diff --git a/Project-Overview.md b/Project-Overview.md
new file mode 100644
index 0000000..35a4dcd
--- /dev/null
+++ b/Project-Overview.md
@@ -0,0 +1,45 @@
+# Project Overview
+
+Welcome to the **Llama Chile Shop Video Automation Project**! This system automates the end-to-end process of preparing, rendering, uploading, and promoting Fortnite gameplay videos extracted from livestream sessions.
+
+## Project Goal
+
+Build a robust pipeline that transforms raw stream recordings into polished video content (montages, hits, misses, outtakes), enriched with branded intros/outros, AI-generated descriptions, and automated uploads to YouTube, PeerTube, Mastodon, and Bluesky.
+
+## Project Structure
+
+```
+Z:\                          # NAS-mounted project root
+├── 2025.06.20/              # Stream session directory (date-based)
+│   ├── hits/                # One video file per hit clip
+│   ├── misses/              # One video file per death clip
+│   ├── montages/            # Pre-edited full montage files
+│   ├── outtakes/            # Marked for manual post-processing
+│   └── rendered/            # Output folder for all generated clips
+├── assets/                  # Branding assets (intros, outros, fonts, logos)
+├── modules/                 # Modular helper libraries
+├── logs/                    # Diagnostic logs and output
+├── .env                     # API keys and environment configuration
+├── .gitignore
+├── main.py                  # Entry point for daily automation pass
+├── upload_youtube_montage.py # Standalone script to test YouTube upload flow
+└── README.md
+```
+
+## Versioning
+
+* **v0.0.1**: Initial clean commit of montage rendering pipeline with FFmpeg, branding assets, intro+title composition, and output to `rendered/`.
+* **v0.1.1** (WIP): Adds YouTube upload support with dynamic OpenAI-generated descriptions.
+* **v0.1.2**: Adds PeerTube upload using same modularized description system.
+* **v0.2.0**: Adds Mastodon support with dynamic toot generation.
+* **v0.2.1**: Adds Bluesky support with dynamic toot generation.
+
+## Current Focus
+
+Locking down proof-of-concept for `upload_youtube_montage.py`: passing in a montage clip, deriving stream date, generating AI description, posting to YouTube (private in debug), and returning video URL.
+
+## Notes
+
+* Vertical videos are identified by filenames ending in `-vert` or `-vertical`.
+* Montage videos do **not** have accompanying `notes.*` files.
+* All AI-generated descriptions are based on a dynamic prompt sourced from a `notes.*` template, file-type agnostic.