🧠 CodeRecall

Context-aware local AI assistant for developers using Ollama, LanceDB, and VS Code’s Continue extension.

CodeRecall ingests your entire codebase — including Git history and diffs — into a local vector database (LanceDB), enabling RAG-augmented queries via Ollama models right inside VS Code.

No cloud APIs. No latency. Full control.

🚀 Features

🔍 Semantic code search across multiple languages
📜 Git commit + diff embedding for code evolution awareness
🤖 RAG integration with local Ollama models (e.g. LLaMA 3)
💻 VS Code Continue extension support
⚙️ Configurable with a simple config.ini

🧩 Project Structure

CodeRecall/
├── lancedb_ingest.py           # Ingest codebase + Git into LanceDB
├── lancedb_context_provider.py # VS Code Continue context provider
├── config.ini.example          # Ollama + LanceDB settings
├── lancedb-data/               # LanceDB persistence directory
└── config.yaml.example         # Continue extension config

🔧 Setup

It seems you want to update the section 'Install dependencies' in your README.md file.

Here's an updated version of that section:

### 1. Install dependencies

pip install fastapi uvicorn pydantic lancedb ollama

Make sure you have:

✨ FastAPI installed
🚀 Uvicorn installed as the ASGI server
🔥 Pydantic for data modeling and validation
💻 Lancedb for vector database operations
🦙 Ollama installed and running
✅ Continue Extension for VS Code
🐙 Git repo initialized (optional but recommended)

2. Configure `config.ini`

[[ollama]
url = http://localhost:11434

[lancedb]
persist_directory = ./lancedb-data

[s3]
enable = True
bucket_name = my-s3-bucket 
access_key_id = my-access-key
secret_access_key = my-secret-key
region = us-east-1
# Optional, if using third party s3 providers
endpoint = http://minio:9000 

[server]
host = 0.0.0.0
port = 8080

📥 Ingest your project

python lancedb_ingest.py

This loads:

Source code in common languages
Markdown and text docs
Git commit messages and full diffs

🧠 Add as a VS Code Context Provider

`config.yaml` for Continue

name: Local Assistant
version: 1.0.0
schema: v1
models: 
  - name: Ollama Autodetect
    provider: ollama
    model: AUTODETECT
    apiBase: http://localhost:11434
  - name: Ollama Autocomplete
    provider: ollama
    model: qwen2.5-coder:1.5b-base
    apiBase: http://localhost:11434
        roles:
      - autocomplete
  - name: Nomic Embed Text
    provider: ollama
    model: nomic-embed-text
    apiBase: http://localhost:11434
    roles:
      - embed
context:
  - provider: code
  - provider: docs
  - provider: diff
  - provider: terminal
  - provider: problems
  - provider: folder
  - provider: codebase
  # LanceDB Context Provider
  - provider: http
    params:
      url: http://localhost/retrieve

✨ Usage

Launch VS Code.
Open the Continue sidebar.
Set "@HTTP" as your context provider.
Ask your model questions about your codebase, architecture, or commits.

Example prompt:

“How does the Git ingestion pipeline work?”

📌 Notes

Default embedding model is nomic-embed-text (via Ollama).
Change n_results in lancedb_context_provider.py for broader/narrower context.
Works offline, no API keys required.

🧪 Roadmap Ideas

Add OpenAPI spec ingestion
Enable full-text search fallback
Support for multi-repo ingestion
Optional chunking for large diffs or files

🛡 License

Creative Commons Attribution-NonCommercial 4.0 International license.

3.9 KiB Raw Blame History Unescape Escape