tl;dr — I Built a Local AI That Turned my friend's Chaotic WhatsApp Group Chats Into Calendar Events
Описание и применение
tl;dr – локальная LLM, которая читает чат WhatsApp и извлекает дедлайны, события и изменения планов, выдавая чистый резюме и календарный файл.
Задача
Atharva struggled to find college assignment deadlines buried in a 1,000+ message WhatsApp group chat.
Результат
The tool extracted 15 dated events from a 1,178‑message chat and flagged 18 with unclear dates, producing a clean summary and calendar file. Precision was 78% with thinking on, 100% without, and time/date accuracy ranged 57–71%.
Рабочий процесс
- Export WhatsApp chat to a text file
- Run local Gemma 4 model via Ollama with temperature 0 and seed 0
- Parse chat into windows, overlapping to capture boundary changes
- Model extracts items, assigns provisional IDs
- Reconcile step compares titles and IDs to merge or duplicate events
- Resolve relative dates (e.g., "kal", "parso") to absolute dates
- Render final markdown summary and iCalendar file
- Output includes message IDs for traceability
Доказательства
tl;dr reads a WhatsApp export, extracts only the deadlines, events and changes of plan, and gives you a clean summary plus a calendar file you can import.
The model is Gemma 4 (gemma4:e4b), running locally through Ollama with temperature 0, seed 0 and output forced into a JSON schema.
To measure it I hand‑labelled 7 items from the sample chat and wrote a small scorer.
It found 15 dated events and flagged 18 more as "date unclear".
- Дата записи: 2026-10-03T22:41:08.814Z