Build an agentic app with eve, Next.js, and FastAPI using Vercel Services.
Connect an AI chat interface to Python functions and application data. This template combines an eve agent, a Next.js frontend, and a FastAPI backend, with a working task manager that you can use through chat or a dedicated page.
Use it as a starting point for internal tools and assistants that need to read records, update data, or run calculations. The agent's tools are defined in TypeScript and call Python endpoints, so you can keep your backend logic in Python while building the interface in React.
The included tools demonstrate how an agent calls backend functions and presents their results. The statistics tool calculates measures such as mean, median, and quartiles. The text analysis tool returns word counts, estimated reading time, readability, and frequently used words.
FastAPI validates requests and runs the calculations. The chat displays the returned data in dedicated cards. Extend this pattern with Python endpoints for your own application logic.
The agent and the Tasks page use the same FastAPI endpoints and database. People can create, edit, complete, and delete tasks through either interface. When the agent calls its deletion tool, it waits for approval before removing the task.
Tasks are stored in Neon Postgres on Vercel, with SQLModel defining the database model. Local development falls back to SQLite when no database connection is configured.
Alongside task and analysis cards, the agent can present answers as tables, metrics, and other components through json-render. You define the component catalog and React renderers, and the template validates the agent's structured output before displaying it.
The chat also includes follow-up question forms and resumable conversations. Customize the agent's instructions, tools, and components to fit the work your app supports.
All three services deploy in one project through Vercel Services. The FastAPI backend has no public route; the agent and Next.js server reach it through internal service bindings.
During setup, configure Neon and set the EVE_CHAT_PASSWORD environment variable for browser access. Everyone with the password shares the same task list and can access shared conversation links. You can replace this access model with your own authentication provider as you develop the app.
