Python: updated declarative samples and handling of non-pydantic response formats (#5022)

* updated declarative samples and handling of non-pydantic response formats

* fixed from comments

* update docstring

python/packages/ollama/agent_framework_ollama/_chat_client.py

@@ -382,7 +382,10 @@ class OllamaChatClient(
             except Exception as ex:
                 raise ChatClientException(f"Ollama chat request failed : {ex}", ex) from ex
 
-            return self._parse_response_from_ollama(response)
+            return self._parse_response_from_ollama(
+                response,
+                response_format=validated_options.get("response_format"),
+            )
 
         return _get_response()
 
@@ -536,7 +539,12 @@ class OllamaChatClient(
             created_at=response.created_at,
         )
 
-    def _parse_response_from_ollama(self, response: OllamaChatResponse) -> ChatResponse:
+    def _parse_response_from_ollama(
+        self,
+        response: OllamaChatResponse,
+        *,
+        response_format: Any | None = None,
+    ) -> ChatResponse:
         contents = self._parse_contents_from_ollama(response)
 
         return ChatResponse(
@@ -547,6 +555,7 @@ class OllamaChatClient(
                 input_token_count=response.prompt_eval_count,
                 output_token_count=response.eval_count,
             ),
+            response_format=response_format,
         )
 
     def _parse_tool_calls_from_ollama(self, tool_calls: Sequence[OllamaMessage.ToolCall]) -> list[Content]:

python/packages/ollama/tests/test_ollama_chat_client.py

@@ -248,6 +248,33 @@ async def test_cmc(
     assert result.text == "test"
 
 
+@patch.object(AsyncClient, "chat", new_callable=AsyncMock)
+async def test_cmc_response_format_dict(
+    mock_chat: AsyncMock,
+    ollama_unit_test_env: dict[str, str],
+    chat_history: list[Message],
+) -> None:
+    mock_chat.return_value = OllamaChatResponse(
+        message=OllamaMessage(content='{"answer": "test"}', role="assistant"),
+        model="test",
+        eval_count=1,
+        prompt_eval_count=1,
+        created_at="2024-01-01T00:00:00Z",
+    )
+    chat_history.append(Message(text="hello world", role="system"))
+    chat_history.append(Message(text="hello world", role="user"))
+
+    ollama_client = OllamaChatClient()
+    result = await ollama_client.get_response(
+        messages=chat_history,
+        options={"response_format": {"type": "object", "properties": {"answer": {"type": "string"}}}},
+    )
+
+    assert result.value is not None
+    assert isinstance(result.value, dict)
+    assert result.value["answer"] == "test"
+
+
 @patch.object(AsyncClient, "chat", new_callable=AsyncMock)
 async def test_cmc_reasoning(
     mock_chat: AsyncMock,