Spaces:
Sleeping
Sleeping
| license: apache-2.0 | |
| sdk: docker | |
| colorFrom: yellow | |
| 🚀 Upniso Backend (v2.0) | |
| Merit-First Creator Ecosystem & AI Feed Engine | |
| Upniso Backend is a production-ready, high-performance backend powering a merit-based creator discovery system and a brand–creator marketplace. | |
| It is designed to prioritize quality, consistency, and fairness over vanity metrics. | |
| This repository represents the core intelligence and infrastructure layer behind Upniso. | |
| 🏗️ System Architecture | |
| The platform is intentionally split into two isolated execution environments for scalability, trust, and abuse prevention. | |
| 1️⃣ Core Feed Engine (Protected) | |
| Powered by a proprietary Merit Algorithm (281 lines) | |
| Enforces strict quality controls | |
| Includes: | |
| Spam protection | |
| Rolling 24-hour upload limits | |
| Merit-based ranking (not engagement bait) | |
| 2️⃣ Community & Brand Hub (Unlimited) | |
| Open interaction layer for: | |
| Brands | |
| Agencies | |
| Creators | |
| Supports: | |
| Brand briefs | |
| Reviews & feedback | |
| Collaboration without feed penalties | |
| This separation ensures the main feed remains signal-rich, while collaboration remains frictionless. | |
| 🧠 Core Capabilities | |
| Merit-First Discovery | |
| Talent Score | |
| Progress Score | |
| Consistency-weighted ranking | |
| Multi-Format Content Support | |
| Short-form video (Reels) | |
| Long-form video | |
| Text posts (X-style) | |
| Artwork & images | |
| Full-length music tracks | |
| Spam & Abuse Shield | |
| Rolling 24-hour upload quotas (e.g. 5 Reels/day) | |
| Automatic quality preservation | |
| Brand Connect Layer | |
| Fixed $2 brand brief fee | |
| Built-in review & feedback loop | |
| Transparent creator selection | |
| AI-Ready Infrastructure | |
| Semantic scoring & matching | |
| ML models loaded at build time (zero cold start) | |
| CPU-optimized for predictable cost | |
| 🛠️ Tech Stack | |
| Language: Python 3.11 | |
| Framework: FastAPI | |
| Database: Firebase Firestore | |
| AI Engine: Cross-Encoder (MS-Marco MiniLM) | |
| Deployment: Docker | |
| Target Platforms: Hugging Face Spaces, Oracle Cloud | |
| 🔐 Architecture Notes | |
| Firebase is used internally as a control plane for: | |
| Metadata | |
| Usage limits | |
| Platform logic | |
| Clients never interact with Firebase directly. | |
| Designed for API monetization, rate limiting, and future multi-tenant usage. | |
| 🚀 Quick Start (Docker) | |
| Ensure the following environment variable is set: | |
| FIREBASE_KEY_B64=<base64-encoded-firebase-key> | |
| Build the image | |
| docker build -t upniso-backend . | |
| Run the container | |
| docker run -p 7860:7860 upniso-backend | |
| The API will be available at: | |
| http://localhost:7860 | |
| Swagger docs: | |
| http://localhost:7860/docs | |
| 📜 License | |
| This project is licensed under the Apache License 2.0. | |
| Commercial usage is provided via hosted API access and managed deployments. | |
| 🧭 Project Status | |
| Backend: Production-ready | |
| API monetization: In progress | |
| Frontend: Intentionally minimal / decoupled |