Files & vector stores

About

Prism proxies the OpenAI Files and Vector Stores APIs. Upload documents for assistant file search, fine-tuning data, or batch processing. Vector stores index uploaded files for semantic retrieval during assistant runs.

Like the Assistants API, files and vector stores are stored on OpenAI’s servers. Use the OpenAI SDK pointed at Prism.

Files

Endpoints

Upload a file

from openai import OpenAI

client = OpenAI(
    base_url="https://gateway.futureagi.com/v1",
    api_key="sk-prism-your-key",
)

# Upload for use with Assistants
file = client.files.create(
    file=open("report.pdf", "rb"),
    purpose="assistants",
)
print(f"File ID: {file.id}")
print(f"Size: {file.bytes} bytes")

curl -X POST https://gateway.futureagi.com/v1/files \
  -H "Authorization: Bearer sk-prism-your-key" \
  -F file=@report.pdf \
  -F purpose=assistants

Purpose values

List and manage files

# List all files
files = client.files.list()
for f in files.data:
    print(f"{f.id}: {f.filename} ({f.bytes} bytes, purpose={f.purpose})")

# Get file metadata
file = client.files.retrieve("file-abc123")

# Download file content
content = client.files.content("file-abc123")
with open("downloaded.pdf", "wb") as f:
    f.write(content.read())

# Delete a file
client.files.delete("file-abc123")

Vector stores

Vector stores index uploaded files for semantic search. They’re used with the Assistants API file_search tool.

Endpoints

Create a vector store and add files

# Create a vector store
vector_store = client.beta.vector_stores.create(
    name="Product Documentation",
)
print(f"Vector store: {vector_store.id}")

# Upload and add a file
file = client.files.create(
    file=open("docs.pdf", "rb"),
    purpose="assistants",
)

client.beta.vector_stores.files.create(
    vector_store_id=vector_store.id,
    file_id=file.id,
)

Batch upload

Add multiple files at once:

# Upload several files
file_ids = []
for path in ["chapter1.pdf", "chapter2.pdf", "chapter3.pdf"]:
    f = client.files.create(file=open(path, "rb"), purpose="assistants")
    file_ids.append(f.id)

# Batch add to vector store
batch = client.beta.vector_stores.file_batches.create(
    vector_store_id=vector_store.id,
    file_ids=file_ids,
)
print(f"Batch status: {batch.status}")

Search a vector store

Search indexed files directly (outside of an assistant run):

results = client.beta.vector_stores.search(
    vector_store_id=vector_store.id,
    query="return policy",
)

for result in results.data:
    print(f"Score: {result.score:.4f}")
    print(f"Content: {result.content[0].text[:200]}")
    print()

Use with an assistant

Attach a vector store to an assistant for automatic file search during runs:

assistant = client.beta.assistants.create(
    name="Support Agent",
    instructions="Answer questions using the product documentation.",
    model="gpt-4o",
    tools=[{"type": "file_search"}],
    tool_resources={
        "file_search": {
            "vector_store_ids": [vector_store.id],
        }
    },
)

See Assistants API for the full assistant workflow.

Manage vector stores

# List vector stores
stores = client.beta.vector_stores.list()
for vs in stores.data:
    print(f"{vs.id}: {vs.name} ({vs.file_counts.completed} files)")

# List files in a vector store
files = client.beta.vector_stores.files.list(vector_store_id=vector_store.id)

# Remove a file from a vector store
client.beta.vector_stores.files.delete(
    vector_store_id=vector_store.id,
    file_id="file-abc123",
)

# Delete a vector store
client.beta.vector_stores.delete(vector_store.id)

Supported file types

Max file size: 512 MB. Max files per vector store: 10,000.

Next Steps