Ask questions across three years of a Telegram group
At a glance
A group member uses source-lm and NotebookLM to search exported conversations with citations. This workflow is listed under Life Admin, Community Knowledge, Document Search.
- Agents / tools
- source-lm, Telegram Desktop, NotebookLM
SourceBuilding a tool for these workflows? See weekly rates.
What they did
1) Install the source-lm Chrome extension and sign into NotebookLM in Chrome. Use Telegram Desktop, which supports JSON and HTML exports; phone and web clients cannot export the history. 2) In the target conversation, export its chat history and keep result.json, or all the messages*.html pages together. 3) Open the destination notebook and use source-lm's extension popup to select the export files. Convert and upload the generated Markdown sources; the default packing budget is 400,000 words per file. 4) Ask which providers were repeatedly recommended, or how an administrative process changed over time. Open the answer citations to inspect the original messages and senders. 5) For a later update, re-export the chat into the same notebook and run the extension again; it reads existing source filenames to upload only newer messages.
How it went
The author's 19.78 MB JSON export became four Markdown sources totaling 10.64 MB.
Worth knowing
Do not rename uploaded sources: their filenames hold the incremental cursor. Media content is not transcribed, and notebook source limits still apply.
