LocalLLM Hub (v1 - Legacy)

⚠️ This is the legacy v1 codebase. Active development has moved to localllm-hub-v2.

Why v2?

v1 was a monolithic Python service with custom routing, agent loops, and RAG implementation. v2 rebuilds the entire system on industry-standard frameworks:

• LangGraph for agent orchestration (vs custom loops)
• LlamaIndex for RAG pipeline (vs custom implementation)
• LiteLLM for model routing (vs manual routing)
• MCP for tool protocol (vs custom tools)

Result: More maintainable, extensible, and production-ready.

What's Here

This codebase contains the original implementation with:

• Custom embedding service
• Manual task routing
• Basic semantic search
• Email classification
• Transcription pipeline

Migration

If you're using v1, see the v2 migration guide.

For new projects, start with v2: https://github.com/kqb/localllm-hub

Original Components

• embeddings/ — Unified embedding service (Ollama mxbai-embed-large)
• triage/ — Task router + urgency classifier
• transcriber/ — Batch voice memo transcription (whisper.cpp)
• classifier/ — Email/content classifier (local LLM)
• search/ — Unified semantic search across all data sources

Archive Status

This codebase is kept for:

• Historical reference
• Comparison benchmarking
• Migration support

Active development: https://github.com/kqb/localllm-hub (v2 branch)

License

MIT