOLS=true

📂 Project Structure

• src/: Core logic (config, embeddings, vectorstore, agents)
• knowledge_base/: Directory for documents to ingest (.txt, .md, .pdf)
• workspace/: Working area for agent file operations
• tests/: Quality and connectivity tests
• scripts/: Utility scripts (ingestion, linting)

🧪 Testing & Quality

Run the linting suite:

./scripts/lint.sh

Run specific tests:

# Test settings
uv run python tests/test_config.py

# Test embeddings connectivity
uv run pytest tests/test_embeddings.py -s

📜 License

MIT