Sprint transcript extractor + product summaries

Creates one .txt transcript file per "Sprint" video published in the last 30 days from the hardcoded YouTube playlists, then generates a per-product summary file using the OpenAI API.

Setup

python -m venv .venv
.\.venv\Scripts\Activate.ps1
pip install -r requirements.txt

Run

python .\sprint_transcripts.py

The script will read OPENAI_API_KEY (and optional OPENAI_MODEL) from your user environment variables. If you prefer a local file, it will also read from a .env file if present. You can copy .env.example to .env and fill in your key:

Copy-Item .env.example .env

Or set the environment variables for just this session:

$env:OPENAI_API_KEY="YOUR_KEY_HERE"
# optional: choose a different model
# $env:OPENAI_MODEL="gpt-5"
# optional: tell yt-dlp which JS runtimes to try (comma-separated)
# $env:YTDLP_JS_RUNTIMES="node,deno"
# optional: point yt-dlp at ffmpeg if it's not on PATH
# $env:YTDLP_FFMPEG_LOCATION="C:\Path\To\ffmpeg\bin"

Output folder: a subfolder named with the run date (YYYY-MM-DD) in the current directory.

Prompt customization

The summary prompt is loaded from prompt.txt in the project folder if present. You can edit this file to tweak the bullet style. The template supports:

• {product} for the product name
• {max_bullets} for the current bullet limit