Chat Your History — Adaptive Q&A Bot

Situation

The client needed an adaptive Q&A flow where the bot asks follow-up questions based on user responses. Building this from scratch would have meant rebuilding conversation management, state handling, and flow logic entirely.

Options Considered

• Build a custom chatbot from scratch — full control, but significantly longer development time
• Use Botpress Cloud — define and connect conversation flows using nodes and cards without rebuilding the conversation engine

Decision

Used Botpress. It let me focus on the question logic and flow design rather than the infrastructure. The tradeoff was that prompt tuning required constant iteration as the client's requirements evolved — eventually the client started updating prompts himself to get the exact behavior he wanted.

Situation

Botpress's built-in database was too expensive for the project's budget. There was also a requirement that if a user leaves mid-conversation, they should resume exactly where they left off on their next visit.

Options Considered

• Use Botpress's native database — simpler integration, but costly
• Build a custom backend with an external database and handle state manually

Decision

Built a custom backend with Neon PostgreSQL. Conversation state is tracked via Botpress hooks and custom APIs, enabling seamless resume without relying on Botpress's storage — at a fraction of the cost.

Situation

A requirement came in that users should be able to upload an image and have the bot understand what it is about and continue the conversation from that context.

Decision

Built a custom API endpoint that sends the uploaded image to OpenAI's API, interprets the result, and feeds it back into the bot flow — extending Botpress's capabilities without modifying the core flow structure.