Skip to content

AGENTS.md

Latest commit

 

History

History
78 lines (59 loc) · 3.18 KB

AGENTS.md

File metadata and controls

78 lines (59 loc) · 3.18 KB

AGENTS.md

Purpose

This repository is a Streamlit app that uses LLMs and Retrieval Augmented Generation (RAG) to summarize and chat with YouTube videos, plus a Library page to save outputs.

Tech Stack

  • Languages: Python
  • Frameworks / libraries: Streamlit, LangChain, YouTube Transcript API, peewee
  • Databases: ChromaDB (vector store)

How to Navigate

Main entrypoint is main.py. Streamlit pages live in pages/. Shared logic is in modules/. Prompts are in prompts/. Tests are in tests/.

Repository Structure

High-level architecture: The app consists of three main components: Summary, Chat and Library. Summary generates summaries of YouTube videos, Chat leverages RAG to answer user questions based on video content, and Library manages saved summaries and answers from Q&A sessions.

.
├── main.py                  # Streamlit entrypoint and app shell
├── pages/                   # Streamlit pages
│   ├── chat.py              # RAG-based Q&A UI flow
│   ├── library.py           # Saved summaries/answers UI
│   └── summary.py           # Summary UI flow
├── modules/                 # Core logic (summarization, RAG, persistence, UI helpers)
│   ├── helpers.py           # Shared helpers (config, tokens, providers)
│   ├── persistance.py       # SQLite models and library CRUD
│   ├── rag.py               # Chunking, embeddings, and RAG response
│   ├── summary.py           # Summary prompt + model invocation
│   ├── transcription.py     # Whisper-based transcription
│   ├── ui.py                # Streamlit UI helpers/settings
│   └── youtube.py           # YouTube metadata and transcript retrieval
├── prompts/                 # LLM system/user prompt templates
├── tests/                   # pytest suites
├── data/                    # Local SQLite database and saved outputs
├── config.json              # Default models and UI/config settings
├── docker-compose.yml       # ChromaDB + app services for local Docker setup
├── requirements.txt         # Python dependencies

Key Flows

  • Summary: pages/summary.py -> modules/summary.py -> prompt files in prompts/.
  • Chat (RAG): pages/chat.py -> modules/rag.py -> ChromaDB.
  • Library: pages/library.py -> modules/persistance.py (SQLite).

Coding Standards

  • Follow SOLID principles.
  • Code should be: simple, readable and maintainable.
  • Format with Black.
  • Use snake_case for variables/functions, PascalCase for classes.
  • Provide Google-style docstrings for all classes and functions.
  • Assume features will evolve over time; write code that is easy to extend and modify.

Commit messages

  • Use the Conventional Commits format: type(scope): description.
  • Use these types: feat, fix, docs, style, refactor, perf, test, chore, ci.
  • Use imperative mood: "add" not "added".
  • Keep the subject <= 72 characters.
  • Include scope when clear (e.g., chat, summary, library, rag etc.).
  • If multiple logical changes, summarize the most important one.

Testing

  • Use pytest; tests are in tests/.

Run (local)

  • uv sync
  • docker-compose up -d chromadb
  • uv run streamlit run main.py