Getting Started

The Vector Inference Platform is available to all Vector Institute faculty & researchers. For an up-to-date list of available models and their specifications, visit inference.vectorinstitute.ai.

Getting an API Key

Access is managed via invite links. To get started:

1. Contact the AI Engineering team to request access. The request must come from Vector faculty sent to inference@vectorinstitute.ai. Please send a list of email IDs for whom access is required (i.e. the faculty themselves or students from lab of Vector faculty).
2. The emails will receive an invite link.
3. Click the link in the invite email. It is valid for 48 hours and takes you to the Vector AI Gateway dashboard.
4. Your API key is generated and displayed once: copy and save it immediately. The key will not be shown again.
5. You can look at your usage across models at https://proxy.vectorinstitute.ai/dashboard/usage.

API keys have the format vp_xxxxxxxx.yyyyyy....

Prerequisites

Install the OpenAI Python client:

pip install openai

Usage

The platform exposes an OpenAI-compatible API at https://proxy.vectorinstitute.ai/v1. Use it as a drop-in replacement for any OpenAI client by changing the base_url and model parameters.

from openai import OpenAI

client = OpenAI(
    base_url="https://proxy.vectorinstitute.ai/v1",
    api_key="vp_xxxxxxxx.yyyyyy..."
)

stream = client.chat.completions.create(
    model="<model-id>",  # see inference.vectorinstitute.ai for available models
    messages=[{"role": "user", "content": "Explain attention mechanisms in transformers."}],
    stream=True,
)

for chunk in stream:
    if chunk.choices:
        print(chunk.choices[0].delta.content or "", end="", flush=True)

You can also use curl:

curl https://proxy.vectorinstitute.ai/v1/chat/completions \
  -H "Authorization: Bearer vp_xxxxxxxx.yyyyyy..." \
  -H "Content-Type: application/json" \
  -d '{
    "model": "<model-id>",
    "messages": [{"role": "user", "content": "Hello!"}]
  }'

Listing Available Models

Retrieve the current list of enabled models programmatically:

curl https://proxy.vectorinstitute.ai/v1/models \
  -H "Authorization: Bearer vp_xxxxxxxx.yyyyyy..."

Or visit inference.vectorinstitute.ai for a visual overview.