LocalForge

LocalForge is an offline-first AI development environment that keeps your code, embeddings, and AI workflows fully local. It pairs a Python backend that talks to a local LLM (Ollama) with a React-based IDE-style frontend featuring a file explorer, Monaco editor, embedding search, and local AI assistant.

Features

• Local LLM integration via Ollama for summarization, suggestions, and Q&A.
• Offline vector database using SQLite + embeddings for code search.
• Local indexing workflows (server-side path indexing or browser-based file import).
• Privacy-first: no external API calls, all data stays on your machine.
• Docker-ready for optional local container deployment.

Architecture

frontend (React + Monaco)  -->  backend (FastAPI)
                                   |-> SQLite vector store
                                   |-> local embeddings (sentence-transformers)
                                   |-> local LLM (Ollama)

Requirements

• Python 3.11+
• Node.js 18+ (for frontend)
• Ollama running locally (offline) with a downloaded model

Backend Setup

cd backend
python -m venv .venv
source .venv/bin/activate
pip install -r requirements.txt
uvicorn app:app --host 0.0.0.0 --port 8000

Offline Embeddings

• By default, the backend uses sentence-transformers/all-MiniLM-L6-v2.
• For fully offline usage, download the model once and set:

export EMBEDDING_MODEL_PATH=/path/to/local/model

If sentence-transformers is unavailable, the backend falls back to a hashing vectorizer (still offline, but lower quality).

Frontend Setup

cd frontend
npm install
npm run dev

Open http://localhost:5173.

Using the Frontend

1. Select a local folder using the Project Files picker.
2. Click Index in LocalForge to embed and store your files.
3. Use Embedding Search to query the indexed code.
4. Ask the Local LLM Assistant for summaries or suggestions.

Example Scripts

python backend/scripts/index_project.py /path/to/project
python backend/scripts/query_llm.py "Summarize the indexing workflow"

API Endpoints

• POST /index: Index files from a local path on the backend host.
• POST /add-batch: Add documents from the frontend or scripts.
• POST /search: Search indexed content by embedding similarity.
• POST /query: Retrieve context + query the local LLM.

Docker (Optional)

docker compose up --build

• Backend: http://localhost:8000
• Frontend: http://localhost:4173

Ensure Ollama is running on the host with your model downloaded. The compose file uses host.docker.internal so the backend can reach your host Ollama server.

Offline Workflow Tips

• Download your Ollama model while online, then disconnect.
• Download the sentence-transformers model to EMBEDDING_MODEL_PATH before going offline.
• Keep your code, embeddings, and queries on-device.

Roadmap Ideas

• Add filesystem watching for automatic background indexing.
• Add multi-project workspaces and saved sessions.
• Add inline code actions driven by the local LLM.