implement support for response_format

src/ai_migrate/llm_providers/base.py

@@ -13,6 +13,7 @@ async def generate_completion(
         temperature: float = 0.1,
         max_tokens: int = 8192,
         model: str | None = None,
+        response_format: dict[str, Any] | None = None,
     ) -> tuple[dict[str, Any], list[dict[str, Any]]]:
         """Generate a completion from the LLM.
 
@@ -22,6 +23,8 @@ async def generate_completion(
             temperature: The temperature to use for generation
             max_tokens: The maximum number of tokens to generate
             model: Optional model override
+            response_format: Optional response format specification, e.g. for JSON Schema
+                           Example: {"type": "json_schema", "schema": {...}}
 
         Returns:
             A tuple of (response, messages)

src/ai_migrate/llm_providers/openai.py

@@ -30,6 +30,8 @@ async def generate_completion(
         tools: list[ToolDefinition] | None = None,
         temperature: float = 0.1,
         max_tokens: int = 8192,
+        model: str | None = None,
+        response_format: dict[str, Any] | None = None,
     ) -> tuple[dict[str, Any], list[dict[str, Any]]]:
         """Generate a completion
 
@@ -38,17 +40,24 @@ async def generate_completion(
             tools: Optional tools to provide
             temperature: The temperature to use for generation
             max_tokens: The maximum number of tokens to generate
+            model: Optional model override
+            response_format: Optional response format specification
 
         Returns:
             A tuple of (response, messages)
         """
-        response = await self._openai_client.chat.completions.create(
-            model="gpt-4o",
-            messages=messages,
-            temperature=temperature,
-            tools=[_tool(t) for t in tools or []],
-            max_tokens=max_tokens,
-        )
+        kwargs = {
+            "model": model or "gpt-4o",
+            "messages": messages,
+            "temperature": temperature,
+            "tools": [_tool(t) for t in tools or []],
+            "max_tokens": max_tokens,
+        }
+
+        if response_format:
+            kwargs["response_format"] = response_format
+
+        response = await self._openai_client.chat.completions.create(**kwargs)
         response = response.model_dump()
 
         return response, messages