Generowanie tekstu

Praca z DeepSeek w Pythonie

James Chapman

AI Curriculum Lead, DataCamp

Jak generowane są odpowiedzi?

Tekst najbardziej prawdopodobny jako dokończenie rozmowy

response = client.chat.completions.create(
    model="deepseek-ai/DeepSeek-V4-Pro",

    messages=[{"role": "user",
             "content": "Hi! I'm James and I'm a Curriculum Manager at DataCamp."}]

)
print(response.choices[0].message.content)

Hi James! It's great to meet you. As a **Curriculum Manager at DataCamp**, you must
play a key role in shaping the learning experience for data science and analytics
students. How can I assist you today?

Odpowiedź jest niedeterministyczna (z natury losowa)

Chatbot obsługi klienta odbierający dwa semantycznie podobne zapytania.

Kontrolowanie losowości odpowiedzi

temperature: kontrola determinizmu od 0 (silnie deterministyczny) do 2 (bardzo losowy)
Ignorowany przy włączonym rozumowaniu → wyłącz je, by go użyć

response = client.chat.completions.create(
    model="deepseek-ai/DeepSeek-V4-Pro",
    extra_body={"thinking": {"type": "disabled"}},
    messages=[{"role": "user",
             "content": "Hi! I'm James and I'm a Curriculum Manager at DataCamp."}],

    temperature=2

)
print(response.choices[0].message.content)

Hi James! Welcome to DataCamp - it's great to connect with someone who works on
shaping learning experiences for data skills. What does your role typically involve
as a Curriculum Manager at **DataCamp?** Perhaps one-on95 kWielpulses.Z.blog.from...

Generowanie tekstu: marketing

prompt = "Generate a powerful tagline for a new electric vehicle 
that highlights innovation and sustainability."

response = client.chat.completions.create(
    model="deepseek-ai/DeepSeek-V4-Pro",
    messages=[{"role": "user", "content": prompt}]
)

print(response.choices[0].message.content)

**"Drive the Future - Zero Emissions, Infinite Innovation."**  
This tagline captures the essence of cutting-edge technology and eco-conscious...

Porządkowanie odpowiedzi

prompt = "Generate a powerful tagline for a new electric vehicle 
that highlights innovation and sustainability. Return only the tagline
without formatting."

# Create a request to the chat model
# ...

"Drive the Future - Clean, Smart, Electric."

Generowanie tekstu: opis produktu

prompt = """Write a compelling e-commmerce product description for the UltraFit
Smartwatch in paragraph form. Highlight its key features:

10-day battery life, 24/7 heart rate and sleep tracking, built-in GPS, water
resistance up to 50 meters, and lightweight design.

Use a persuasive and engaging tone to appeal to fitness enthusiasts and busy
professionals.
"""


response = client.chat.completions.create(
    model="deepseek-ai/DeepSeek-V4-Pro",
    messages=[{"role": "user", "content": prompt}])

Generowanie tekstu: opis produktu

print(response.choices[0].message.content)

"The UltraFit Smartwatch is your all-in-one health and fitness companion. 
Stay on top of your wellness with 24/7 heart rate and sleep tracking, 
while the built-in GPS keeps you on course during runs. 
With a 10-day battery life, you can track your progress without constant recharging. 
Designed for both workouts and daily wear, its  lightweight build and 50-meter
water resistance make it the perfect smartwatch for an active lifestyle."

Iteruj! Dodaj więcej szczegółów do promptu i ponów zapytanie
Następny film → przekazywanie modelowi przykładów

Czas na praktykę!

Praca z DeepSeek w Pythonie