CatchUp Local: Catch Up on 500 Messages Without Reading 500

This is a submission for the Hacktoberfest Open-Source AI Challenge — Week 3: Slow and Steady.

What I Built

Ever returned to a group chat after a busy day and found hundreds of messages waiting for you?

Reading everything takes time. Skimming risks missing the one message that actually matters: a decision, a deadline, or a task assigned to you.

CatchUp Local turns exported WhatsApp group chats into concise, evidence-backed briefings.

It helps identify:

• Important updates and decisions.
• Action items, with owners and deadlines when explicitly supported by the conversation.
• Unanswered questions and unresolved disagreements.
• Original messages worth reading directly.

The key design principle is traceability. Instead of asking users to trust an AI-generated summary blindly, CatchUp Local connects briefing items to their supporting messages.

The goal is simple: spend less time catching up and more time acting on what matters.

Demo

[Add a screenshot or short video demonstrating the application here.]

The demonstration uses a synthetic group conversation so the workflow can be explored without sharing anyone's private messages.

Code

GitHub repository: https://github.com/ayushshandilya-dev/catchup-local

Explore the implementation, setup instructions, tests, and documentation in the repository.

How I Built It

CatchUp Local uses Python for chat parsing and application logic, Ollama for local inference with an open-weight model, and a lightweight web interface for reviewing results.

The processing pipeline separates deterministic parsing from AI interpretation:

1. Parse the exported conversation while preserving message IDs, timestamps, and senders.
2. Use the local model to identify candidate decisions, action items, and important updates.
3. Validate the structured output and its references against the original messages.
4. Let users inspect the evidence behind each briefing item.

The system is designed to avoid inventing deadlines, assigning unsupported responsibilities, or presenting uncertain interpretations as confirmed facts.

Why Does Open Innovation Matter?

Group conversations can contain personal, professional, and sensitive information. Sending every message to a hosted AI service is not always an acceptable trade-off for convenience.

Using an open-weight model with local inference makes it possible to build this workflow without relying on a cloud AI API. It also makes experimentation more accessible: developers can inspect the architecture, adapt the prompts, evaluate failure cases, and improve the system.

Open innovation is not just about making source code available. It is also about giving people more control over the tools they use and the environments in which those tools operate.

Local inference does not guarantee absolute privacy, but it reduces the need to send conversation data to an external AI provider.

The Lazy Payoff

The intended payoff is fewer minutes spent scrolling through repetitive messages and fewer missed decisions buried in long conversations.

CatchUp Local is designed to make that payoff measurable through evaluation of decision extraction, action-item accuracy, and evidence references.

[Add actual test results and measured reading-time comparisons after running the evaluation. Do not report unverified performance or time savings.]