Wyszukiwanie w sieci, które działa

Praca z API Responses OpenAI

James Chapman

AI Curriculum Manager, DataCamp

Potrzeba aktualnych informacji

LLM-y → mają datę graniczną wiedzy
- Może być >1 rok:
  - np. GPT 4.1 — czerwiec 2024
- Problem przy kodowaniu, badaniach i nie tylko...

Data graniczna wiedzy w praktyce

response = client.responses.create(
    model="gpt-5.4-mini",
    input="What is the current temperature in Tokyo, Japan?"
)

print(response.output_text)

I can't access real-time data, so I can't tell you the current temperature in
Tokyo right now.

Pierwsze kroki z wyszukiwaniem w sieci

response = client.responses.create(
    model="gpt-5.4-mini",

    tools=[{"type": "web_search"}],

    input="What is the current temperature in Tokyo, Japan?"
)

print(response.output_text)

Current temperature in Tokyo (Japan): 46°F (8°C), foggy. (As of Nov 25, 2025,
Tokyo time.)

Wyszukiwanie w sieci od środka

for item in response.output:
    print(item)

ResponseReasoningItem(id='...', summary=[], type='reasoning', content=None,
                      encrypted_content=None, status=None)

ResponseFunctionWebSearch(id='...', action=ActionSearch(query='weather: Tokyo, Japan',
                          type='search', sources=None), status='completed',
                          type='web_search_call')

ResponseReasoningItem(id='...', summary=[], type='reasoning', content=None,
                      encrypted_content=None, status=None)

ResponseOutputMessage(id='...', content=[ResponseOutputText(annotations=[],
                      text='Current temperature in Tokyo (Japan)...', type='output_text',
                      logprobs=[])], role='assistant', status='completed', type='message')

Uwzględnianie wszystkich źródeł

response = client.responses.create(
    model="gpt-5.4-mini",
    tools=[{"type": "web_search"}],
    input="What is the current temperature in Tokyo, Japan?",

    include=["web_search_call.action.sources"]

)

for item in response.output:
    if item.type == "web_search_call":
        print(item)

ResponseFunctionWebSearch(id='...', action=ActionSearch(query='weather: Tokyo, Japan',
                          type='search', sources=[ActionSearchSource(type='api', url=None,
                          name='oai-weather')]), status='completed', type='web_search_call')

Inne typy źródeł

response = client.responses.create(
    model="gpt-5.4-mini",
    tools=[{"type": "web_search"}],
    input="What is the latest Python version released?",
    include=["web_search_call.action.sources"]
)

for item in response.output:
    if item.type == "web_search_call":
        print(item.action.sources)

Inne typy źródeł

[{'type': 'url', 'url': 'https://www.python.org/doc/versions/'},
 {'type': 'url', 'url': 'https://test.python.org/doc/versions/'},
 {'type': 'url', 'url': 'https://www.techradar.com/pro/...'}]

[{'type': 'url', 'url': 'https://www.python.org/downloads/latest/python3.14/'},
 {'type': 'url', 'url': 'https://www.python.org/downloads/release/python-3140rc1/'},
 {'type': 'url', 'url': 'https://blog.python.org/2025/10/python-3140-final-is-here.html'},
 {'type': 'url', 'url': 'https://www.python.org/downloads/release/python-3140rc2/'},
 {'type': 'url', 'url': 'https://peps.python.org/pep-0745/'},
 {'type': 'url', 'url': 'https://blog.python.org/2025/05/python-3140-beta-1-is-here.html'},
 {'type': 'url', 'url': 'https://www.python.org/downloads/release/python-3140rc3/'},
 {'type': 'url', 'url': 'https://blog.python.org/2025/10/'},
 {'type': 'url', 'url': 'https://blog.python.org/2025/07/python-314-release-candidate-1-is-go.html'},
 {'type': 'url', 'url': 'https://blog.python.org/2025/01/python-3140-alpha-4-is-out.html'},
 {'type': 'url', 'url': 'https://mail.python.org/archives/list/python-announce-list...'}]

Czas na praktykę!

Praca z API Responses OpenAI