fix(llm_client): remove response_format json_object for local LLM compatibility

[ChinmayShringi]

Problem

chat_json() uses response_format={"type": "json_object"}, but LM Studio and Ollama do not support this parameter (only json_schema or text), causing API calls to fail when using local LLMs.

Related references:

• LM Studio: Failed to load model - Gemma 3 1B MLX lmstudio-ai/lmstudio-bug-tracker#534
• Similar to issue Generated content is too outrageous/unrealistic #110 (API call failures)

Solution

Remove response_format from chat_json(). The method already has robust markdown code fence cleanup logic (L93-97) that correctly parses JSON from raw LLM output, making response_format unnecessary.

This follows the same approach as commit 985f89f (handling <think> tags from reasoning models) — improving compatibility with diverse model outputs.

Changed Files

• backend/app/utils/llm_client.py: Remove response_format in chat_json(), rely on prompt instructions + markdown cleanup

Testing

• LM Studio + qwen3.5-9b: Full prediction pipeline (ontology → graph → prepare → simulate → report) passes
• Does not affect OpenAI API users (removing response_format is backwards compatible)

---

Original PR: 666ghj/MiroFish#122
Original Author: @ImL1s