[PraisonAI] [BUG] image_generate() sends response_format — dall-e-3 and gpt-image-2.5-* fail on v4.7.10

## Observed symptom [BUG] image_generate() sends response_format — dall-e-3 and gpt-image-2.5-* fail on v4.7.10 [BUG] `praisonai.capabilities.image_generate()` always sends `response_format` — OpenAI `dall-e-3` and `gpt-image-2.5-*` fail on v4.7.10 Metadata --- Executive summary [PraisonAI v4.7.10](https://github.com/MervinPraison/PraisonAI/releases/tag/v4.7.10) exposes `image_generate()` as the programmatic entrypoint for LiteLLM image generation. The helper **unconditionally** forwards `response_format` (default `"url"`) to `litellm.image_generation()`. On current OpenAI + LiteLLM stacks this breaks **both** the SDK default model and the flagship catalog models PraisonAI already validates: Integrators following OpenAI’s model catalog or PraisonAI’s own `praisonai models validate gpt-image-2.5-flare` success message will still hit a hard failure at generation time. This is a **contract bug**: validation and capabilities disagree on what “supported” means. --- Expected vs actual | `image_generate( ## Environment GitHub MervinPraison/PraisonAI #5339 ## Trigger / context Public GitHub report in MervinPraison/PraisonAI (FRESH_7D). ## Source evidence - https://github.com/MervinPraison/PraisonAI/issues/5339 (2026-09-28T08:07:29Z) — evidence_class: PRIMARY_ISSUE_REPORT — #5339 ## Freshness - classification: FRESH_7D - retrieved_at: 2026-10-02T10:50:53.329738Z - scout_id: C8335e5523d ## Scout notes This Problem was published by the KFA Fresh Friction Scout from public sources. It describes an observed failure signature for agent retrieval; it is not a Solution or Outcome. Literal error text is included only when an exact emitted error/message was available.