MeetUs

Setup

• Import the file into your pc

• Then go the file location

Create Virtual Environment

python -m venv myenv

Enabling virtual Environment

myenv\Scripts\activate

Starting app

streamlit run app.py

Usage Guidelines

User Interface

The application interface consists of a sidebar for navigation and a main content area for specific functionalities.

Meeting Agenda:

Upload relevant documents (e.g., meeting objectives, project proposals). Interact with the chat interface to provide details about the meeting. Download the generated agenda as a PDF.

Upload Meeting:

Upload a video recording of the meeting. Transcribe the video and download the transcript as a PDF.

Meeting Tracking:

Upload meeting agenda and transcript documents (PDF format). Analyze meeting flow and effectiveness through a chat interface. Download the analysis report as a PDF.

Meeting Summary:

Upload meeting agenda and transcript documents (PDF format). Generate a concise summary based on the uploaded documents. Download the meeting summary as a PDF.

Dependencies:

The code relies on the following Python libraries:

streamlit: For creating interactive web applications. fpdf: For generating PDF documents. langchain & related sub-packages: For natural language processing tasks like text processing, embedding generation, and retrieval. pydotenv: For managing environment variables securely. google-auth & related libraries: For authentication with Google APIs (if using Groq LLM).

Architecture Breakdown:

Streamlit App:

The main entry point of the application, providing a user interface with interactive elements like file uploaders, buttons, and text boxes. Handles user interactions and navigates between different pages based on user choices.

Language Models:

Hugging Face Embeddings: Used for generating embeddings of text documents, enabling semantic understanding and comparison. ChatGroq: A large language model from Groq, employed for generating text responses, such as meeting agendas, summaries, and analysis. Document Processing:

PyPDFLoader: Loads PDF documents into a structured format. RecursiveCharacterTextSplitter: Splits text into smaller chunks for efficient processing. Chroma Vectorstore: Stores embeddings of document chunks for fast retrieval.

Prompt Engineering:

ChatPromptTemplate: Defines prompts for interacting with the language models, providing context and instructions. MessagesPlaceholder: Used to insert conversation history into prompts. Retrieval Chains:

create_history_aware_retriever: Combines the language model and retriever to maintain conversation context. create_retrieval_chain: Creates a chain for retrieving relevant information from the vectorstore based on user queries.

QA Chains:

create_stuff_documents_chain: Chains the language model and document retriever for question-answering tasks. Conversation History:

ChatMessageHistory: Stores conversation history to maintain context and enable more coherent responses.

Additional Modules:

FPDF: For generating PDF documents from generated text. google_auth & related libraries: For authentication with Google APIs (if using Groq LLM). whisper: For speech recognition in video transcription. moviepy: For video editing and audio extraction. pydub: For audio manipulation. speech_recognition: For audio transcription.

Workflow:

User Interaction: The user interacts with the Streamlit interface, uploading documents or videos as needed. Document Processing: Uploaded documents are processed, split into chunks, and embedded. Prompt Engineering: Prompts are constructed based on user input and conversation history. Retrieval: Relevant information is retrieved from the vectorstore based on the prompts. Language Model Processing: The language model generates text responses based on the retrieved information and prompts. Output: The generated text is displayed to the user or saved as a PDF.

Key Components:

Embedding Generation: Creates semantic representations of text for efficient retrieval and comparison. Prompt Engineering: Carefully crafted prompts guide the language model to generate relevant and informative responses. Retrieval: Efficiently retrieves relevant information from the vectorstore based on user queries. Language Model: Generates human-like text responses based on the provided prompts and information. User Interface: Provides a user-friendly way to interact with the application and view results.

Explaination Video

Explaination.video.1.1.1.mp4

(Sorry for the bad quality I had to compress the video under 10mb to upload it on github)

Google Drive Link

https://drive.google.com/file/d/1OyGjb2l2yzZ_DgSQe36a8kC68r0KkJim/view?usp=sharing