TeliChat - Code-centric white-box conversational agent

Hi Product Hunt 👋 I’m Hans, founder of TeliChat. We built TeliChat because we kept seeing the same problem in enterprise customer service and complex business processes: ReAct-style agents are flexible, but often too black-box, slow, expensive, and risky when strict business rules, permissions, state transitions, and external system calls are involved. Workflow agents are more controllable, but they often make conversations rigid. Real users don’t follow a perfect flow. They provide information out of order, correct themselves, jump between topics, or add new requests midway. TeliChat is our attempt to solve this with a more engineering-oriented, white-box conversational agent architecture. The core idea is simple: Let the LLM do what it is good at: * intent recognition * information extraction * natural language generation Let code handle what must be deterministic: * business rules * permission checks * API calls * state management * debugging and traceability TeliChat manages multi-turn conversations through a “ChatTree + InfoItem + Python code” model. This allows teams to build agents that can handle non-linear conversations while still keeping business logic explicit, debuggable, and traceable. It is designed for use cases like ticket changes and cancellations, insurance claims, after-sales support, account opening, approvals, work orders, and other rule-heavy business processes. The core product is not open source yet, but we have published example code on GitHub to show how agents can be built with TeliChat: https://github.com/hanswang73/te... We also shared more of our thinking here: https://telichat.io/blog/telichat We would love to hear your feedback, criticism, and questions. Especially if you have worked with ReAct agents, workflow builders, Rasa CALM, Dify, Claude Code, or similar tools, I’d be very interested to hear what worked for you and what didn’t. Thanks for checking out TeliChat 🙏