Repo AI：以本機 AI 助理分析文件及整理工作

What I Built
Repo AI — Do less. Get more done.

Repo AI is a local-first AI productivity assistant designed to reduce the repetitive work involved in reading documents, identifying action items, and tracking deadlines.

Instead of manually reviewing lengthy documents, users can upload supported TXT, PDF, and DOCX files and receive an AI-generated summary, extracted tasks, relevant events, and potential blockers.

The application also includes task management, analysis history, and a daily productivity briefing that brings pending tasks, upcoming deadlines, recent document activity, and potential blockers together in one place.

The goal is simple:
spend less time organizing information and more time acting on it.

Code
GitHub repository: https://github.com/sanalnayyar2006/RepoAI.git

The repository contains the application, document-processing pipeline, local model integration, database layer, tests, and setup documentation.

How I Built It

RepoAI is built with Python and Streamlit, with SQLite for persistence and Ollama for local inference.

The configured model is qwen2.5:3b, an open-weight model that runs locally through Ollama.

The application processes supported documents, extracts structured information, and presents the results through a productivity-focused interface.

Key components include:
Streamlit: User interface and interactive dashboard.
Ollama: Local model inference.
Qwen 2.5 3B: Open-weight language model for document analysis.
SQLite: Persistent storage for analyses and tasks.
Python document-processing libraries: Text extraction from supported document formats.
Automated tests: Coverage for document processing, analysis, persistence, and application behavior.
The application is designed around local processing rather than depending on a hosted AI API for its core local workflow. Model-generated results still require human review because extraction can be incomplete or inaccurate.

Why Does Open Innovation Matter?

Open innovation makes it possible to experiment with AI without making a proprietary hosted API the foundation of the application.

Using an open-weight model through Ollama gives users more control over where their documents are processed. It also makes local experimentation possible without requiring every document to be sent to an external inference provider.

This architecture creates opportunities for developers to inspect, modify, evaluate, and extend the system according to their own needs.

RepoAI demonstrates how open-weight models can support practical productivity workflows, while highlighting the importance of evaluating extraction quality rather than assuming every generated result is correct.

The Lazy Payoff
Repo AI reduces the manual effort involved in reviewing documents and organizing their action items.

Instead of separately identifying tasks, deadlines, and potential blockers, users can begin with a structured analysis and a consolidated productivity briefing.