GameplayCollector V2A library that checks what it keeps.
A collector driven by a local AI that researches, selects and archives videos to Drive with hash checks.
Overview
A library that checks what it keeps.
A collector driven by a local AI that researches, selects and archives videos to Drive with hash checks.
Product definition and AI-assisted development, connecting requirements, implementation and review.
Local pilot with 73 tests passing. The first delivery archived 18 videos (7.16 GB) with size and MD5 verified; re-running the same 18 IDs produced 0 new downloads.
The problem
Building an editing library without broken, duplicate or off-topic files.
The solution
You type a subject; a local model (Ollama, CPU only) plans the searches and ranks real results. The program validates schema, paths, duration and limits before downloading, runs FFmpeg QC and only sends to Drive once size and hash match.
Architecture
Ask — Plain-language request.
An explanatory diagram of the running workflow. This is not live system monitoring.
Features
Searches planned by a local AI
Implemented¹FFmpeg QC and hash checks
Implemented¹Private Drive upload through rclone
Implemented¹¹ Identified in code or implementation documentation; not equivalent to a new production test.
Technologies
Technical challenges
When a source asks for a bot check, the batch pauses and reports it: no cookies, proxies or workarounds. The AI gets no secrets, runs no commands and cannot invent IDs.
Real screenshots
No reviewed screenshot is available to publish yet.
Lessons learned
A small model is useful when the program around it validates everything it returns.
Links
Public repository not provided.
Next conversation
Let’s build something.An interesting problem?
Let’s build.
Web systems, automation and tools that turn an idea into working software.