🎬 YouTube Companion Agent - Powered by Portia AI

An intelligent AI-powered companion for YouTube that helps you summarize, fact-check, rank ideas, and get recommendations — all through an interactive chat interface.

This project combines:

• Streamlit UI for a clean and interactive frontend.

• Local (via Ollama) /remote LLMs for natural conversation and fallback responses.

• Portia plans (MCP) to orchestrate structured tasks like fact-checking, summarization, and recommendations.

• A custom built/implementation of Youtube MCP server.

• Session state management so conversations and results persist across interactions.

Architecture:

Uses local ollama powered intent classifer function to build the correct Portia plans and to handle out of context prompts , Keeps the chat interactive and on going

Use cases:

✨ Features

🔹 Interactive Chat Interface

Chat with the agent naturally.

Detects intents (summarize_video, fact_check, comments_summary, rank_ideas, etc.).

Falls back to LLM response when no structured intent is found.

🔹 Video Summaries

Paste a YouTube link and get quick, AI-generated summaries.

Comment summaries for community sentiment analysis.

🔹Fact-Checking

Extracts claims from videos and runs fact-checks.

Displays confidence score, trusted sources count, and a list of supporting links.

Results are shown in a visual card layout.

🔹 Recommendations

Get related YouTube videos with title, channel, thumbnail, and direct link.

Nicely formatted for quick browsing.

🔹 Idea Ranking

Save your own project/idea notes.

Agent ranks ideas by confidence and displays them in a stylized leaderboard.

🔹 Persistent Session State

Previous fact-checks, recommendations, and summaries are saved.

No duplication — only new results are appended.

Custom recommendations :

Facts checking / Confidence score :

Comments summary of youtube video :

Content summary of youtube video :

Select your next content creation idea based on current trend :

Note : The application is not yet deployed as it's using a user implemnted custom MCP server, and local ollama models.

Streamlit → frontend & chat interface

Ollama → local LLM for fallback conversations

Portia MCP → orchestrated plans for structured tasks

YouTube API wrapper (custom MCP) → fetching video details, comments, etc.

Session State → to persist results across reruns

🤝Contributing

Pull requests and feature suggestions are welcome!

📜 License

MIT License.