Spaces:
Runtime error
Runtime error
| title: Origin ZeroGPU API | |
| emoji: 🎓 | |
| sdk: gradio | |
| app_file: app.py | |
| python_version: 3.12.12 | |
| models: | |
| - sentence-transformers/paraphrase-multilingual-mpnet-base-v2 | |
| short_description: Graduation project similarity search and idea enhancement | |
| # Origin ZeroGPU Gradio adapter | |
| **Experimental staging build**, not a production deployment. The original research retrieval implementation must remain frozen. Run `export_vectors.py` from the main Origin GitHub checkout first, and copy the ORIGINAL `data/processed/projects_clean.csv` to `assets/projects_clean.csv` and generated `assets/project_vectors.npz` here. Missing assets deliberately fail startup. | |
| ZeroGPU GPU callback `search` only computes the query embedding and ranks results. The `enhance` callback calls Gemini without holding a GPU. Use **Use via API** to inspect the deployed names and signatures. Do not paste Gemini keys in code; set `GEMINI_API_KEY` in Space Settings -> Secrets. One stage can work without Gemini; enhancement requires a key. | |
| The resulting public Space is only a research demonstration; public API clients can consume GPU quota and Gemini usage. Check terms and permission to republish corpus. No originality percentages or plagiarism claims are produced. | |