FilVault Bridge

FilVault Bridge is a single-page web demo that shows how AI conversation memory can live outside any one LLM provider, using Filecoin (via Filebase/IPFS) as the storage layer. What it does: The app walks through a five-step flow:

Chat with Model A — You talk with Llama 3.1 8B (HuggingFace) about a topic: decisions, preferences, facts. Summarize — A second LLM call extracts structured memory: topic, key facts, decisions, preferences, and a short summary. Save to Filecoin — That JSON is uploaded to Filebase (S3-compatible IPFS/Filecoin storage). You get back a CID (content address). Load into Model B — You paste the CID into a second chat panel running Llama 3.3 70B (Groq). The app fetches the JSON and injects it into Model B’s system prompt. Prove portability — You ask Model B a follow-up. It should answer with full context from the Model A conversation, even though it’s a different model on a different provider. The core pattern is: LLM → structured JSON → Filecoin → CID → different LLM. Its for Anyone frustrated with AI lock-in