AI Chat
Your AI already knows the answer — because it knows your work.
Multi-provider AI chat grounded in your specific knowledge base via Graph RAG. Every answer cites the actual notes it used. You can ask questions that span years of your own writing — and get precise, sourced answers in seconds.
The Problem
Generic AI (ChatGPT, Claude) doesn't know anything about you. Every session starts from zero. The AI that knows your work, your research, and your history is dramatically more useful than a general assistant.
The Solution
Adaptive RAG with 4 retrieval modes (vector, summary, rerank, chronological) selected by intent classification. Graph RAG expands results with 1-hop knowledge graph traversal. Multi-hop chain-of-thought RAG for complex queries (3 hops, 4s budget). Contradiction detection. Function calling: create tasks, save notes, search memories from chat. Supports GPT-4, Claude, Gemini, Groq (free), or local Ollama.
Key benefits
Answers grounded in YOUR notes — not the internet, not a generic LLM
Ask in natural language: 'What did I learn about transformers last month?'
Multi-hop reasoning traces connections across notes you forgot you wrote
5 AI providers: GPT-4, Claude, Gemini, Groq (free), local Ollama (private)
Function calling: AI creates tasks, saves notes, and searches memories on your behalf
Bring your own API key (BYOK) for any cloud provider
Technical Depth
Query intent classification determines retrieval mode before any LLM call — zero latency heuristic. Graph RAG: 1-hop traversal retrieves related nodes, expanding context by 40-120%. Multi-hop CoT: retrieve → reason → retrieve again, up to 3 hops with 4s total budget. Contradiction detection via semantic negation detection across retrieved chunks.