"I'll send it by Friday." — Everyone says it. Nobody tracks it. PromiseKeeper reads your chats & emails and builds the to-do list you forgot to write.
Every day, real commitments get buried in WhatsApp threads, emails, and meeting notes: "I'll call the landlord tomorrow", "can you sign this by Friday?", "remind me to pay rent on the 1st." Missed promises cost friendships, money, and jobs, and nobody wants to copy-paste them into a task app.
Paste any conversation. PromiseKeeper finds the commitments, works out who owes what, parses the deadline (even "next Tuesday at 5pm", "in 2 weeks", "before the 20th"), ranks urgency, and exports a calendar file with reminders.
$ python promisekeeper.py --demo
YOU OWE (3)
🔴 2026-11-04 09:00 I'll start tomorrow at 9am
🟡 2026-11-06 Can you review the contract by Friday?
🟢 2026-11-20 I need to renew my passport before the 20th
OWED TO YOU (3)
🔴 2026-11-03 Please don't forget to call the landlord today (Sam)
🟡 2026-11-10 I'll book the venue next Tuesday and let you know ASAP (Sam)
🟢 2026-12-01 Remind me to pay rent on the 1st of December (Sam)
| PromiseKeeper | Typical AI to-do apps | |
|---|---|---|
| Privacy | 100% offline, nothing leaves your machine | Uploads your private chats |
| Cost | Free, zero dependencies | API keys & subscriptions |
| Direction-aware | Separates "you owe" vs "owed to you" | Flat task list |
| Output | .ics calendar, JSON, terminal |
Locked in their app |
| Speed | Milliseconds | Network round-trips |
git clone https://github.com/theinsidecoder/promisekeeper && cd promisekeeper
python promisekeeper.py --demo # try it
python promisekeeper.py chat.txt --me Sam --ics promises.ics # real chat -> calendar
cat email.txt | python promisekeeper.py --json # pipe anything
python -m unittest -v # run the testsSupports plain text, Name: message chats, and WhatsApp exports (12/03/2026, 10:15 - Sam: ...).
Import the .ics into Google Calendar, Apple Calendar, or Outlook.
- Segment: split into speakers and sentences.
- Detect intent: commitment patterns (I'll, let me, I need to) vs request patterns (can you, please, don't forget).
- Assign ownership: commitments bind the speaker; requests bind the other person;
--medecides the direction. - Parse time: a rule-based temporal parser handles relative dates, weekdays (
nextvsthis), calendar dates with year roll-over, clock times, and phrases like EOD and end of month. - Prioritise: due within 24h or urgent words (ASAP) → 🔴, within 3 days → 🟡, else 🟢.
- Export: table, JSON, or
.icswith a 1-hour-before alarm.
Deterministic, explainable, testable. Every decision traces back to a readable rule.
promisekeeper.py # parser, extractor, CLI, ICS exporter (~200 lines)
test_promisekeeper.py # 11 unit tests covering dates, ownership, exports
- Optional LLM mode for ambiguous phrasing
- Multi-language date parsing
-
watchmode for live chat exports
MIT. PRs welcome.