Feature : Added workflow support for Llama Vision 3.2

[AHB102]

Description

Issue : #692
Added a workflow block for Llama Vision 3.2

List any dependencies that are required for this change.
None

Please delete options that are not relevant.

• New feature
• This change requires a documentation update

How has this change been tested, please provide a testcase or example of how you tested the change?

No testing has been performed

Any specific deployment considerations

Currently, only the Llama Vision 3.2 model workflow block is present. API keys for the model have not been provided.

Docs

• Docs updated? No, documentation needs updates

[probicheaux]

Hello @AHB102 ! First of all, thanks for the contribution! Is this PR ready for review? It appears to just rename a few variables but functionally still calls the OpenAI API.

When you're ready for review, please add a test as in https://github.com/roboflow/inference/blob/main/tests/workflows/unit_tests/core_steps/models/foundation/test_openai.py

Tag me or @PawelPeczek-Roboflow if you have further questions!

[AHB102]

@probicheaux So, I'm still working on this. I'm trying out OpenRouter's Inference API for LLama Vision, which uses an OpenAI client. I've tweaked it a bit, but most of it is based on how OpenAI and Anthropic workflows do things. I opened a PR to get some feedback on what might be wrong. I've also started writing unit tests, but I haven't actually run them yet. I'm a bit stuck on how to test this workflow part. I haven't worked with a codebase this big before, so I'm still learning the ropes. Some things might be a bit off. 😅

[PawelPeczek-Roboflow]

ok, have taken a look - looks good - I will add test later on this week, looks good, if I did not discover anything I would simply approve

[probicheaux]

@AHB102 things are looking good! if you're interested in trying to test it, it is possible to build one of the dockerfiles locally and connect workflows to localhost, or to mock out the response from the openrouter call as in

inference/tests/workflows/unit_tests/core_steps/models/foundation/test_openai.py

Line 384 in a4e51e2

def test_execute_gpt_4v_request() -> None:

It sounds like you and @PawelPeczek-Roboflow are in correspondence, so I'll let you two take it from here!

[PawelPeczek-Roboflow]

I am sorry, running out of time this week to handle that - will try to push through finish line next week

[PawelPeczek-Roboflow]

really sorry, got buffer overflow with work on another topics. This would need to wait till after Christmas unfortunately

[AHB102]

@PawelPeczek-Roboflow Hello, I've been unavailable for the past few days due to exams. Is there anything I need to address ?

[PawelPeczek-Roboflow]

I am dragging that change in and will fix what is needed.
Thanks a lot for contribution and sorry for very long delay, I am back from vacation now - will take that and provide tests and alignments.
Will link follow up PR if you are interested to see what changes I'll be making.

[PawelPeczek-Roboflow]

this is PR with initial refactor: #927

I will add PR with tests soon

[AHB102]

@PawelPeczek-Roboflow Yes, I'd be interested in seeing the follow-up PR 😁

[PawelPeczek-Roboflow]

#928