Extract Structured Data from Unstructured Text

1. 1
   Set Up Your Environment

   To perform AI-powered Data Extraction, you need libraries to interact with a large language model (LLM) and to define data schemas. We'll use `openai` for the LLM and `instructor` to apply the schema. Install them via pip.
2. 2
   Define Your Target Data Schema

   The core of schema-based extraction is defining the structure you want to pull from the text. Create a Pydantic model that represents your desired data. This schema guides the AI, ensuring the output is typed and validated.
3. 3
   Prepare the AI Client and Input Text

   Patch the OpenAI client using `instructor` to enable schema-based responses. Then, define the unstructured text you want to parse. Make sure your OpenAI API key is set as an environment variable (`OPENAI_API_KEY`).
4. 4
   Execute the Data Extraction

   Call the chat completions API using the `response_model` parameter to pass your `Invoice` schema. `instructor` handles the complex prompting and validation to ensure the LLM's output perfectly conforms to your model.

import os
import pydantic
import openai
import instructor

# 1. Ensure your OPENAI_API_KEY is set in your environment
# For example: export OPENAI_API_KEY='sk-...'
if not os.getenv("OPENAI_API_KEY"):
    raise ValueError("OPENAI_API_KEY environment variable not set!")

# 2. Define the schema for the Data Extraction skill
# This tells the AI exactly what to look for and in what format.
class Invoice(pydantic.BaseModel):
    invoice_id: str
    amount: float
    due_date: str = pydantic.Field(description="The due date in YYYY-MM-DD format")

# 3. Patch the OpenAI client to add the extraction capability
client = instructor.patch(openai.OpenAI())

# 4. Define the unstructured text to process
text_block = "Hi, please find attached invoice INV-123 for $99.99. The payment is due on Feb 25, 2024."

# 5. Execute the extraction by providing the response_model
try:
    invoice_details = client.chat.completions.create(
        model="gpt-4o",
        response_model=Invoice,
        messages=[
            {"role": "user", "content": f"Extract the invoice details from the following text: {text_block}"}
        ]
    )

    print("--- Extracted Data ---")
    print(invoice_details.model_dump_json(indent=2))

except Exception as e:
    print(f"An error occurred: {e}")

# Expected output:
# -- Extracted Data ---
# {
#   "invoice_id": "INV-123",
#   "amount": 99.99,
#   "due_date": "2024-02-25"
# }