Pilotbase
Universal database GUI for SQL, NoSQL, and vector DBs
7 followers
Universal database GUI for SQL, NoSQL, and vector DBs
7 followers
• One web UI for 19+ database types: PostgreSQL, MySQL, MongoDB, Redis, Qdrant, ChromaDB, Weaviate, and more • Self-host with dockerfile, fully open source (MIT license) — zero vendor lock-in • Built-in AI agent (LangGraph + ReAct) that queries databases in plain English • Works with local Ollama or any OpenAI-compatible LLM • Docker Compose setup in 3 minutes • AI Agent re-uses the connection that is already present in Pilotbase to do SELECT queries on databases when required by user.
Finally tried Pilotbase with my Postgres and a local Ollama setup, and the docker compose came together in a couple of minutes like advertised. The plain English querying feels surprisingly natural, even on a messy schema.
How does the AI agent handle schema discovery across the 19+ database types, and does it keep an internal map of table relationships or query it on the fly each time?
How does the AI agent actually decide when to fall back on plain SQL versus pulling from vector stores like Qdrant or ChromaDB, and can you override that routing logic?
How does the AI agent handle schema introspection across the 19+ database types, especially something like Redis or vector stores like Qdrant where the "schema" looks pretty different from a traditional relational DB?