Samouczek: pierwsze kroki z interfejsem Gemini API


Wyświetl w AI od Google Uruchom w Google Colab Wyświetl źródło w GitHubie

Z tego krótkiego wprowadzenia dowiesz się, jak używać pakietu SDK Pythona na potrzeby interfejsu Gemini API. daje dostęp do dużych modeli językowych Gemini (Google). W tym krótkim wprowadzeniu dowiesz się, jak:

  1. Aby korzystać z Gemini, skonfiguruj środowisko programistyczne i dostęp do interfejsu API.
  2. Generuj odpowiedzi tekstowe na podstawie danych wejściowych.
  3. Generuj odpowiedzi tekstowe na podstawie danych wejściowych multimodalnych (tekstu i obrazów).
  4. Używaj Gemini do rozmów wieloetapowych (czatu).
  5. Używaj wektorów dystrybucyjnych dla dużych modeli językowych.

Wymagania wstępne

Możesz uruchomić to krótkie wprowadzenie w Google Colab, który uruchamia ten notatnik bezpośrednio w przeglądarce i nie wymaga dodatkowych konfigurację środowiska.

Aby dokończyć to krótkie wprowadzenie lokalnie, upewnij się, że programowanie spełnia te wymagania:

  • Python w wersji 3.9 lub nowszej
  • Instalacja programu jupyter w celu uruchomienia notatnika.

Konfiguracja

Zainstaluj pakiet SDK Pythona

Pakiet Python SDK dla interfejsu Gemini API znajduje się w sekcji google-generativeai. Zainstaluj zależność za pomocą pip:

pip install -q -U google-generativeai

Importuj pakiety

Zaimportuj niezbędne pakiety.

import pathlib
import textwrap

import google.generativeai as genai

from IPython.display import display
from IPython.display import Markdown

def to_markdown(text):
  text = text.replace('•', '  *')
  return Markdown(textwrap.indent(text, '> ', predicate=lambda _: True))

# Used to securely store your API key
from google.colab import userdata

Skonfiguruj klucz interfejsu API

Aby móc korzystać z Gemini API, musisz najpierw uzyskać klucz API. Jeśli jeszcze go nie masz, utwórz klucz jednym kliknięciem w Google AI Studio.

Uzyskiwanie klucza interfejsu API

W Colab dodaj klucz do menedżera obiektów tajnych w sekcji „🔑” w panelu po lewej stronie. Nazwij go GOOGLE_API_KEY.

Po uzyskaniu klucza interfejsu API przekaż go do pakietu SDK. Można to zrobić na dwa sposoby:

  • Umieść klucz w zmiennej środowiskowej GOOGLE_API_KEY (pakiet SDK automatycznie odebrać ją stamtąd).
  • Przekaż klucz urządzeniu genai.configure(api_key=...)
# Or use `os.getenv('GOOGLE_API_KEY')` to fetch an environment variable.
GOOGLE_API_KEY=userdata.get('GOOGLE_API_KEY')

genai.configure(api_key=GOOGLE_API_KEY)

Wyświetlenie listy modeli

Teraz możesz wywołać interfejs Gemini API. Użyj list_models, aby zobaczyć dostępne Modele Gemini:

  • gemini-1.5-flash: nasz najszybszy model multimodalny
  • gemini-1.5-pro: nasz najbardziej zaawansowany i inteligentny model multimodalny.
for m in genai.list_models():
  if 'generateContent' in m.supported_generation_methods:
    print(m.name)

Generowanie tekstu na podstawie danych wejściowych

W przypadku promptów tekstowych użyj modelu Gemini 1.5 lub Gemini 1.0 Pro:

model = genai.GenerativeModel('gemini-1.5-flash')

Metoda generate_content sprawdza się w różnych przypadkach użycia, w tym: czat wieloetapowy i multimodalne dane wejściowe w zależności od modelu bazowego. obsługuje. Dostępne modele obsługują tylko tekst i obrazy jako dane wejściowe, a tekst jako dane wyjściowe.

W najprostszym przypadku możesz przekazać ciąg promptu do funkcji GenerativeModel.generate_content :

%%time
response = model.generate_content("What is the meaning of life?")

CPU times: user 110 ms, sys: 12.3 ms, total: 123 ms
Wall time: 8.25 s

W prostych przypadkach wystarczy akcesor response.text. Do wyświetlenia w formacie Markdown, użyj funkcji to_markdown:

to_markdown(response.text)

The query of life's purpose has perplexed people across centuries, cultures, and continents. While there is no universally recognized response, many ideas have been put forth, and the response is frequently dependent on individual ideas, beliefs, and life experiences.

1.  **Happiness and Well-being:** Many individuals believe that the goal of life is to attain personal happiness and well-being. This might entail locating pursuits that provide joy, establishing significant connections, caring for one's physical and mental health, and pursuing personal goals and interests.

2.  **Meaningful Contribution:** Some believe that the purpose of life is to make a meaningful contribution to the world. This might entail pursuing a profession that benefits others, engaging in volunteer or charitable activities, generating art or literature, or inventing.

3.  **Self-realization and Personal Growth:** The pursuit of self-realization and personal development is another common goal in life. This might entail learning new skills, pushing one's boundaries, confronting personal obstacles, and evolving as a person.

4.  **Ethical and Moral Behavior:** Some believe that the goal of life is to act ethically and morally. This might entail adhering to one's moral principles, doing the right thing even when it is difficult, and attempting to make the world a better place.

5.  **Spiritual Fulfillment:** For some, the purpose of life is connected to spiritual or religious beliefs. This might entail seeking a connection with a higher power, practicing religious rituals, or following spiritual teachings.

6.  **Experiencing Life to the Fullest:** Some individuals believe that the goal of life is to experience all that it has to offer. This might entail traveling, trying new things, taking risks, and embracing new encounters.

7.  **Legacy and Impact:** Others believe that the purpose of life is to leave a lasting legacy and impact on the world. This might entail accomplishing something noteworthy, being remembered for one's contributions, or inspiring and motivating others.

8.  **Finding Balance and Harmony:** For some, the purpose of life is to find balance and harmony in all aspects of their lives. This might entail juggling personal, professional, and social obligations, seeking inner peace and contentment, and living a life that is in accordance with one's values and beliefs.

Ultimately, the meaning of life is a personal journey, and different individuals may discover their own unique purpose through their experiences, reflections, and interactions with the world around them.

Jeśli interfejs API nie zwrócił wyniku, użyj GenerateContentResponse.prompt_feedback aby sprawdzić, czy nie został zablokowany ze względu na zagrożenia dotyczące bezpieczeństwa.

response.prompt_feedback

safety_ratings {
  category: HARM_CATEGORY_SEXUALLY_EXPLICIT
  probability: NEGLIGIBLE
}
safety_ratings {
  category: HARM_CATEGORY_HATE_SPEECH
  probability: NEGLIGIBLE
}
safety_ratings {
  category: HARM_CATEGORY_HARASSMENT
  probability: NEGLIGIBLE
}
safety_ratings {
  category: HARM_CATEGORY_DANGEROUS_CONTENT
  probability: NEGLIGIBLE
}

Gemini może wygenerować wiele możliwych odpowiedzi na 1 prompt. Te Możliwe odpowiedzi to candidates. Możesz je przejrzeć, aby wybrać tę najodpowiedniejszą jako odpowiedź.

Wyświetl kandydujące odpowiedzi z: GenerateContentResponse.candidates:

response.candidates

[
  content {
    parts {
      text: "The query of life\'s purpose has perplexed people across centuries, cultures, and continents. While there is no universally recognized response, many ideas have been put forth, and the response is frequently dependent on individual ideas, beliefs, and life experiences.\n\n1. **Happiness and Well-being:** Many individuals believe that the goal of life is to attain personal happiness and well-being. This might entail locating pursuits that provide joy, establishing significant connections, caring for one\'s physical and mental health, and pursuing personal goals and interests.\n\n2. **Meaningful Contribution:** Some believe that the purpose of life is to make a meaningful contribution to the world. This might entail pursuing a profession that benefits others, engaging in volunteer or charitable activities, generating art or literature, or inventing.\n\n3. **Self-realization and Personal Growth:** The pursuit of self-realization and personal development is another common goal in life. This might entail learning new skills, pushing one\'s boundaries, confronting personal obstacles, and evolving as a person.\n\n4. **Ethical and Moral Behavior:** Some believe that the goal of life is to act ethically and morally. This might entail adhering to one\'s moral principles, doing the right thing even when it is difficult, and attempting to make the world a better place.\n\n5. **Spiritual Fulfillment:** For some, the purpose of life is connected to spiritual or religious beliefs. This might entail seeking a connection with a higher power, practicing religious rituals, or following spiritual teachings.\n\n6. **Experiencing Life to the Fullest:** Some individuals believe that the goal of life is to experience all that it has to offer. This might entail traveling, trying new things, taking risks, and embracing new encounters.\n\n7. **Legacy and Impact:** Others believe that the purpose of life is to leave a lasting legacy and impact on the world. This might entail accomplishing something noteworthy, being remembered for one\'s contributions, or inspiring and motivating others.\n\n8. **Finding Balance and Harmony:** For some, the purpose of life is to find balance and harmony in all aspects of their lives. This might entail juggling personal, professional, and social obligations, seeking inner peace and contentment, and living a life that is in accordance with one\'s values and beliefs.\n\nUltimately, the meaning of life is a personal journey, and different individuals may discover their own unique purpose through their experiences, reflections, and interactions with the world around them."
    }
    role: "model"
  }
  finish_reason: STOP
  index: 0
  safety_ratings {
    category: HARM_CATEGORY_SEXUALLY_EXPLICIT
    probability: NEGLIGIBLE
  }
  safety_ratings {
    category: HARM_CATEGORY_HATE_SPEECH
    probability: NEGLIGIBLE
  }
  safety_ratings {
    category: HARM_CATEGORY_HARASSMENT
    probability: NEGLIGIBLE
  }
  safety_ratings {
    category: HARM_CATEGORY_DANGEROUS_CONTENT
    probability: NEGLIGIBLE
  }
]

Domyślnie model zwraca odpowiedź po zakończeniu całego generowania proces tworzenia konta. Możesz też transmitować odpowiedź w trakcie jej generowania. model zwróci fragmenty odpowiedzi zaraz po ich wygenerowaniu.

Aby przesyłać odpowiedzi strumieniowo, użyj funkcji GenerativeModel.generate_content(..., stream=True).

%%time
response = model.generate_content("What is the meaning of life?", stream=True)

CPU times: user 102 ms, sys: 25.1 ms, total: 128 ms
Wall time: 7.94 s

for chunk in response:
  print(chunk.text)
  print("_"*80)

The query of life's purpose has perplexed people across centuries, cultures, and
________________________________________________________________________________
 continents. While there is no universally recognized response, many ideas have been put forth, and the response is frequently dependent on individual ideas, beliefs, and life experiences
________________________________________________________________________________
.

1.  **Happiness and Well-being:** Many individuals believe that the goal of life is to attain personal happiness and well-being. This might entail locating pursuits that provide joy, establishing significant connections, caring for one's physical and mental health, and pursuing personal goals and aspirations.

2.  **Meaning
________________________________________________________________________________
ful Contribution:** Some believe that the purpose of life is to make a meaningful contribution to the world. This might entail pursuing a profession that benefits others, engaging in volunteer or charitable activities, generating art or literature, or inventing.

3.  **Self-realization and Personal Growth:** The pursuit of self-realization and personal development is another common goal in life. This might entail learning new skills, exploring one's interests and abilities, overcoming obstacles, and becoming the best version of oneself.

4.  **Connection and Relationships:** For many individuals, the purpose of life is found in their relationships with others. This might entail building
________________________________________________________________________________
 strong bonds with family and friends, fostering a sense of community, and contributing to the well-being of those around them.

5.  **Spiritual Fulfillment:** For those with religious or spiritual beliefs, the purpose of life may be centered on seeking spiritual fulfillment or enlightenment. This might entail following religious teachings, engaging in spiritual practices, or seeking a deeper understanding of the divine.

6.  **Experiencing the Journey:** Some believe that the purpose of life is simply to experience the journey itself, with all its joys and sorrows. This perspective emphasizes embracing the present moment, appreciating life's experiences, and finding meaning in the act of living itself.

7.  **Legacy and Impact:** For others, the goal of life is to leave a lasting legacy or impact on the world. This might entail making a significant contribution to a particular field, leaving a positive mark on future generations, or creating something that will be remembered and cherished long after one's lifetime.

Ultimately, the meaning of life is a personal and subjective question, and there is no single, universally accepted answer. It is about discovering what brings you fulfillment, purpose, and meaning in your own life, and living in accordance with those values.
________________________________________________________________________________

Podczas strumieniowego przesyłania danych niektóre atrybuty odpowiedzi nie są dostępne, dopóki nie wykonasz iteracji we wszystkich fragmentach odpowiedzi. Oto przykład:

response = model.generate_content("What is the meaning of life?", stream=True)

Atrybut prompt_feedback działa:

response.prompt_feedback

safety_ratings {
  category: HARM_CATEGORY_SEXUALLY_EXPLICIT
  probability: NEGLIGIBLE
}
safety_ratings {
  category: HARM_CATEGORY_HATE_SPEECH
  probability: NEGLIGIBLE
}
safety_ratings {
  category: HARM_CATEGORY_HARASSMENT
  probability: NEGLIGIBLE
}
safety_ratings {
  category: HARM_CATEGORY_DANGEROUS_CONTENT
  probability: NEGLIGIBLE
}

Atrybuty takie jak text nie:

try:
  response.text
except Exception as e:
  print(f'{type(e).__name__}: {e}')

IncompleteIterationError: Please let the response complete iteration before accessing the final accumulated
attributes (or call `response.resolve()`)

Generuj tekst na podstawie danych wejściowych z obrazem i tekstem

Gemini oferuje różne modele, które obsługują wprowadzanie multimodalne (Gemini 1.5 modeli), dzięki czemu można wprowadzać zarówno tekst, jak i obrazy. Zapoznaj się z wymagania dotyczące obrazów w promptach.

Jeśli prompt zawiera zarówno tekst, jak i obrazy, użyj Gemini 1.5 z parametrem Metoda GenerativeModel.generate_content do generowania tekstowych danych wyjściowych:

Dołączmy obraz:

curl -o image.jpg https://t0.gstatic.com/licensed-image?q=tbn:ANd9GcQ_Kevbk21QBRy-PgB4kQpS79brbmmEG7m3VOTShAn4PecDU5H5UxrJxE3Dw1JiaG17V88QIol19-3TM2wCHw

% Total    % Received % Xferd  Average Speed   Time    Time     Time  Current
                                 Dload  Upload   Total   Spent    Left  Speed

  0     0    0     0    0     0      0      0 --:--:-- --:--:-- --:--:--     0
100  405k  100  405k    0     0  6982k      0 --:--:-- --:--:-- --:--:-- 7106k

import PIL.Image

img = PIL.Image.open('image.jpg')
img

png

Użyj modelu Gemini 1.5 i przekaż obraz do modelu za pomocą funkcji generate_content.

model = genai.GenerativeModel('gemini-1.5-flash')

response = model.generate_content(img)

to_markdown(response.text)

Chicken Teriyaki Meal Prep Bowls with brown rice, roasted broccoli and bell peppers.

Aby w promptie podać zarówno tekst, jak i obrazy, przekaż listę zawierającą ciągi znaków oraz obrazy:

response = model.generate_content(["Write a short, engaging blog post based on this picture. It should include a description of the meal in the photo and talk about my journey meal prepping.", img], stream=True)
response.resolve()

to_markdown(response.text)

Meal prepping is a great way to save time and money, and it can also help you to eat healthier. This meal is a great example of a healthy and delicious meal that can be easily prepped ahead of time.

This meal features brown rice, roasted vegetables, and chicken teriyaki. The brown rice is a whole grain that is high in fiber and nutrients. The roasted vegetables are a great way to get your daily dose of vitamins and minerals. And the chicken teriyaki is a lean protein source that is also packed with flavor.

This meal is easy to prepare ahead of time. Simply cook the brown rice, roast the vegetables, and cook the chicken teriyaki. Then, divide the meal into individual containers and store them in the refrigerator. When you're ready to eat, simply grab a container and heat it up.

This meal is a great option for busy people who are looking for a healthy and delicious way to eat. It's also a great meal for those who are trying to lose weight or maintain a healthy weight.

If you're looking for a healthy and delicious meal that can be easily prepped ahead of time, this meal is a great option. Give it a try today!

Rozmowy na czacie

Gemini umożliwia swobodne prowadzenie rozmów na różnych etapach. Klasa ChatSession upraszcza proces, zarządzając stanem rozmowy, więc w przeciwieństwie do generate_content, nie musisz przechowywać historię rozmów w formie listy.

Zainicjuj czat:

model = genai.GenerativeModel('gemini-1.5-flash')
chat = model.start_chat(history=[])
chat

<google.generativeai.generative_models.ChatSession at 0x7b7b68250100>

ChatSession.send_message zwraca ten sam typ GenerateContentResponse co GenerativeModel.generate_content Twoja wiadomość i odpowiedź są też dodawane do historii czatu:

response = chat.send_message("In one sentence, explain how a computer works to a young child.")
to_markdown(response.text)

A computer is like a very smart machine that can understand and follow our instructions, help us with our work, and even play games with us!

chat.history

[
  parts {
    text: "In one sentence, explain how a computer works to a young child."
  }
  role: "user",
  parts {
    text: "A computer is like a very smart machine that can understand and follow our instructions, help us with our work, and even play games with us!"
  }
  role: "model"
]

Aby kontynuować rozmowę, możesz nadal wysyłać wiadomości. Użyj stream=True argument do przesyłania strumieniowego czatu:

response = chat.send_message("Okay, how about a more detailed explanation to a high schooler?", stream=True)

for chunk in response:
  print(chunk.text)
  print("_"*80)

A computer works by following instructions, called a program, which tells it what to
________________________________________________________________________________
 do. These instructions are written in a special language that the computer can understand, and they are stored in the computer's memory. The computer's processor
________________________________________________________________________________
, or CPU, reads the instructions from memory and carries them out, performing calculations and making decisions based on the program's logic. The results of these calculations and decisions are then displayed on the computer's screen or stored in memory for later use.

To give you a simple analogy, imagine a computer as a
________________________________________________________________________________
 chef following a recipe. The recipe is like the program, and the chef's actions are like the instructions the computer follows. The chef reads the recipe (the program) and performs actions like gathering ingredients (fetching data from memory), mixing them together (performing calculations), and cooking them (processing data). The final dish (the output) is then presented on a plate (the computer screen).

In summary, a computer works by executing a series of instructions, stored in its memory, to perform calculations, make decisions, and display or store the results.
________________________________________________________________________________

Obiekty glm.Content zawierają listę obiektów glm.Part, z których każdy zawiera tekst (ciąg znaków) lub inline_data (glm.Blob), gdzie obiekt blob zawiera dane binarne i mime_type. Historia czatu jest dostępna na liście glm.Content obiekty w ChatSession.history:

for message in chat.history:
  display(to_markdown(f'**{message.role}**: {message.parts[0].text}'))

**user**: In one sentence, explain how a computer works to a young child.

**model**: A computer is like a very smart machine that can understand and follow our instructions, help us with our work, and even play games with us!

**user**: Okay, how about a more detailed explanation to a high schooler?

**model**: A computer works by following instructions, called a program, which tells it what to do. These instructions are written in a special language that the computer can understand, and they are stored in the computer's memory. The computer's processor, or CPU, reads the instructions from memory and carries them out, performing calculations and making decisions based on the program's logic. The results of these calculations and decisions are then displayed on the computer's screen or stored in memory for later use.

To give you a simple analogy, imagine a computer as a chef following a recipe. The recipe is like the program, and the chef's actions are like the instructions the computer follows. The chef reads the recipe (the program) and performs actions like gathering ingredients (fetching data from memory), mixing them together (performing calculations), and cooking them (processing data). The final dish (the output) is then presented on a plate (the computer screen).

In summary, a computer works by executing a series of instructions, stored in its memory, to perform calculations, make decisions, and display or store the results.

Policz tokeny

Duże modele językowe mają okno kontekstu, a jego długość jest często mierzona za pomocą liczby tokenów. Dzięki interfejsowi Gemini API możesz: określać liczbę tokenów na dowolny obiekt genai.protos.Content. W w najprostszym przypadku możesz przekazać ciąg zapytania do funkcji GenerativeModel.count_tokens w następujący sposób:

model.count_tokens("What is the meaning of life?")

total_tokens: 7

Możesz też sprawdzić token_count dla: ChatSession:

model.count_tokens(chat.history)

total_tokens: 501

Korzystanie z wektorów dystrybucyjnych

Umieszczanie to technika używana do przedstawiania informacji w formie listy liczb zmiennoprzecinkowych w tablicy. Za pomocą Gemini możesz reprezentować tekst (słowa, zdania i bloki) tekstu) w formie wektorowej, co ułatwia porównanie wektory dystrybucyjne. na przykład dwa teksty o podobnej tematyce, powinna mieć podobne wektory dystrybucyjne, które można rozpoznać za pomocą technik porównawczych, takich jak podobieństwo cosinusowe. Więcej informacji i dlaczego warto korzystać z reprezentacji właściwościowych, zapoznaj się z sekcją Wektory dystrybucyjne .

Do generowania wektorów dystrybucyjnych użyj metody embed_content. Metoda obsługuje dla następujących zadań (task_type):

Typ zadania Opis
RETRIEVAL_QUERY Określa, że dany tekst jest zapytaniem w ustawieniach wyszukiwania/pobierania.
RETRIEVAL_DOCUMENT Określa, że dany tekst jest dokumentem w ustawieniu wyszukiwania/pobierania. Użycie tego typu zadania wymaga: title.
SEMANTIC_SIMILARITY Określa, który tekst będzie używany na potrzeby funkcji podobieństwo semantycznego (STS).
KLASYFIKACJA Określa, że wektory dystrybucyjne będą używane do klasyfikacji.
KLASTEROWANIE Określa, że wektory dystrybucyjne będą używane do grupowania.

To spowoduje wygenerowanie wektora dystrybucyjnego dla pojedynczego ciągu tekstowego na potrzeby pobierania dokumentu:

result = genai.embed_content(
    model="models/embedding-001",
    content="What is the meaning of life?",
    task_type="retrieval_document",
    title="Embedding of single string")

# 1 input > 1 vector output
print(str(result['embedding'])[:50], '... TRIMMED]')

[-0.003216741, -0.013358698, -0.017649598, -0.0091 ... TRIMMED]

Aby obsługiwać partie ciągów znaków, przekaż listę ciągów w content:

result = genai.embed_content(
    model="models/embedding-001",
    content=[
      'What is the meaning of life?',
      'How much wood would a woodchuck chuck?',
      'How does the brain work?'],
    task_type="retrieval_document",
    title="Embedding of list of strings")

# A list of inputs > A list of vectors output
for v in result['embedding']:
  print(str(v)[:50], '... TRIMMED ...')

[0.0040260437, 0.004124458, -0.014209415, -0.00183 ... TRIMMED ...
[-0.004049845, -0.0075574904, -0.0073463684, -0.03 ... TRIMMED ...
[0.025310587, -0.0080734305, -0.029902633, 0.01160 ... TRIMMED ...

Funkcja genai.embed_content akceptuje ciągi tekstowe lub listy ciągów, ale bazuje na typie genai.protos.Content (takim jak GenerativeModel.generate_content). Obiekty glm.Content są głównymi jednostkami konwersacji w interfejsie API.

Obiekt genai.protos.Content jest multimodalny, a obiekt embed_content obsługuje tylko wektory dystrybucyjne tekstu. Taki układ zapewnia interfejsowi API możliwość rozszerzenia na wektory dystrybucyjne multimodalne.

response.candidates[0].content

parts {
  text: "A computer works by following instructions, called a program, which tells it what to do. These instructions are written in a special language that the computer can understand, and they are stored in the computer\'s memory. The computer\'s processor, or CPU, reads the instructions from memory and carries them out, performing calculations and making decisions based on the program\'s logic. The results of these calculations and decisions are then displayed on the computer\'s screen or stored in memory for later use.\n\nTo give you a simple analogy, imagine a computer as a chef following a recipe. The recipe is like the program, and the chef\'s actions are like the instructions the computer follows. The chef reads the recipe (the program) and performs actions like gathering ingredients (fetching data from memory), mixing them together (performing calculations), and cooking them (processing data). The final dish (the output) is then presented on a plate (the computer screen).\n\nIn summary, a computer works by executing a series of instructions, stored in its memory, to perform calculations, make decisions, and display or store the results."
}
role: "model"

result = genai.embed_content(
    model = 'models/embedding-001',
    content = response.candidates[0].content)

# 1 input > 1 vector output
print(str(result['embedding'])[:50], '... TRIMMED ...')

[-0.013921871, -0.03504407, -0.0051786783, 0.03113 ... TRIMMED ...

Podobnie historia czatu zawiera listę obiektów genai.protos.Content, który można przekazać bezpośrednio do funkcji embed_content:

chat.history

[
  parts {
    text: "In one sentence, explain how a computer works to a young child."
  }
  role: "user",
  parts {
    text: "A computer is like a very smart machine that can understand and follow our instructions, help us with our work, and even play games with us!"
  }
  role: "model",
  parts {
    text: "Okay, how about a more detailed explanation to a high schooler?"
  }
  role: "user",
  parts {
    text: "A computer works by following instructions, called a program, which tells it what to do. These instructions are written in a special language that the computer can understand, and they are stored in the computer\'s memory. The computer\'s processor, or CPU, reads the instructions from memory and carries them out, performing calculations and making decisions based on the program\'s logic. The results of these calculations and decisions are then displayed on the computer\'s screen or stored in memory for later use.\n\nTo give you a simple analogy, imagine a computer as a chef following a recipe. The recipe is like the program, and the chef\'s actions are like the instructions the computer follows. The chef reads the recipe (the program) and performs actions like gathering ingredients (fetching data from memory), mixing them together (performing calculations), and cooking them (processing data). The final dish (the output) is then presented on a plate (the computer screen).\n\nIn summary, a computer works by executing a series of instructions, stored in its memory, to perform calculations, make decisions, and display or store the results."
  }
  role: "model"
]

result = genai.embed_content(
    model = 'models/embedding-001',
    content = chat.history)

# 1 input > 1 vector output
for i,v in enumerate(result['embedding']):
  print(str(v)[:50], '... TRIMMED...')

[-0.014632266, -0.042202696, -0.015757175, 0.01548 ... TRIMMED...
[-0.010979066, -0.024494737, 0.0092659835, 0.00803 ... TRIMMED...
[-0.010055617, -0.07208932, -0.00011750793, -0.023 ... TRIMMED...
[-0.013921871, -0.03504407, -0.0051786783, 0.03113 ... TRIMMED...

Zaawansowane przypadki użycia

W kolejnych sekcjach omawiamy zaawansowane przypadki użycia i szczegóły niższego poziomu Pakiet Python SDK dla interfejsu Gemini API.

Ustawienia bezpieczeństwa

Argument safety_settings pozwala określić, co model blokuje i zezwala zarówno na prompty, jak i odpowiedzi. Domyślnie ustawienia bezpieczeństwa blokują treści o średnim lub wysokim prawdopodobieństwie niebezpiecznych treści w każdym wymiarów. Dowiedz się więcej o bezpieczeństwie ustawieniach.

Wpisz wątpliwy prompt i uruchom model z domyślnymi ustawieniami bezpieczeństwa. i nie zwróci żadnych kandydatów:

response = model.generate_content('[Questionable prompt here]')
response.candidates

[
  content {
    parts {
      text: "I\'m sorry, but this prompt involves a sensitive topic and I\'m not allowed to generate responses that are potentially harmful or inappropriate."
    }
    role: "model"
  }
  finish_reason: STOP
  index: 0
  safety_ratings {
    category: HARM_CATEGORY_SEXUALLY_EXPLICIT
    probability: NEGLIGIBLE
  }
  safety_ratings {
    category: HARM_CATEGORY_HATE_SPEECH
    probability: NEGLIGIBLE
  }
  safety_ratings {
    category: HARM_CATEGORY_HARASSMENT
    probability: NEGLIGIBLE
  }
  safety_ratings {
    category: HARM_CATEGORY_DANGEROUS_CONTENT
    probability: NEGLIGIBLE
  }
]

prompt_feedback poinformuje Cię, który filtr bezpieczeństwa zablokował prośbę:

response.prompt_feedback

safety_ratings {
  category: HARM_CATEGORY_SEXUALLY_EXPLICIT
  probability: NEGLIGIBLE
}
safety_ratings {
  category: HARM_CATEGORY_HATE_SPEECH
  probability: NEGLIGIBLE
}
safety_ratings {
  category: HARM_CATEGORY_HARASSMENT
  probability: NEGLIGIBLE
}
safety_ratings {
  category: HARM_CATEGORY_DANGEROUS_CONTENT
  probability: NEGLIGIBLE
}

Teraz przekaż ten sam prompt do modelu z nowo skonfigurowanymi ustawieniami zabezpieczeń, i możesz otrzymać odpowiedź.

response = model.generate_content('[Questionable prompt here]',
                                  safety_settings={'HARASSMENT':'block_none'})
response.text

Pamiętaj też, że każdy kandydat ma własne uprawnienie safety_ratings, na wypadek, gdyby prompt – wszystkie odpowiedzi nie przejdą weryfikacji bezpieczeństwa.

Kodowanie wiadomości

Poprzednie sekcje korzystały z pakietu SDK, aby ułatwić Ci wysyłanie promptów do interfejsu API. Ta sekcja zawiera w pełni napisany odpowiednik poprzedniej sekcji Dzięki temu lepiej zrozumiesz szczegóły niższego poziomu Pakiet SDK koduje komunikaty.

SDK próbuje przekonwertować Twoją wiadomość na obiekt genai.protos.Content, zawiera listę genai.protos.Part obiektów, z których każdy zawiera jeden z tych elementów:

  1. text (ciąg znaków)
  2. inline_data (genai.protos.Blob), gdzie obiekt blob zawiera binarne data i mime_type.
  3. i innych typów danych.

Można również przekazać dowolną z tych klas jako równoważny słownik.

Tak więc pełny odpowiednik poprzedniego przykładu to:

model = genai.GenerativeModel('gemini-1.5-flash')
response = model.generate_content(
    genai.protos.Content(
        parts = [
            genai.protos.Part(text="Write a short, engaging blog post based on this picture."),
            genai.protos.Part(
                inline_data=genai.protos.Blob(
                    mime_type='image/jpeg',
                    data=pathlib.Path('image.jpg').read_bytes()
                )
            ),
        ],
    ),
    stream=True)

response.resolve()

to_markdown(response.text[:100] + "... [TRIMMED] ...")

Meal prepping is a great way to save time and money, and it can also help you to eat healthier. By ... [TRIMMED] ...

Rozmowy wieloetapowe

Pokazana wcześniej klasa genai.ChatSession może obsługiwać wiele przypadków użycia, przyjmuje pewne założenia. Jeśli Twój przypadek użycia nie pasuje do tego czatu implementacji, warto pamiętać, że genai.ChatSession to tylko kod w pobliżu GenerativeModel.generate_content Może też obsługiwać rozmowy wieloetapowe, a nie tylko pojedyncze żądania.

Pojedyncze wiadomości są genai.protos.Content obiektami lub są zgodne słowniki, tak jak w poprzednich sekcjach. Wiadomość w postaci słownika wymaga kluczy role i parts. Elementem role w rozmowie może być user (wyświetla prompty) lub model (wyświetla odpowiedzi).

Po przesłaniu listy genai.protos.Content obiektu będzie on traktowany jako czat wieloetapowy:

model = genai.GenerativeModel('gemini-1.5-flash')

messages = [
    {'role':'user',
     'parts': ["Briefly explain how a computer works to a young child."]}
]
response = model.generate_content(messages)

to_markdown(response.text)

Imagine a computer as a really smart friend who can help you with many things. Just like you have a brain to think and learn, a computer has a brain too, called a processor. It's like the boss of the computer, telling it what to do.

Inside the computer, there's a special place called memory, which is like a big storage box. It remembers all the things you tell it to do, like opening games or playing videos.

When you press buttons on the keyboard or click things on the screen with the mouse, you're sending messages to the computer. These messages travel through special wires, called cables, to the processor.

The processor reads the messages and tells the computer what to do. It can open programs, show you pictures, or even play music for you.

All the things you see on the screen are created by the graphics card, which is like a magic artist inside the computer. It takes the processor's instructions and turns them into colorful pictures and videos.

To save your favorite games, videos, or pictures, the computer uses a special storage space called a hard drive. It's like a giant library where the computer can keep all your precious things safe.

And when you want to connect to the internet to play games with friends or watch funny videos, the computer uses something called a network card to send and receive messages through the internet cables or Wi-Fi signals.

So, just like your brain helps you learn and play, the computer's processor, memory, graphics card, hard drive, and network card all work together to make your computer a super-smart friend that can help you do amazing things!

Aby kontynuować rozmowę, dodaj odpowiedź i kolejną wiadomość.

messages.append({'role':'model',
                 'parts':[response.text]})

messages.append({'role':'user',
                 'parts':["Okay, how about a more detailed explanation to a high school student?"]})

response = model.generate_content(messages)

to_markdown(response.text)

At its core, a computer is a machine that can be programmed to carry out a set of instructions. It consists of several essential components that work together to process, store, and display information:

**1. Processor (CPU):**
   -   The brain of the computer.
   -   Executes instructions and performs calculations.
   -   Speed measured in gigahertz (GHz).
   -   More GHz generally means faster processing.

**2. Memory (RAM):**
   -   Temporary storage for data being processed.
   -   Holds instructions and data while the program is running.
   -   Measured in gigabytes (GB).
   -   More GB of RAM allows for more programs to run simultaneously.

**3. Storage (HDD/SSD):**
   -   Permanent storage for data.
   -   Stores operating system, programs, and user files.
   -   Measured in gigabytes (GB) or terabytes (TB).
   -   Hard disk drives (HDDs) are traditional, slower, and cheaper.
   -   Solid-state drives (SSDs) are newer, faster, and more expensive.

**4. Graphics Card (GPU):**
   -   Processes and displays images.
   -   Essential for gaming, video editing, and other graphics-intensive tasks.
   -   Measured in video RAM (VRAM) and clock speed.

**5. Motherboard:**
   -   Connects all the components.
   -   Provides power and communication pathways.

**6. Input/Output (I/O) Devices:**
   -   Allow the user to interact with the computer.
   -   Examples: keyboard, mouse, monitor, printer.

**7. Operating System (OS):**
   -   Software that manages the computer's resources.
   -   Provides a user interface and basic functionality.
   -   Examples: Windows, macOS, Linux.

When you run a program on your computer, the following happens:

1.  The program instructions are loaded from storage into memory.
2.  The processor reads the instructions from memory and executes them one by one.
3.  If the instruction involves calculations, the processor performs them using its arithmetic logic unit (ALU).
4.  If the instruction involves data, the processor reads or writes to memory.
5.  The results of the calculations or data manipulation are stored in memory.
6.  If the program needs to display something on the screen, it sends the necessary data to the graphics card.
7.  The graphics card processes the data and sends it to the monitor, which displays it.

This process continues until the program has completed its task or the user terminates it.

Konfiguracja generowania

Argument generation_config umożliwia zmianę parametrów generowania. Każdy prompt wysyłany do modelu zawiera wartości parametrów, które określają, model generuje odpowiedzi.

model = genai.GenerativeModel('gemini-1.5-flash')
response = model.generate_content(
    'Tell me a story about a magic backpack.',
    generation_config=genai.types.GenerationConfig(
        # Only one candidate for now.
        candidate_count=1,
        stop_sequences=['x'],
        max_output_tokens=20,
        temperature=1.0)
)

text = response.text

if response.candidates[0].finish_reason.name == "MAX_TOKENS":
    text += '...'

to_markdown(text)

Once upon a time, in a small town nestled amidst lush green hills, lived a young girl named...

Co dalej?

  • Projektowanie promptów to proces tworzenia promptów, które wywołują pożądane odpowiedzi modelowych. Pisanie dobrze uporządkowanych promptów to to kluczowy element zapewnienia dokładnych i wysokiej jakości odpowiedzi udzielanych w danym języku model atrybucji. Poznaj sprawdzone metody dotyczące próśb pisania.
  • Gemini oferuje kilka wersji modelu, które można dopasować do różnych zastosowań takie jak typy danych wejściowych i złożoność, wdrożenia czatu zadania związane z językiem okien dialogowych i ograniczeniami rozmiaru. Dowiedz się więcej o dostępnych Modele Gemini.