Externe APIs aufrufen

Entwicklung von KI-Systemen mit der OpenAI-API

Francesca Donadoni

Curriculum Manager, DataCamp

Python-Requests-Bibliothek für APIs

Ein Diagramm mit einer Anfrage an die Art Institute of Chicago API

Python-Requests-Bibliothek für APIs

Bibliothek importieren
```
import requests
```

URL angeben

url = "https://api.artic.edu/api/v1/artworks/search"

Query-Parameter angeben
```
querystring = {"q":keyword}
```

Antwort abrufen

response = requests.request("GET", url, params=querystring)

¹ https://api.artic.edu/docs/#quick-start

Funktion kapseln

import requests

def get_artwork(keyword):
    url = "https://api.artic.edu/api/v1/artworks/search"
    querystring = {"q":keyword}
    response = requests.request("GET", url, params=querystring)
    return response.text

Kontext hinzufügen

response = client.chat.completions.create(
    model="gpt-4o-mini",

    messages=[
{"role": "system",
 "content": "You are an AI assistant, a specialist in history 
    of art. You should interpret the user prompt, and based on it extract one 
    keyword for recommending artwork related to their preference."},

{"role": "user", 
 "content": "I don't have much time to visit the museum and would 
    like some recommendations. I like the seaside and quiet places."}
],

Funktion zu Tools hinzufügen

    function_definition=[{"type": "function",
            "function" : {
                "name": "get_artwork",
                "description": "This function calls the Art Institute of Chicago API 
                                to find artwork that matches a keyword",
                "parameters": {

                    "type": "object",
                    "properties": {
                        "artwork keyword": {
                            "type": "string",
                            "description": "The keyword to be passed to the 
                                            get_artwork function."}}},
                "result": {"type": "string"}
            }} ] )

Alles zusammenführen

import json

if response.choices[0].finish_reason=='tool_calls':
    function_call = response.choices[0].message.tool_calls[0].function

    if function_call.name == "get_artwork":
        artwork_keyword = json.loads(function_call.arguments)["artwork keyword"]
        artwork = get_artwork(artwork_keyword)

        if artwork: 
            print(f"Here are some recommendations: 
                    {[i['title'] for i in json.loads(artwork)['data']]}")
        else:
            print("Apologies, I couldn't make any recommendations based on the request.")

    else:
        print("Apologies, I couldn't find any artwork.")

else:
    print("I am sorry, but I could not understand your request.")

Abschließende Antwort

Here are some recommendations: ['Seaside, Port of Honfleur', 
'Little Landscape at the Seaside (Kleine Landschaft am Meer)', 
'Museum of Contemporary Art, Niterói, Rio de Janeiro, Brazil, Four Sketches',
'A seaside outing', 'Stacks of Wheat (End of Day, Autumn)', 'Stack of Wheat', 
'Stacks of Wheat (Sunset, Snow Effect)', 'Stacks of Wheat (End of Summer)', 
'Stack of Wheat (Thaw, Sunset)', 'Waitress at a Seaside Teahouse']

Ein impressionistisches Gemälde vom Meer mit Boot

¹ Seaside, Port of Honfleur | The Art Institute of Chicago

Lass uns üben!

Entwicklung von KI-Systemen mit der OpenAI-API