Skip to content

[Feature]: OAI-Compatible / LiteLLM support #66

Description

@overbit

Problem statement

When integrating Pullfrog into enterprise environments or customized AI stacks, teams frequently prefer routing their LLM requests through an internal API gateway or a unified proxy layer. Currently, Pullfrog supports fixed providers (like OpenAI, Anthropic, or OpenRouter), but lacks a generalized, plug-and-play integration for LiteLLM.

Without this, users who leverage LiteLLM to handle failovers, load-balancing, spend tracking, or custom self-hosted models (e.g., via vLLM or Ollama hosted internally) cannot easily connect Pullfrog to their infrastructure without hacky workarounds or treating it strictly as a pseudo-OpenAI endpoint (which can break provider-specific parameter schemas).

Proposed solution

Add native support for LiteLLM as a Bring Your Own Key (BYOK) provider.

This involves:

Adding litellm to the allowed list/enum of providers in Pullfrog's configuration schemas (e.g., Zod schemas).

Implementing a dedicated LiteLLMProvider that respects LiteLLM's unified routing capabilities.

Allowing the configuration of a LITELLM_BASE_URL (or proxy endpoint) alongside standard API credentials, giving users complete flexibility over their custom gateway routing and dynamic model passing.

Use case

as the rest of providers

Alternatives considered

No response

Additional context

No response

Activity

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Metadata

Metadata

Assignees

No one assigned

    Labels

    enhancementNew feature or request

    Type

    No type

    Projects

    No projects

      Milestone

      No milestone

      Relationships

      None yet

      Development

      No branches or pull requests

      Issue actions