
The provider management system has been rebuilt from the ground up. Now you can add your own AI providers, edit existing ones, and configure custom models with specific capability settings. Available in both the open-source edition and LobeHub Cloud.
Full control over your providers
Previously, you were limited to a fixed set of pre-configured providers. The new system lets you:
- Add custom providers: Connect to any OpenAI-compatible API endpoint
- Edit existing setups: Modify provider settings without waiting for updates
- Configure custom models: Define your own model entries with specific capabilities (vision, function calling, file support)
- Remove unused providers: Clean up your list to focus on what you actually use
Multimodal ready
The system supports language, image, and voice capabilities today. Configure which models can process images, which support tool calling, and which handle file uploads. Video and music generation support is planned for upcoming releases.
Feedback
Have suggestions for the provider management system? Join the conversation in GitHub Discussions.
Version details
Features
What's improved
- v1.47.7Remove redundant payload remapping in client-fetch.
- v1.47.6Refactor provider code.
- v1.47.5Improve ai provider code.
- v1.46.4Refactor some implement for the next performance improvement.
- v1.46.1Add auth support for PROXY_URL.
- v1.45.17Refactor Minimax with LobeOpenAICompatibleFactory.
- v1.45.16Improve ai provider code.
- v1.45.13Improve model config form modal.
- v1.45.2Update siliconcloud model list.
- v1.43.5Fix style warning in antd 5.23.0 and some error logs.
- v1.43.3Upgrade @clerk/nextjs to v6.
What's fixed
- v1.47.3Fix hydration error.
- v1.47.2Fix api key in api key form.
- v1.46.7Improve validation for provider and model in parseFilesConfig, temporarily disable S3 client integrity check for Cloudflare R2.
- v1.46.6Gemini models HarmBlockThreshold.
- v1.46.3Fix azure in new ai provider.
- v1.45.15Fix pull models error in new ai provider.
- v1.45.12Fix enable_search parameter intro condition in Qwen.
- v1.45.11Support Gemini 2.0 HarmBlockThreshold.
- v1.45.10Fix some providers issues.
- v1.45.9Fix pin package manager to pnpm\@9 for docker.
- v1.45.8Refactor dynamic import in RSC.
- v1.45.7Fix released at for undefined condition.
- v1.45.6Fix *_MODEL_LIST env in new provider.
- v1.45.5Revert officeparser.
- v1.45.4Fix GitHub and huggingface provider config unusable.
- v1.45.3Fix some ai provider known issues.
- v1.45.1Fix remark gfm regex breaks in Safari versions
- v1.44.3Fix provider enabled issue.
- v1.44.2Add provider id validate.
- v1.44.1Fix model select not auto update and sort issue.
- v1.43.6Fix portal suspense error when first open.
- v1.43.4Fix format short number.
- v1.43.2Fix heatmap and manifest.
- v1.43.1Fix stats data query issue.
