Skip to content

Feature Request: Extensible Provider Architecture #3

Description

@DevBhuyan

Summary

As OpenBalancer grows, adding support for new inference providers should require minimal code changes.

Currently, each new provider integration may require modifications across multiple parts of the codebase. As the number of supported providers increases, this can make maintenance and contribution more difficult.

Goal

Enable new providers to be added through a well-defined extension mechanism with minimal changes to existing code.

Ideally, adding a provider such as:

  • Azure OpenAI
  • Anthropic
  • AWS Bedrock
  • Google Vertex AI
  • Together AI
  • Fireworks AI
  • DeepInfra

should primarily involve implementing a provider-specific adapter and registering it with the platform.

Benefits

  • Easier community contributions
  • Faster integration of emerging providers
  • Reduced maintenance burden
  • Cleaner separation between routing logic and provider implementations
  • Improved long-term scalability of the project

Possible Directions

Potential approaches could include:

  • Provider plugin architecture
  • Provider registry pattern
  • Capability-based provider discovery
  • Shared base classes for OpenAI-compatible providers

The exact implementation is open for discussion.

Help Wanted

Contributors with experience in API gateway design, plugin systems, or multi-provider AI infrastructure are welcome to share ideas and submit proposals.

Metadata

Metadata

Assignees

No one assigned

    Labels

    No labels
    No labels

    Projects

    No projects

    Milestone

    No milestone

    Relationships

    None yet

    Development

    No branches or pull requests

    Issue actions