2025
If you work with teams that coordinate on Discord or Slack, you probably know the feeling of opening a channel after a few days and wondering: what actually happened here? Threads sprawl, IDs appear everywhere, and the most important points get buried. So I built What The Chat — a small, modular Python tool that fetches channel history, standardizes user references, and generates structured summaries using local or remote LLMs. You can even chat interactively with the history afterwards.
How it works is simple. First, you point the tool to a channel and a time window. It will collect messages — including threads — and create a clean, human-readable transcript. Then, depending on the timespan, it asks the model to write either a Project Event Update (0–2 days), a Periodical Digest (3–7 days), or a Full Project Status Summary (8+ days). The output is terse and structured, designed for quick catch-up. User mentions are normalized so you don’t get lost in raw IDs.
Under the hood, it’s a small package with a clear separation of concerns: platform integrations for Discord and Slack, LLM services for summarization and chat, utilities for formatting, and light data models for messages and history. It can run as a CLI or be imported into your own project to drive a bot, a web view, or internal tooling. If you prefer privacy and speed, use local models via Ollama; if you prefer capability, switch to remote models like gpt-4o.
... the command changed a bit, correct example is down there 👇
Why I like it is that it removes the friction of context switching. After a week away, I can run a single command and get a clean status: completed items, open actions, who said what, and any risks that bubbled up. If I’m unsure about something, I just ask follow-up questions in the interactive session — the model only sees the collected history, so the answers stay grounded.
You can run it from the terminal or import it. Example CLI usage:
python apps/launch_cli.py --since-days 30 --platform discord --channel 123456789012345678 --chat
And of course, you can save both the summary and the full transcript as Markdown for later reference. If you want to wire it into a custom workflow, the package API exposes Discord/Slack fetchers, a SummarizationService, and a ChatService so you can build on top of it.
Many thanks to open-source toolmakers — this project leans on Ollama, LangChain, and the official Slack SDK / discord.py. If you try What The Chat and find it helpful, feel free to star the repo or open an issue.