fix: re-add openai response_format param, add test

lib/crewai/src/crewai/llms/providers/openai/completion.py

@@ -17,6 +17,7 @@ from crewai.events.types.llm_events import LLMCallType
 from crewai.llms.base_llm import BaseLLM
 from crewai.llms.hooks.transport import HTTPTransport
 from crewai.utilities.agent_utils import is_context_length_exceeded
+from crewai.utilities.converter import generate_model_description
 from crewai.utilities.exceptions.context_window_exceeding_exception import (
     LLMContextLengthExceededError,
 )
@@ -245,6 +246,16 @@ class OpenAICompletion(BaseLLM):
         if self.is_o1_model and self.reasoning_effort:
             params["reasoning_effort"] = self.reasoning_effort
 
+        if self.response_format is not None:
+            if isinstance(self.response_format, type) and issubclass(
+                self.response_format, BaseModel
+            ):
+                params["response_format"] = generate_model_description(
+                    self.response_format
+                )
+            elif isinstance(self.response_format, dict):
+                params["response_format"] = self.response_format
+
         if tools:
             params["tools"] = self._convert_tools_for_interference(tools)
             params["tool_choice"] = "auto"
@@ -303,8 +314,11 @@ class OpenAICompletion(BaseLLM):
         """Handle non-streaming chat completion."""
         try:
             if response_model:
+                parse_params = {
+                    k: v for k, v in params.items() if k != "response_format"
+                }
                 parsed_response = self.client.beta.chat.completions.parse(
-                    **params,
+                    **parse_params,
                     response_format=response_model,
                 )
                 math_reasoning = parsed_response.choices[0].message

lib/crewai/tests/cassettes/test_openai_response_format_none.yaml

@@ -0,0 +1,112 @@
+interactions:
+- request:
+    body: '{"messages":[{"role":"user","content":"Say hello in one word"}],"model":"gpt-4o"}'
+    headers:
+      accept:
+      - application/json
+      accept-encoding:
+      - gzip, deflate, zstd
+      connection:
+      - keep-alive
+      content-length:
+      - '81'
+      content-type:
+      - application/json
+      host:
+      - api.openai.com
+      user-agent:
+      - OpenAI/Python 1.109.1
+      x-stainless-arch:
+      - arm64
+      x-stainless-async:
+      - 'false'
+      x-stainless-lang:
+      - python
+      x-stainless-os:
+      - MacOS
+      x-stainless-package-version:
+      - 1.109.1
+      x-stainless-read-timeout:
+      - '600'
+      x-stainless-retry-count:
+      - '0'
+      x-stainless-runtime:
+      - CPython
+      x-stainless-runtime-version:
+      - 3.12.10
+    method: POST
+    uri: https://api.openai.com/v1/chat/completions
+  response:
+    body:
+      string: !!binary |
+        H4sIAAAAAAAAAwAAAP//jJJNT9wwEIbv+RXunDdVNtoPuteiCoEQJ7SiFYqMPcm6OB7LnvAhtP8d
+        OWE3oYDUiw9+5h2/73heMiHAaNgIUDvJqvU2/1nf1K44vVzePG6vr5cPV2V5/nt7eqG36lcLs6Sg
+        u7+o+KD6rqj1FtmQG7AKKBlT1/l6tSjKYr1a96AljTbJGs/5gvKyKBd5cZIXqzfhjozCCBvxJxNC
+        iJf+TBadxifYiGJ2uGkxRtkgbI5FQkAgm25AxmgiS8cwG6Eix+h612doLX2bwoB1F2Xy5jprJ0A6
+        RyxTtt7W7RvZH41Yanygu/iPFGrjTNxVAWUklx6NTB56us+EuO0Dd+8ygA/Ueq6Y7rF/bl4O7WCc
+        8AgPjImlnWgWs0+aVRpZGhsn8wIl1Q71qByHKzttaAKySeSPXj7rPcQ2rvmf9iNQCj2jrnxAbdT7
+        vGNZwLR+X5UdR9wbhojhwSis2GBI36Cxlp0dNgPic2Rsq9q4BoMPZliP2lfqxwkWSyXna8j22SsA
+        AAD//wMAmJrFFCcDAAA=
+    headers:
+      CF-RAY:
+      - 9a3c18dff8580f53-EWR
+      Connection:
+      - keep-alive
+      Content-Encoding:
+      - gzip
+      Content-Type:
+      - application/json
+      Date:
+      - Mon, 24 Nov 2025 21:46:08 GMT
+      Server:
+      - cloudflare
+      Set-Cookie:
+      - FILTERED
+      Strict-Transport-Security:
+      - max-age=31536000; includeSubDomains; preload
+      Transfer-Encoding:
+      - chunked
+      X-Content-Type-Options:
+      - nosniff
+      access-control-expose-headers:
+      - X-Request-ID
+      alt-svc:
+      - h3=":443"; ma=86400
+      cf-cache-status:
+      - DYNAMIC
+      openai-organization:
+      - FILTERED
+      openai-processing-ms:
+      - '1096'
+      openai-project:
+      - FILTERED
+      openai-version:
+      - '2020-10-01'
+      x-envoy-upstream-service-time:
+      - '1138'
+      x-openai-proxy-wasm:
+      - v0.1
+      x-ratelimit-limit-project-requests:
+      - '10000'
+      x-ratelimit-limit-requests:
+      - '10000'
+      x-ratelimit-limit-tokens:
+      - '30000000'
+      x-ratelimit-remaining-project-requests:
+      - '9999'
+      x-ratelimit-remaining-requests:
+      - '9999'
+      x-ratelimit-remaining-tokens:
+      - '29999992'
+      x-ratelimit-reset-project-requests:
+      - 6ms
+      x-ratelimit-reset-requests:
+      - 6ms
+      x-ratelimit-reset-tokens:
+      - 0s
+      x-request-id:
+      - req_670507131d6c455caf0e8cbc30a1a792
+    status:
+      code: 200
+      message: OK
+version: 1

lib/crewai/tests/cassettes/test_openai_response_format_with_dict.yaml

@@ -0,0 +1,113 @@
+interactions:
+- request:
+    body: '{"messages":[{"role":"user","content":"Return a JSON object with a ''status''
+      field set to ''success''"}],"model":"gpt-4o","response_format":{"type":"json_object"}}'
+    headers:
+      accept:
+      - application/json
+      accept-encoding:
+      - gzip, deflate, zstd
+      connection:
+      - keep-alive
+      content-length:
+      - '160'
+      content-type:
+      - application/json
+      host:
+      - api.openai.com
+      user-agent:
+      - OpenAI/Python 1.109.1
+      x-stainless-arch:
+      - arm64
+      x-stainless-async:
+      - 'false'
+      x-stainless-lang:
+      - python
+      x-stainless-os:
+      - MacOS
+      x-stainless-package-version:
+      - 1.109.1
+      x-stainless-read-timeout:
+      - '600'
+      x-stainless-retry-count:
+      - '0'
+      x-stainless-runtime:
+      - CPython
+      x-stainless-runtime-version:
+      - 3.12.10
+    method: POST
+    uri: https://api.openai.com/v1/chat/completions
+  response:
+    body:
+      string: !!binary |
+        H4sIAAAAAAAAA4xSwW6cMBC98xXWnJeKkC274Zr01vbUVopKhLxmACfGdj1D1Wi1/14ZNgtpU6kX
+        hObNe7z3mGMiBOgGSgGql6wGb9Lb9r4191/46eaD+/j509f9Lvs2PP+47u/usIdNZLjDIyp+Yb1T
+        bvAGWTs7wyqgZIyqV7tim+XZrng/AYNr0ERa5zndujTP8m2a7dOsOBN7pxUSlOJ7IoQQx+kZLdoG
+        f0Epss3LZEAi2SGUlyUhIDgTJyCJNLG0DJsFVM4y2sn1sbJCVEAseaQKyvg+KoVEFVT2tGYFbEeS
+        0bQdjVkB0lrHMoae/D6ckdPFoXGdD+5Af1Ch1VZTXweU5Gx0Q+w8TOgpEeJhamJ8FQ58cIPnmt0T
+        Tp/L81kOluoX8OaMsWNplvH11eYNsbpBltrQqkhQUvXYLMyldTk22q2AZBX5by9vac+xte3+R34B
+        lELP2NQ+YKPV67zLWsB4l/9au1Q8GQbC8FMrrFljiL+hwVaOZj4ZoGdiHOpW2w6DD3q+m9bXO8TD
+        tmizYg/JKfkNAAD//wMA0CE0wkADAAA=
+    headers:
+      CF-RAY:
+      - 9a3c18d7de3c80dc-EWR
+      Connection:
+      - keep-alive
+      Content-Encoding:
+      - gzip
+      Content-Type:
+      - application/json
+      Date:
+      - Mon, 24 Nov 2025 21:46:06 GMT
+      Server:
+      - cloudflare
+      Set-Cookie:
+      - FILTERED
+      Strict-Transport-Security:
+      - max-age=31536000; includeSubDomains; preload
+      Transfer-Encoding:
+      - chunked
+      X-Content-Type-Options:
+      - nosniff
+      access-control-expose-headers:
+      - X-Request-ID
+      alt-svc:
+      - h3=":443"; ma=86400
+      cf-cache-status:
+      - DYNAMIC
+      openai-organization:
+      - FILTERED
+      openai-processing-ms:
+      - '424'
+      openai-project:
+      - FILTERED
+      openai-version:
+      - '2020-10-01'
+      x-envoy-upstream-service-time:
+      - '443'
+      x-openai-proxy-wasm:
+      - v0.1
+      x-ratelimit-limit-project-requests:
+      - '10000'
+      x-ratelimit-limit-requests:
+      - '10000'
+      x-ratelimit-limit-tokens:
+      - '30000000'
+      x-ratelimit-remaining-project-requests:
+      - '9999'
+      x-ratelimit-remaining-requests:
+      - '9999'
+      x-ratelimit-remaining-tokens:
+      - '29999983'
+      x-ratelimit-reset-project-requests:
+      - 6ms
+      x-ratelimit-reset-requests:
+      - 6ms
+      x-ratelimit-reset-tokens:
+      - 0s
+      x-request-id:
+      - req_71bc4c9f29f843d6b3788b119850dfde
+    status:
+      code: 200
+      message: OK
+version: 1

lib/crewai/tests/cassettes/test_openai_response_format_with_pydantic_model.yaml

@@ -0,0 +1,116 @@
+interactions:
+- request:
+    body: '{"messages":[{"role":"user","content":"What is the capital of France? Be
+      concise."}],"model":"gpt-4o","response_format":{"type":"json_schema","json_schema":{"name":"AnswerResponse","strict":true,"schema":{"description":"Response
+      model with structured fields.","properties":{"answer":{"description":"The answer
+      to the question","title":"Answer","type":"string"},"confidence":{"description":"Confidence
+      score between 0 and 1","title":"Confidence","type":"number"}},"required":["answer","confidence"],"title":"AnswerResponse","type":"object","additionalProperties":false}}}}'
+    headers:
+      accept:
+      - application/json
+      accept-encoding:
+      - gzip, deflate, zstd
+      connection:
+      - keep-alive
+      content-length:
+      - '571'
+      content-type:
+      - application/json
+      host:
+      - api.openai.com
+      user-agent:
+      - OpenAI/Python 1.109.1
+      x-stainless-arch:
+      - arm64
+      x-stainless-async:
+      - 'false'
+      x-stainless-lang:
+      - python
+      x-stainless-os:
+      - MacOS
+      x-stainless-package-version:
+      - 1.109.1
+      x-stainless-read-timeout:
+      - '600'
+      x-stainless-retry-count:
+      - '0'
+      x-stainless-runtime:
+      - CPython
+      x-stainless-runtime-version:
+      - 3.12.10
+    method: POST
+    uri: https://api.openai.com/v1/chat/completions
+  response:
+    body:
+      string: !!binary |
+        H4sIAAAAAAAAAwAAAP//jFLLbtswELzrK4g9SwFtyA/pmKCH9pRbUVSBwJArmTVFElyqbWr43wvK
+        jqW0KdALDzs7w5ndPWWMgVZQM5AHEeXgTfHQfemO1f0H8cK73fDp82b/iz6uH4+WnC0hTwz3/A1l
+        fGXdSTd4g1E7e4FlQBExqa5225Kv+W5bTsDgFJpE630sSles+bos+L7g2yvx4LREgpp9zRhj7DS9
+        yaJV+BNqxvPXyoBEokeob02MQXAmVUAQaYrCRshnUDob0U6uTw0ISz8wNFA38CiCpgbyJrV0WqGV
+        2EDN76rqvBQI2I0kkn87GrMAhLUuipR/sv50Rc43s8b1Prhn+oMKnbaaDm1AQc4mYxSdhwk9Z4w9
+        TUMZ3+QEH9zgYxvdEafvqvIiB/MWZnC1uoLRRWEWdb7J35FrFUahDS2mClLIA6qZOq9AjEq7BZAt
+        Qv/t5j3tS3Bt+/+RnwEp0UdUrQ+otHybeG4LmI70X223IU+GgTB81xLbqDGkRSjsxGgu9wP0QhGH
+        ttO2x+CDvhxR51tZ7ZFvpFjtIDtnvwEAAP//AwAvoKedTQMAAA==
+    headers:
+      CF-RAY:
+      - 9a3c18cf7fe04253-EWR
+      Connection:
+      - keep-alive
+      Content-Encoding:
+      - gzip
+      Content-Type:
+      - application/json
+      Date:
+      - Mon, 24 Nov 2025 21:46:05 GMT
+      Server:
+      - cloudflare
+      Set-Cookie:
+      - FILTERED
+      Strict-Transport-Security:
+      - max-age=31536000; includeSubDomains; preload
+      Transfer-Encoding:
+      - chunked
+      X-Content-Type-Options:
+      - nosniff
+      access-control-expose-headers:
+      - X-Request-ID
+      alt-svc:
+      - h3=":443"; ma=86400
+      cf-cache-status:
+      - DYNAMIC
+      openai-organization:
+      - FILTERED
+      openai-processing-ms:
+      - '448'
+      openai-project:
+      - FILTERED
+      openai-version:
+      - '2020-10-01'
+      x-envoy-upstream-service-time:
+      - '465'
+      x-openai-proxy-wasm:
+      - v0.1
+      x-ratelimit-limit-project-requests:
+      - '10000'
+      x-ratelimit-limit-requests:
+      - '10000'
+      x-ratelimit-limit-tokens:
+      - '30000000'
+      x-ratelimit-remaining-project-requests:
+      - '9999'
+      x-ratelimit-remaining-requests:
+      - '9999'
+      x-ratelimit-remaining-tokens:
+      - '29999987'
+      x-ratelimit-reset-project-requests:
+      - 6ms
+      x-ratelimit-reset-requests:
+      - 6ms
+      x-ratelimit-reset-tokens:
+      - 0s
+      x-request-id:
+      - req_765510cb1e614ed6a83e665bf7c5a07b
+    status:
+      code: 200
+      message: OK
+version: 1

lib/crewai/tests/llms/openai/test_openai.py

@@ -528,3 +528,50 @@ def test_openai_streaming_with_response_model():
 
         assert "input" not in call_kwargs
         assert "text_format" not in call_kwargs
+
+
+@pytest.mark.vcr(filter_headers=["authorization"])
+def test_openai_response_format_with_pydantic_model():
+    """
+    Test that response_format with a Pydantic BaseModel returns structured output.
+    """
+    from pydantic import BaseModel, Field
+
+    class AnswerResponse(BaseModel):
+        """Response model with structured fields."""
+
+        answer: str = Field(description="The answer to the question")
+        confidence: float = Field(description="Confidence score between 0 and 1")
+
+    llm = LLM(model="gpt-4o", response_format=AnswerResponse)
+    result = llm.call("What is the capital of France? Be concise.")
+
+    assert isinstance(result, AnswerResponse)
+    assert result.answer is not None
+    assert 0 <= result.confidence <= 1
+
+
+@pytest.mark.vcr(filter_headers=["authorization"])
+def test_openai_response_format_with_dict():
+    """
+    Test that response_format with a dict returns JSON output.
+    """
+    import json
+
+    llm = LLM(model="gpt-4o", response_format={"type": "json_object"})
+    result = llm.call("Return a JSON object with a 'status' field set to 'success'")
+
+    parsed = json.loads(result)
+    assert "status" in parsed
+
+
+@pytest.mark.vcr(filter_headers=["authorization"])
+def test_openai_response_format_none():
+    """
+    Test that when response_format is None, the API returns plain text.
+    """
+    llm = LLM(model="gpt-4o", response_format=None)
+    result = llm.call("Say hello in one word")
+
+    assert isinstance(result, str)
+    assert len(result) > 0