I have an LLM-agent project where the video component refuses to behave the way it is prompted, and I also need to wire in the Gemini API. Both issues matter to me equally. Video prompt issue Right now the clip renders, but the behavior on screen doesn’t reflect the prompt text. It feels as if the agent hands off the wrong parameters somewhere in the chain. I need the logic traced, the root cause isolated, and the fix committed so the video consistently follows the prompt every run. Gemini API integration The second task is to drop Gemini seamlessly into the same codebase. I already have API credentials ready; what’s missing is the clean wrapper, error handling, and a quick demo call that proves the model responds through my agent framework. Deliverables • Clean, well-commented code showing the video prompt bug fixed • Gemini API module integrated and callable from the project’s existing interface • A brief README or inline notes so I can reproduce your steps and extend them later I keep everything in Git, so you’ll push to a feature branch and open a pull request once you’re satisfied with the tests. If you can jump on a quick call to walk me through your changes, even better—otherwise concise written explanations will do.