Invalid image input when using Ollama as model backend

Describe the bug
docker本地部署接入Ollama视觉模型gemma3:12b分析图片报错，报错信息如下：
Error: 400 invalid image input
at APIError.generate (/app/node_modules/.pnpm/openai@4.83.0_ws@8.17.1_zod@3.23.8/node_modules/openai/error.js:45:20)
at OpenAI.makeStatusError (/app/node_modules/.pnpm/openai@4.83.0_ws@8.17.1_zod@3.23.8/node_modules/openai/core.js:299:33)
at OpenAI.makeRequest (/app/node_modules/.pnpm/openai@4.83.0_ws@8.17.1_zod@3.23.8/node_modules/openai/core.js:343:30)
at process.processTicksAndRejections (node:internal/process/task_queues:95:5)
at async /app/node_modules/.pnpm/@langchain+openai@0.4.3_@langchain+core@0.3.29_ws@8.17.1/node_modules/@langchain/openai/dist/chat_models.cjs:1519:29
at async RetryOperation._fn (/app/node_modules/.pnpm/p-retry@4.6.2/node_modules/p-retry/index.js:50:12)

To Reproduce
Steps to reproduce the behavior:

1. 上传图片节点（图片格式 png，大小800kb左右）
2. 使用AI问答节点并将图片节点作为上下文

Screenshots

Desktop (please complete the following information):

• OS: Mac OS M1 Pro
• Browser edge
• Version 15.3.2

Additional context
docker desktop： 4.32.2

Start with the Ollama model-backend integration and the image-upload-to-AI-Q&A context path described in the reproduction steps. Reproduce the request with a PNG through Docker, then trace the payload into the OpenAI-compatible client; done means the Ollama vision model accepts the image without an "invalid image input" error.