[FEATURE] OpenAI Model Provider

[pgrayy]

Problem Statement

I would like Strands Agents to support an OpenAI model provider. This would allow us to connect to OpenAI models directly. Additionally, any OpenAI compatible provider (e.g., LiteLLM) could reuse the request/response formatting logic and thus reduce code duplication.

Proposed Solution

Implement the provider in a similar manner to the existing providers (e.g., LiteLLM). As part of this work, we can also setup an abstract OpenAI class so that any OpenAI compatible provider can reuse the request/response formatting logic.

Use Case

• Call OpenAI models directly

Alternatives Solutions

N/A

Additional Context

N/A

[niklas-palm]

Is this not already supported?
LiteLLM support

[pgrayy]

Is this not already supported? LiteLLM support

Yes it is, but we want to add direct support for a few reasons:

1. Slight optimization as we avoid the layer of abstraction LiteLLM creates.
2. Eliminate extra dependencies when users only want to run OpenAI.
3. Reduce code duplication across OpenAI compatible providers.
   • For example, LiteLLM could derive much of its request/response formatting logic from this proposed OpenAI model provider as the logic will be the same.

[pgrayy]

#65

[pgrayy]

PR is merged into main and will go out with our next PyPI release.