🎯 Vector Embeddings

📌 Overview

This project integrates ChromaDB, Google Gemini AI, and YouTube Transcript API to handle vector embeddings, text processing, and CRUD operations. It supports semantic search, text embeddings, and AI-generated keynotes from video transcripts.

🚀 Features

• ChromaDB integration for vector storage
• Google Gemini AI for text embeddings and summarization
• YouTube Transcript API to fetch video transcripts
• CRUD operations (Create, Read, Update, Delete) on embeddings
• Persistent and ephemeral database modes using chromadb
• File handling for storing transcripts and notes

🛠️ Setup & Installation

1️⃣ Install Dependencies

Ensure you have Python installed, then run:

pip install -r requirements.txt

2️⃣ Set Up Environment Variables

Create a .env file with:

GOOGLE_API_KEY=your_gemini_api_key

3️⃣ Run the Project

python main.py

---

📝 Key Functionalities

🔹 ChromaDB Client

Defined in utils.py, the get_client() function initializes a ChromaDB client:

def get_client(client_type=None, path=None):
    if client_type.lower() != 'persistent' and path is None:
        return chromadb.EphemeralClient()
    return chromadb.PersistentClient(path=path)

• Persistent Mode: Saves embeddings for future retrieval.
• Ephemeral Mode: Stores data only for the current session.

🔹 Creating & Managing Collections

Collections store vector embeddings and text data.

def get_or_create_collection(client, name='my_collection', embedding_function=None, data_loader=None):
    return client.get_or_create_collection(
        name=name,
        embedding_function=embedding_function,
        data_loader=data_loader
    )

🔹 Generating AI-Based Keynotes

Extracts keynotes from a transcript using Gemini AI:

response = genai_model.generate_content(prompt + transcript, stream=False)
with open(notes_path, "w") as file:
    file.write(response.text)

---

📌 Usage Guide

1️⃣ Fetch and Store Video Transcripts

transcript = YouTubeTranscriptApi.get_transcript(video_id, languages=['en', 'en-US'])
formatted_transcript = TextFormatter().format_transcript(transcript)

2️⃣ Generate and Store Keynotes

response = genai_model.generate_content("Extract keynotes: " + transcript)
collection.upsert(ids=[video_id], documents=[response.text])

3️⃣ Perform CRUD Operations

• Insert data: collection.upsert()
• Retrieve data: collection.get()
• Update existing data: collection.update()
• Delete data: collection.delete()

---

🔗 Additional Information

• ChromaDB Docs: https://github.com/chroma-core/chroma
• Google Gemini AI: https://ai.google.dev/
• YouTube Transcript API: https://pypi.org/project/youtube-transcript-api/

---

📜 License

This project is licensed under the GNU GENERAL PUBLIC LICENSE - see the LICENSE file for details.

---

📬 Contact

For any inquiries or contributions, please feel free to reach out.

• GitHub Profile: kivanc57
• Email: kivancgordu@hotmail.com