Hi All -- Would anyone like to use AI to improve the efficiency of our process?

I asked ChatGPT (or a version of it) if it could create a process whereby we could analyze several games at once. It responded positively -- see below.

@JonasRath, isn't this what we've been wanting to do for years? Does anyone want to take this on?
EOH
------------------------------------

Absolutely, — that’s a brilliant next evolution for FishTrap, and it’s totally within reach. Here's how we could map it:

🧠 FishTrap: Batch Mode Blueprint
🗂️ 1. Drop Zone (Input Bucket)
Designate a folder (e.g., fishtrap/input/) as the “bucket.”
You drop .pgn files there — one or many.
Optional: support .zip archives of PGNs for bulk uploads.
⚙️ 2. Watcher Script
A lightweight Python script (using watchdog or a simple polling loop) monitors the folder.
When new PGNs appear, it triggers the analysis pipeline.
🧪 3. Batch Analysis Engine
For each PGN:Parse games using python-chess.
Run your mapped engine analysis (e.g., Fairy-Stockfish, ffish, mvstockfish).
Store results in a structured format (e.g., JSON, CSV, or trainer-aligned summary).
📤 4. Output Archive
Results are saved to fishtrap/output/, with:One file per game (e.g., game_001.json)
Or a consolidated report (e.g., batch_summary.csv)
Optional: auto-generate trainer-ready summaries or error logs.
🛠️ 5. Configurable Parameters
You could define:Engine depth
Time per move
Which engines to benchmark
Output verbosity (e.g., full eval tree vs. top line only)

Would you like me to sketch out the Python scaffolding for this — maybe a main.py that wraps the watcher and batch processor? Or would you prefer a CLI-style tool where you just run fishtrap-batch input/ output/ and it processes everything in one go?

Either way, this is a perfect match for your mapped workflows and trainer-aligned analysis.

atomic FBI

Super Admins

Admins