├── README.md
└── using_structured_output_and_json_strict_mode_openai.py


/README.md:
--------------------------------------------------------------------------------
 1 | ## Learn More About Using OpenAI's Structured Outputs
 2 | 
 3 | For a detailed explanation of the techniques used in this project, check out our blog post:
 4 | 
 5 | [Using Structured Output and JSON Strict Mode with OpenAI](https://www.firecrawl.dev/blog/using-structured-output-and-json-strict-mode-openai)
 6 | 
 7 | This article provides in-depth insights into:
 8 | 
 9 | - The benefits of using structured outputs
10 | - How to implement JSON Strict Mode with OpenAI's models
11 | - Best practices for combining web scraping with AI-powered data extraction
12 | 
13 | Reading this blog post will give you a deeper understanding of the concepts behind this project and how to make the most of these powerful tools.
14 | 


--------------------------------------------------------------------------------
/using_structured_output_and_json_strict_mode_openai.py:
--------------------------------------------------------------------------------
 1 | # pip install firecrawl openai
 2 | # set FIRECRAWL_API_KEY and OPENAI_API_KEY environment variables
 3 | 
 4 | 
 5 | from firecrawl import FirecrawlApp
 6 | from openai import OpenAI
 7 | import os
 8 | 
 9 | # Initialize the FirecrawlApp with your API key
10 | firecrawl_app = FirecrawlApp(api_key=os.environ['FIRECRAWL_API_KEY'])
11 | 
12 | # Scrape data from mendable.ai
13 | url = 'https://mendable.ai'
14 | scraped_data = firecrawl_app.scrape_url(url)
15 | 
16 | # Initialize OpenAI client
17 | client = OpenAI(
18 |     api_key=os.environ['OPENAI_API_KEY']
19 | )
20 | 
21 | # Define the OpenAI API request
22 | messages = [
23 |     {
24 |         "role": "system",
25 |         "content": "You are a helpful assistant that extracts structured data from web pages."
26 |     },
27 |     {
28 |         "role": "user",
29 |         "content": f"Extract the headline and description from the following HTML content: {scraped_data['content']}"
30 |     }
31 | ]
32 | 
33 | response_format = {
34 |     "type": "json_schema",
35 |     "json_schema": {
36 |         "name": "extracted_data",
37 |         "schema": {
38 |             "type": "object",
39 |             "properties": {
40 |                 "headline": {
41 |                     "type": "string"
42 |                 },
43 |                 "description": {
44 |                     "type": "string"
45 |                 }
46 |             },
47 |             "required": ["headline", "description"],
48 |             "additionalProperties": False
49 |         }
50 |     }
51 | }
52 | 
53 | # Call the OpenAI API to extract structured data
54 | chat_completion = client.chat.completions.create(
55 |     model="gpt-4o-mini-2024-07-18",
56 |     messages=messages,
57 |     response_format=response_format
58 | )
59 | 
60 | # Extracted data
61 | # Access the content of the first choice in the response
62 | extracted_data = chat_completion.choices[0].message.content
63 | 
64 | # Print the extracted data
65 | print(extracted_data)


--------------------------------------------------------------------------------