perceptra-ai

This project provides a unified REST API that abstracts multiple AI service providers — including Ollama, OpenAI, Google Gemini, and Hugging Face — behind a single consistent interface. It enables developers to switch or combine providers without changing client code. The API supports text generation, chat completions, image-based prompts (multimodal), and embeddings. Features include optional authentication via API key, real-time streaming responses, automatic provider routing, and auto-generated OpenAPI documentation. It is designed for easy deployment using Docker Compose or local Python execution. Configuration is managed through environment variables, allowing toggling of individual providers and setting model-specific parameters. The system also exposes health checks and model listing endpoints. Extensible architecture allows adding new providers. Management commands (via Makefile) simplify lifecycle operations like start, stop, logs, and shell access. MIT licensed.