gitwtfhub

wtf is saytrace?

rajdeepkushwaha5/saytrace — explained in plain English

Analysis updated 2026-05-18

0PythonAudience · generalComplexity · 4/5Setup · hard

TL;DR

A fully local tool that listens to your conversations, extracts the promises made in them, and answers questions about who owes what with a citation back to the exact moment it was said.

Mindmap

mindmap
  root((SayTrace))
    What it does
      Tracks spoken promises
      Resolves deadlines
      Cites exact audio moment
    Tech stack
      Python
      faster-whisper
      Qwen via Ollama
      Supermemory Local
      SQLite
    Use cases
      Personal commitment tracking
      Relationship memory
      Follow-up reminders
    Audience
      Individuals
      Privacy conscious users
    Privacy
      Fully local
      No cloud APIs
      User controlled deletion

Code map

Detail Auto

An interactive map of this repo's files and how they connect — its source is parsed live in your browser. Click Visualize to build it.

filefunction / class

Why would anyone build with this?

REASON 1

Capture a conversation and automatically extract who promised what and by when.

REASON 2

Get reminded before a commitment you made is about to be broken.

REASON 3

Ask what was said in a past conversation and get an answer backed by the exact recorded moment.

REASON 4

Track a personal reliability score based on which of your promises were kept versus broken.

What's in the stack?

PythonFastAPINext.jsSQLiteOllama

How it stacks up

rajdeepkushwaha5/saytrace0xallam/my-recipe0xhassaan/nn-from-scratch
Stars00
LanguagePythonPythonPython
Last pushed2022-11-22
MaintenanceDormant
Setup difficultyhardmoderatemoderate
Complexity4/52/54/5
Audiencegeneralgeneraldeveloper

Figures from each repo's GitHub metadata at analysis time.

How do you spin it up?

Difficulty · hard Time to first run · 1h+

Requires running Ollama, Supermemory Local, and a FastAPI backend together locally, with no cloud API shortcut available.

Wtf does this do

SayTrace is a tool that listens to your real conversations and keeps a private record of the promises made in them, whether you made them to someone else or someone made them to you. The idea is that important commitments made casually in a hallway, on a call, or at dinner are easy to forget, so SayTrace captures what was said, figures out who owes what to whom, and can remind you before a promise is broken. Everything runs on your own machine rather than in the cloud. Audio recorded from your desktop or phone is transcribed locally using faster-whisper, and a language model called Qwen, run through Ollama, extracts the actual promises, decisions, and people involved, along with converting relative dates like tonight or next Monday into real deadlines based on when the conversation happened. The operational record of promises, deadlines, and audio timing is stored locally in SQLite, while a separate local memory system called Supermemory handles semantic search and per-person profiles. No cloud AI APIs or external API keys are used anywhere in the pipeline. You can capture a conversation from a desktop microphone, a pasted note, a phone capture page, or an iPhone Shortcut. Once captured, SayTrace tracks each promise through its lifecycle, marking it open, kept, broken, or snoozed, and shows a reliability score based on resolved promises. It also flags things like overdue promises, days with too many commitments, and conflicting deadlines. When you ask about a promise, SayTrace answers with a citation back to the exact recorded moment it was spoken, rather than a generic AI summary, and you can replay that audio directly. The project was built for the Supermemory Local hackathon and is designed around explicit, user-initiated capture rather than always-on ambient recording. Users remain responsible for getting consent from other people before recording them, depending on their local laws. Anyone using it can inspect, correct, export as Markdown, or delete their stored memory at any time. This is aimed at individuals who want to track personal and professional commitments without sending private conversations to a cloud service.

Yoink these prompts

Prompt 1
Explain how SayTrace resolves a relative deadline like 'by Friday' into an actual date based on when the conversation happened.
Prompt 2
Walk me through SayTrace's local architecture, from microphone capture through Whisper transcription to Supermemory indexing.
Prompt 3
How does SayTrace decide whether a promise is you_owe or they_owe when extracting it from a conversation?
Prompt 4
Show me how to set up SayTrace locally with Ollama and Supermemory Local running on my own machine.
Prompt 5
Explain the privacy guarantees SayTrace makes about not using any cloud AI APIs.

Frequently asked questions

wtf is saytrace?

A fully local tool that listens to your conversations, extracts the promises made in them, and answers questions about who owes what with a citation back to the exact moment it was said.

What language is saytrace written in?

Mainly Python. The stack also includes Python, FastAPI, Next.js.

How hard is saytrace to set up?

Setup difficulty is rated hard, with roughly 1h+ to a first successful run.

Who is saytrace for?

Mainly general.

View the repo → Decode another repo

This repo across BitVibe Labs

Don't trust strangers blindly. Verify against the repo.