OGX is an open-source agentic API server written in Python that provides a drop-in replacement for the OpenAI API, enabling developers to build AI applications that can run on any infrastructure from local laptops to cloud environments. The project was formerly known as Llama Stack and underwent a rebranding to reflect its evolution into a model-agnostic, multi-SDK, production-grade platform. The repository is distributed via PyPI and Docker Hub, with an active community supported through Discord channels and regular Thursday community calls.
The core functionality of OGX centers on providing OpenAI-compatible endpoints while supporting multiple large language models and infrastructure providers. The platform offers Chat Completions and Embeddings endpoints that maintain full compatibility with OpenAI clients, allowing developers to swap between models like Llama, GPT, Gemini, and Mistral without modifying application code. Beyond standard completions, OGX implements a Responses API for server-side agentic orchestration that includes tool calling, Model Context Protocol server integration, and built-in file search capabilities for retrieval-augmented generation in single API calls. The platform also provides Vector Stores and Files endpoints for managed document storage and search, batch processing capabilities through the Batches endpoint, and a Skills system for managing versioned skill bundles as zip archives with SKILL.md manifests that agents can invoke.
A distinguishing feature of OGX is its multi-SDK support, allowing developers to use the Anthropic SDK with the Messages endpoint or the Google GenAI SDK with the Interactions endpoint alongside the standard OpenAI API. The Responses API implementation is Open Responses conformant, passing the official conformance test suite. The pluggable provider architecture enables flexible deployment options, with support for development environments like Ollama and production systems like vLLM, while maintaining API consistency across different infrastructure choices.
The repository demonstrates active maintenance and community engagement based on GitGenius tracking data. Across 1,384 tracked issues and pull requests, the median response latency is 0.0 hours with a mean of 45.2 hours, indicating rapid issue handling. The most active issue labels are enhancement with 450 items, bug with 429 items, and stale with 405 items. Key contributors include leseb with 649 tracked events, ashwinb with 463 events, and franciscojavierarceo with 433 events. The project shares overlapping contributors with microsoft/vscode, ogx-ai/llama-stack, and llamastack/llama-stack, indicating connections within the broader AI infrastructure ecosystem.
OGX provides official client SDKs for both Python and TypeScript, with the Python SDK available as the ogx_client package on PyPI. An alternative OpenAPI Generator-based Python client called ogx_open_client is also available for teams with specific OpenAPI tooling requirements. The project maintains comprehensive documentation at ogx-ai.github.io, including API reference materials, provider documentation, OpenAI API compatibility matrices, and getting started guides. The repository includes unit tests and integration tests with automated CI/CD workflows, and developers can contribute through the established CONTRIBUTING.md guidelines.
