AI RAG Assistant
Ask DevToolkit Hub docs with local or Hugging Face RAG
About AI RAG Assistant
How AI RAG Assistant Works
Your question retrieves the best tool-doc passages, then optionally a Hugging Face generation model writes a concise answer from that context.
The AI RAG Assistant retrieves DevToolkit Hub knowledge with a real retrieval-augmented pipeline. Offline TF-IDF always works; with a Hugging Face token you can use MiniLM embeddings and Flan-T5 generative answers.
Frequently asked questions
Do I need a Hugging Face token?+
No for basic retrieval. A free HF token unlocks semantic embeddings and generative answers.
Where is my HF token stored?+
Only in your browser localStorage. DevToolkit Hub never receives it.
What models are used?+
sentence-transformers/all-MiniLM-L6-v2 for embeddings and google/flan-t5-base for optional generation.
Limitations
Public HF Inference rate limits apply. Large or cold models may return a temporary loading error — retry after a few seconds.