Tyler-d

AI-powered local-first "second brain" optimized for Polish and privacy.

Project Structure

• services/orchestrator: Core logic and RAG management (Rust).
• services/ocr: Document processing service (Python/gRPC).
• services/gui: Web-based user interface (Vue 3 / Nuxt UI v4).
• services/cli: Command-line interface for headless operations (Rust).
• services/llm_gateway: Unified API for local and remote LLM providers (Rust).
• protos: Shared gRPC definitions.
• docker-compose.yml: Local infrastructure (Ollama, Qdrant).

Interfaces

GUI

Modern, accessible web interface built with Nuxt UI v4 and Tailwind CSS v4.

• Features: Chat with knowledge base, file ingestion (PDF/Images), and text entry.
• Development: cd services/gui && npm install && npm run dev

CLI

Fast and lightweight command-line tool for ingestion and querying.

• Commands: ingest, ask, models.
• Development: cd services/cli && cargo run -- --help

Getting Started

1. Start Infrastructure

docker compose up -d

2. Prepare Local LLM

Ensure Ollama is running, then pull the models (Bielik for text, bge-m3 for embeddings):

docker exec -it tyler-d-ollama ollama pull speakleash/bielik-v3-4.5b-instruct:q4_k_m
docker exec -it tyler-d-ollama ollama pull bge-m3

3. Development

• Orchestrator: cd services/orchestrator && cargo build
• OCR: cd services/ocr && pip install -r requirements.txt
• GUI: cd services/gui && npm install && npm run dev
• CLI: cd services/cli && cargo build

License

Non-commercial / Personal use.