Skip to main content

Uploading Files to a Vector Store

In this lesson, you'll learn how to create a new vector store, upload a file into it, and then query it using GPT-4o with the file_search tool.

Prerequisites​

Make sure you have:

  • A PDF file ready to upload (e.g., Nexovious Project Info.pdf)
  • OpenAI SDK installed
pip install openai

The Code​

from openai import OpenAI

client = OpenAI()

file_path = "../files/Nexovious Project Info.pdf"

# 1. Create vector store
vector_store = client.vector_stores.create(
    name="Nexovious File Store",
    metadata={
        "description": "Vector store for Nexovious project information",
        "type": "product information"
    }
)
print(f"Created vector store: {vector_store.id}")

# 2. Upload file content directly to vector store for indexing
with open(file_path, "rb") as f:
    client.vector_stores.file_batches.upload_and_poll(
        vector_store_id=vector_store.id,
        files=[f]
    )

print("File uploaded and indexed in vector store.")

# 3. Query the vector store
response = client.responses.create(
    model="gpt-4o-mini",
    input="What services Nexovious provides?",
    tools=[{
        "type": "file_search",
        "vector_store_ids": [vector_store.id],
        "max_num_results": 1
    }]
)

# 4. Extract assistant response text
for item in response.output:
    if getattr(item, "type", None) == "message":
        for content in item.content:
            if getattr(content, "type", None) == "output_text":
                print(content.text)

Explanation​

  • client.vector_stores.create(...): Initializes a new vector store.
  • upload_and_poll(...): Uploads and waits until the file is fully indexed.
  • vector_store.id: Used for querying the indexed content.
  • response.output: Contains the AI's grounded response.

Use Case​

This setup is ideal for:

  • Uploading internal knowledge bases
  • Allowing GPT to answer questions from custom documents
  • Integrating document search into your chat agents