Summarising Microsoft Teams Transcripts using Python and Azure Open AI

This Python-based tool uses Azure OpenAI to summarise Microsoft Teams meeting transcripts. It processes .docx transcript files, chunks them, and uses AI to generate comprehensive summaries.

Diagram

The diagram below shows the current flow:

How it works

1. Read the Teams transcript file
2. Split the document into manageable chunks
3. Use Azure OpenAI to process each chunk
4. Compile a final summary of the document

Prerequisites

• Python 3.12
• Azure Open AI subscription
• pip (Python package installer)

Setup

1. Clone the repository

git clone https://github.com/thomast1906/teams-transcript-summariser-python-openai.git
cd teams-transcript-summariser-python-openai

2. Create and activate a virtual environment:

On macOS and Linux

python3 -m venv venv
source vnv/bin/activate

On Windows

python -m venv venv
venv\Scripts\activate

3. Install dependencies:

pip install -r requirements.txt

4. Configure environment variables in a .env file:

FILE_PATH="<file_path>.docx"
AZURE_OPENAI_API_KEY="azure_openai_api_key"
AZURE_OPENAI_ENDPOINT="azure_openai_endpoint"
AZURE_OPENAI_DEPLOYMENT_NAME="azure_openai_deployment_name"

# prompts from prompt.json
INITIAL_PROMPT_SUMMARY="initial_prompt_summary_per_user_questions_answered"
FINAL_PROMPT_SUMMARY="final_prompt_summary_per_user_questions_answered"

Usage

Run the summariser:

python summary.py

Features

• Chunks large transcripts for efficient processing
• Utilises Azure OpenAI for intelligent summarisation
• Generates summaries focusing on key areas: Overview, Concerns, Current Model, Questions/Thoughts, To-Do Lists, and Outstanding Tasks
• Supports three summary types: Overall, Per User, and Per User with Questions Answered
• Logs processing steps and errors

Prompts

The script uses prompts to guide the AI model on what to generate. The prompts are stored in prompt.json and are read by the script.

1. Overall Summary

• Purpose: Generates a general summary of the entire Teams transcript
• Prompts: initial_prompt_summary and final_prompt_summary

2. User-Specific Summary

• Purpose: Creates a summary with references to each participant and their contribution timestamps
• Prompts: initial_prompt_summary_per_user and final_prompt_summary_per_user

3. Comprehensive User Summary with Q&A

• Purpose: Produces a detailed summary including user references, timestamps, and addressed questions
• Prompts: initial_prompt_summary_per_user_questions_answered and final_prompt_summary_per_user_questions_answered

Example output of each summary: (Formatted raw .txt files for markdown)

1. Overall Summary
2. User-Specific Summary
3. Comprehensive User Summary with Q&A

Logging

The script logs its progress to the console, including chunking status, API requests, and execution time.

Error Handling

Includes error handling for file operations and AI model interactions.