Commercial project

AI Message Analysis & Response Assistant

An AI tool that identifies relevant requests in selected Telegram communities and generates context-aware replies using a knowledge base.

The task: Recognise relevant service requests in message streams and make responses consistent with product knowledge.
My role: Development of message classification, AI response generation and persistent conversation state.
Delivered: A configurable Telegram processing pipeline with intent rules, knowledge-based generation, conversation history and sending limits.

How it works

Turn a message stream into relevant conversations.

01
Find relevant requests
The tool reads configured chats and channels. Intent rules and exclusion filters separate relevant requests from unrelated messages.
02
Generate a contextual response
A language model uses the selected knowledge base and the current conversation history to generate a reply.
03
Manage the conversation
The system records successful replies, limits the length of each dialogue and respects account-level sending intervals.

My contribution

Explicit intent rules decide which messages qualify; AI handles the wording of the response using relevant context.

Curated product information gives the model a consistent reference for answers. Generated responses still require quality evaluation.

Configured fallback providers and request timeouts handle unavailable models without sending an empty response.

Dry-run mode processes incoming messages without posting replies, with separate test history and counters.

For the technical team

Python Telegram AI SQLite

Have a similar task?

Describe what you need in a few sentences. I’ll help clarify the scope and propose the next step.

First we discuss the task, then agree on the scope, estimate and milestones.