Hub

AI Lab · Hugging Face · RAG

Real models. Optional cloud. Local-first RAG.

Attractive AI features without forcing accounts: a local retrieval pipeline always works, and Hugging Face Inference models unlock summarization, sentiment, embeddings, and generative answers when you paste your own free token.

Hugging Face token (optional)

Stored only in your browser localStorage. Used for summarization (facebook/bart-large-cnn), sentiment, embeddings, and generative RAG. Get a free token at huggingface.co/settings/tokens.

No token — local TF-IDF RAG still works offline.

Tool RAG Assistant

Ask questions over DevToolkit Hub docs with TF-IDF or HF embeddings.

Document RAG

Paste your notes, chunk them, retrieve, and answer with Flan-T5.

AI Summarizer

facebook/bart-large-cnn via Hugging Face Inference.

Sentiment + Themes

DistilBERT sentiment and BART-MNLI zero-shot labels.

Semantic Similarity

MiniLM embeddings + cosine similarity — RAG building block.

AI Text Humanizer

On-device rewrite for stiff / AI-sounding drafts.

Pipeline shape

  1. Chunk knowledge (tool docs or your paste)
  2. Retrieve with TF-IDF (offline) or MiniLM embeddings (HF)
  3. Optionally generate with Flan-T5 using retrieved context