🧠 RAG Agent
AI-powered document intelligence — fully local, on-device
Hybrid Search
BM25 + Dense
LLM Reranker
ReAct Agent
PDF · DOCX · XLSX · PPTX · CSV · MD · HTML · TXT
Select one or more files. Supported: PDF, TXT, DOCX, XLSX, XLS, PPTX, CSV, MD, HTML.
Paste any public URL — webpage, PDF, DOCX, etc. The content will be chunked and added to the knowledge base.
Search the web for a topic and index the top results. Uses DuckDuckGo — no API key needed.
Each result URL is crawled to the chosen depth, same as the URL crawl.
Each result URL is crawled to the chosen depth, same as the URL crawl.
1 20
1 3
1 10
⚠️ Topic filter is required for Wikipedia — without it, depth ≥ 2 follows every inline link and fills your page budget in seconds.
Set topic = article name, e.g.
On HF free CPU: keep depth=1, max 10 pages to avoid timeout.
Set topic = article name, e.g.
Elizabeth_Olsen.On HF free CPU: keep depth=1, max 10 pages to avoid timeout.
1 3
1 50
Index these file types:
⏳ Initializing...
Chunks in knowledge base: 0
Agent Tools
Use Agent mode to activate these tools.
🔍 rag_search — search your documents
🧮 calculator — evaluate math expressions
📝 summarise — summarise a document section
💬 sentiment — analyse tone & sentiment
🌐 translate — translate to any language
✅ finish — return the final answer
Pipeline
Query type, confidence score, and retrieved chunks appear here after each query.