Release v0.25.0

[lchoquel]

Release v0.25.0

Bumps version from 0.24.1 to 0.25.0.

Changelog

Added

• GPT-5.5 model support. New gpt-5.5 entry on the openai, azure_openai and gateway backends.

• gpt-image-2 image generation (OpenAI). New model entry with img-gen routing and constraints on openai, azure_openai.

Fixed

• PDF input declared on Azure OpenAI and gateway for the GPT-5.4 series. gpt-5.4, gpt-5.4-mini, gpt-5.4-nano, gpt-5.4-pro (and gpt-5.5) now advertise pdf in their inputs on azure_openai and the gateway, matching the existing OpenAI direct entries. Verified end-to-end with live document tests against both backends.

Changed

• Img-gen taxonomies aligned for gpt-image-2. Convention: gpt_image for taxonomies/values shared across all OpenAI GPT Image models (legacy + gpt-image-2); gpt_image_legacy when the value applies only to gpt-image-1/-1-mini/-1.5 (gpt-image-2 uses unavailable where a legacy-only option does not apply). Renames:

  • AspectRatioTaxonomy.OPENAI_GPT_IMAGE_LEGACY → GPT_IMAGE_LEGACY (value "openai_gpt_image_legacy" → "gpt_image_legacy")
  • AspectRatioTaxonomy.OPENAI_GPT_IMAGE_2 → GPT_IMAGE_2 (value "openai_gpt_image_2" → "gpt_image_2")
  • OutputFormatTaxonomy.GPT → GPT_IMAGE_LEGACY (value "gpt" → "gpt_image_legacy")
  • OutputCompressionTaxonomy.GPT_IMAGE → GPT_IMAGE_LEGACY (value "gpt_image" → "gpt_image_legacy")
  • InputFidelityTaxonomy.OPENAI_IMAGE → GPT_IMAGE_LEGACY (value "openai_image" → "gpt_image_legacy")
  • NumImagesTaxonomy.GPT → GPT_IMAGE (value "gpt" → "gpt_image")
  • InferenceTaxonomy.GPT → GPT_IMAGE (value "gpt" → "gpt_image")
  • Removed dead AspectRatioTaxonomy.GPT.
  • InputImagesTaxonomy.GPT_IMAGE unchanged (already correct — shared across legacy and gpt-image-2).

• HuggingFace and Fal img-gen workers honor model_choice rule. huggingface_img_gen and fal workers now pop "model" from args_dict instead of hardcoding inference_model.model_id, mirroring how prompt is already extracted. The model_choice = "model_id" rule is now required on every model under these backends; missing it raises ImgGenParameterError at call time. Rules added to qwen-image (HF) and to all Fal models (flux-pro, flux-pro/v1.1, flux-pro/v1.1-ultra, flux-2, fast-lightning-sdxl).

---

Summary by cubic

Release v0.25.0 adds gpt-5.5 and gpt-image-2, aligns image‑gen rules, and fixes PDF input support on azure_openai and the gateway. It also adds parameter support checks and polishes img‑gen workers.

• New Features

  • Added gpt-5.5 across openai, azure_openai, and the gateway.
  • Added gpt-image-2 with routing and constraints; best-gpt now points to gpt-image-2.
  • Introduced ImgGenParamSupport and static PipeImgGen validation; expanded unit, integration, and e2e tests.

• Bug Fixes

  • Declared pdf input for the GPT‑5.4 series and gpt-5.5 on azure_openai and the gateway.
  • Enforced model_choice and unified img‑gen taxonomies (gpt_image/gpt_image_legacy) to prevent wrong model selection on openai, fal, and huggingface.
  • Improved img‑gen workers: captured Azure token usage with correct zero‑token handling and longer timeout; standardized error messages using model name; removed sensitive debug logs; refreshed gateway model docs (including Gemini renames) and updated remote config.

^{Written for commit a549a29. Summary will update on new commits. Review in cubic}

[greptile-apps]

Greptile Summary

• Adds gpt-image-2 and gpt-5.5 model support across OpenAI, Azure OpenAI, and gateway backends, with a unified ImgGenArgsFactory-based args pipeline replacing per-worker hardcoded logic.
• Introduces ImgGenParamSupport for load-time blueprint validation, renames img-gen taxonomies to gpt_image_legacy/gpt_image convention, and enforces model_choice rules on Fal and HuggingFace workers.
• The best-gpt alias is promoted from gpt-image-1.5 to gpt-image-2; callers relying on transparent background, output_format, or input_fidelity via this alias will now get ImgGenParameterError at runtime.

Confidence Score: 4/5

Safe to merge; only P2 style/logic findings, core paths look correct

No P0/P1 findings. Two P2 issues: an error message uses model_id instead of model_name, and an or-based token extraction can silently swap token categories on a zero count. Neither affects the happy path.

pipelex/plugins/azure_rest/azure_img_gen_worker.py (token extraction logic), pipelex/cogt/img_gen/img_gen_args_factory.py (error message variable)

Important Files Changed

Filename	Overview
pipelex/cogt/img_gen/img_gen_args_factory.py	Adds model_name parameter to make_args_for_model, new INPUT_IMAGES/INPUT_FIDELITY topic handlers, and model_choice taxonomy routing; one minor error message inconsistency (model_id vs model_name)
pipelex/plugins/openai/openai_img_gen_worker.py	Migrates to unified ImgGenArgsFactory, adds images.edit routing for input-image requests, robust size/format fallback detection, and multipart conversion helper; looks correct
pipelex/plugins/openai/openai_img_gen_factory.py	Refactored to support both legacy GPT Image and gpt-image-2 size validation; comprehensive constraint checks for gpt-image-2 dimensions
pipelex/plugins/azure_rest/azure_img_gen_worker.py	Adds model_name pass-through, timeout bump to 600s, and usage token extraction; minor logical issue with or for zero-valued token counts
pipelex/cogt/img_gen/img_gen_param_support.py	New capability-check module reusing factory as single truth source; clean NamedTuple approach with abstain-on-unknown-taxonomy policy
pipelex/pipe_operators/img_gen/pipe_img_gen.py	New _validate_param_support_against_model_rules hook wired into static validation; gracefully skips unresolvable choices
pipelex/kit/configs/inference/backends/openai.toml	Adds gpt-5.5, gpt-image-2 (and dated snapshot), rules for all gpt-image models including model_choice; looks complete
pipelex/kit/configs/inference/backends/azure_openai.toml	PDF input added to GPT-5.4 series, gpt-5.5 added, taxonomy renames applied to all gpt-image models; gpt-image-2 added with correct unavailable rules for output_format/compression/input_fidelity

Flowchart

%%{init: {'theme': 'neutral'}}%%
flowchart TD
    A[OpenAIImgGenWorker._gen_image_list] --> B[ImgGenArgsFactory.make_args_for_model]
    B --> C{model_rules loop}
    C --> D[model_choice → model key]
    C --> E[aspect_ratio → size string]
    C --> F[inference → quality/steps]
    C --> G[input_images → image data URLs]
    C --> H[output_format / compression]
    B --> I[args_dict]
    I --> J{image key present?}
    J -- yes --> K[_convert_image_data_urls_for_openai_sdk]
    K --> L[images.edit **args_dict]
    J -- no --> M[images.generate **args_dict]
    L --> N[ImagesResponse]
    M --> N
    N --> O{response.size?}
    O -- yes --> P[parse WxH from response]
    O -- no --> Q[_get_requested_size from args_dict]
    Q --> P
    N --> R{response.output_format?}
    R -- yes --> S[use reported format]
    R -- no --> T[detect from bytes]
    T --> S
    P --> U[GeneratedImageRawDetails]
    S --> U

Loading

Prompt To Fix All With AI

This is a comment left during a code review.
Path: pipelex/cogt/img_gen/img_gen_args_factory.py
Line: 192

Comment:
**Inconsistent variable in error message**

`model_id` is used here while all other error messages in this method use `model_name`. An operator will see the raw `model_id` (e.g., `"gpt-image-2-2026-04-21"`) instead of the human-readable config name (e.g., `"gpt-image-2"`), making the message harder to act on.

```suggestion
            msg = f"Model '{model_name}' does not support input_fidelity"
```

How can I resolve this? If you propose a fix, please make it concise.

---

This is a comment left during a code review.
Path: pipelex/plugins/azure_rest/azure_img_gen_worker.py
Line: 107-112

Comment:
**`or` short-circuits on zero token count**

`usage_dict.get("prompt_tokens") or usage_dict.get("input_tokens")` evaluates the right-hand side when the left is falsy — including when it is `0`. A zero `prompt_tokens` value (rare but valid) would fall through to `input_tokens`, potentially misattributing token categories. The same applies to the output-tokens line.

How can I resolve this? If you propose a fix, please make it concise.

_{Reviews (1): Last reviewed commit: "Release v0.25.0" | Re-trigger Greptile}

[cubic-dev-ai]

3 issues found across 41 files

Prompt for AI agents (unresolved issues)

Check if these issues are valid — if so, understand the root cause of each and fix them. If appropriate, use sub-agents to investigate and fix each issue separately.


<file name="pipelex/plugins/gateway/gateway_img_gen_worker.py">

<violation number="1" location="pipelex/plugins/gateway/gateway_img_gen_worker.py:77">
P1: Avoid logging the full image-generation args payload; it can leak prompt and input image data. Log only non-sensitive metadata (e.g., model, endpoint, arg keys).</violation>
</file>

<file name="pipelex/plugins/azure_rest/azure_img_gen_worker.py">

<violation number="1" location="pipelex/plugins/azure_rest/azure_img_gen_worker.py:110">
P2: Use `is not None` checks when reading token counts; current truthiness checks drop valid `0` token values.</violation>
</file>

<file name="tests/e2e/pipelex/pipes/pipe_operators/pipe_img_gen/test_pipe_img_gen_validation.py">

<violation number="1" location="tests/e2e/pipelex/pipes/pipe_operators/pipe_img_gen/test_pipe_img_gen_validation.py:112">
P3: Skip these checks when the model exists but has no img-gen rules; otherwise the test can become routing-dependent and flaky.</violation>
</file>

_{Reply with feedback, questions, or to request a fix. Tag @cubic-dev-ai to re-run a review.}