Skip to content

[pull] main from inclusionAI:main#23

Merged
pull[bot] merged 1 commit intoaxistore80-coder:mainfrom
inclusionAI:main
Apr 1, 2026
Merged

[pull] main from inclusionAI:main#23
pull[bot] merged 1 commit intoaxistore80-coder:mainfrom
inclusionAI:main

Conversation

@pull
Copy link
Copy Markdown

@pull pull bot commented Apr 1, 2026

See Commits and Changes for more details.


Created by pull[bot] (v2.0.0-alpha.4)

Can you help keep this open source service alive? 💖 Please sponsor : )

…ible API (#1119)

Enable VLM models (e.g. Qwen3-VL) to receive image_url content parts
through both the Chat Completions and Responses APIs. Images are
extracted from messages, forwarded as base64 data to SGLang/vLLM
backends, and tokenized with HF-compatible image placeholders.

Key changes:
- Add _extract_images_from_messages() for parsing OpenAI-format
  image_url content parts into base64 data, HF tokenizer messages,
  and vLLM vision messages
- Extract _build_messages_list() to module level for Responses API
  input_image/input_text/output_text conversion with validation
- Forward image_data in SGLangBridgeBackend payload
- Validate empty/missing image URLs with clear error messages
- Forward detail field from Responses API input_image content
- Reduce GPU memory fractions in integration tests so concurrent
  module-scoped server fixtures fit on a single CI GPU
- Add 41 unit tests and 6 GPU integration tests (Qwen3-VL-2B)
@pull pull bot locked and limited conversation to collaborators Apr 1, 2026
@pull pull bot added the ⤵️ pull label Apr 1, 2026
@pull pull bot merged commit 92c467e into axistore80-coder:main Apr 1, 2026
Sign up for free to subscribe to this conversation on GitHub. Already have an account? Sign in.

Projects

None yet

Development

Successfully merging this pull request may close these issues.

1 participant