kratul9563/VideoGuard-AI — reverse-engineered prompt
Reverse engineered prompt
Build me a backend app called VideoGuard AI that can audit videos for compliance issues. I want to give it a YouTube link, have it pull the transcript and any text it can read from the video, then compare that against compliance guidance using retrieval and AI reasoning. It should detect likely violations, explain why they were flagged, and return a clean structured JSON report with the findings, evidence, and severity.
Please use the Azure based services mentioned in the project, and wire it up as a usable API with clear endpoints so I can test the full flow end to end. If it helps, include a simple way to run it locally and a sample request and response. I also want basic observability and tracing so it is easier to debug what the agent is doing.
Keep the implementation focused on the core workflow, ingestion, transcript and OCR extraction, compliance lookup, reasoning, and report generation. You can look up current docs online if you need to.
Want more depth? Deep Reverse