wtf is market-pilot?

Market Pilot is a prototype tool for doing market research with AI, built around the idea that a research summary should always be traceable back to real evidence, not just a confident sounding paragraph. Instead of asking an AI to summarize a topic and trusting the result blindly, the project keeps track of where each finding came from, separates things that were directly stated from things the AI inferred, and tries to make it clear when something is uncertain. It works by collecting public market signals, building anonymized profile summaries called persona cards from that information, and then answering research questions in a way that stays grounded in the collected evidence. The final report highlights where sources agree, where they disagree, what the limitations are, and which signals support each claim. The interface can be viewed in either English or Simplified Chinese, and cached translated reports can be read without re-running the underlying research. The project has a React frontend and a FastAPI backend, with the frontend calling a local API that a development proxy forwards to the backend service. The README is upfront that this is a product prototype: any demo data shown must be clearly marked as sanitized, and using real data connectors or running the AI models yourself requires your own authorized provider credentials, since none are bundled with the project. Running it locally requires Node.js, npm, Python, and Git, plus setting up a Python virtual environment for the backend collector. The README also notes that some research workflows will not work at all until valid API credentials are supplied. The project is released under the Apache 2.0 license and includes guidance on data privacy and responsible use of collected information.