Enhance LLM functionality with PRO mode support and improved prompt handling

- Added support for PRO mode in LLM with specific instruction handling and context awareness.
- Updated prompt building functions to include prefill options for better context management.
- Introduced new inline examples for PRO mode in JSON format.
- Enhanced system prompts to reflect PRO mode capabilities and rules.
- Modified API endpoints to accommodate new parameters and ensure backward compatibility.
- Improved test cases to validate new functionality and ensure comprehensive coverage.

backend/tests/test_main_endpoints.py

@@ -138,7 +138,7 @@ def test_post_completions_privacy_mode(monkeypatch):
         captured["kwargs"] = kwargs
         return {"content": "done", "think": ""}
     monkeypatch.setattr(main, "call_ollama", fake_call)
-    monkeypatch.setattr(main, "build_completion_prompts", lambda *a, **k: ("sys", "user"))
+    monkeypatch.setattr(main, "build_completion_prompts", lambda *a, **k: ("sys", "user", ""))
     monkeypatch.setattr(main, "prepare_prompt_context", lambda *a, **k: ("p", "s"))
 
     client = TestClient(main.app)