wtf is octo-smart-summary?

octo-smart-summary is a small Go service that reads conversation threads from the OCTO workplace platform and uses an AI language model to produce a structured summary. Instead of leaving team members to manually catch up on long group chats or meeting transcripts, this service distills them into a scannable brief listing key decisions, unanswered questions, and suggested follow-up actions. The service is part of the larger OCTO ecosystem, an open source workplace platform that includes a backend server, task management, mobile clients, and admin tools. When a conversation ID is submitted, the service fetches the transcript from octo-server, breaks it into chunks that fit the AI model's context window while preserving speaker and time boundaries, sends those chunks to the language model with a summarization prompt, parses the structured JSON output, adds participant and duration details, and returns the result. If the model returns malformed JSON, it retries the prompt once. A practical design choice is that the LLM endpoint is fully configurable through an environment variable, so you can point it at any service that speaks the OpenAI chat completions API format: a self-hosted model, a commercial API, or a gateway in between. The output is always structured JSON rather than free-form prose, so other OCTO services such as the task manager can consume it directly without re-parsing text. The service exposes four HTTP endpoints: one for synchronous summarization, one for streaming output, one health check, and one metrics endpoint. It is written in Go and licensed under Apache 2.0.