Google Colab¶
A Colab notebook is quick to share and quick to fill with real data. Call models from it through the gateway and every call is checked against your policies, charged to the right team and recorded in Request logs, with no provider key in the notebook.
What it is¶
Colab runs ordinary Python in Google's cloud, so you call the gateway with the SDK you already use: the OpenAI SDK, the Anthropic SDK or the Google GenAI SDK. The key lives in Colab's Secrets (the key icon in the left bar), never in a cell.
Colab's built-in AI assistant (Gemini in Colab) uses Google's own models and doesn't go through the gateway. Only the calls your code makes do.
The two settings¶
- In Secrets, add a secret named
VIDAI_API_KEYwith your Vidai key (startsvk_) as its value, and switch on Notebook access for this notebook. - Read it in a cell and give the SDK the gateway's address:
| SDK | Base address | Key |
|---|---|---|
OpenAI (openai) |
https://<your-gateway>/v1 |
api_key=userdata.get("VIDAI_API_KEY") |
Anthropic (anthropic) |
https://<your-gateway> |
api_key=userdata.get("VIDAI_API_KEY") |
Google GenAI (google-genai) |
https://<your-gateway> in http_options |
api_key=userdata.get("VIDAI_API_KEY") |
The gateway's address, and the base URL for each kind of client, are under Connect your app on the Keys page and in System › Settings › Gateway. You get the key from Keys › Issue a key; its secret is shown once. Use a provider's key nowhere.
A minimal call¶
from google.colab import userdata
from openai import OpenAI
client = OpenAI(base_url="https://<your-gateway>/v1", api_key=userdata.get("VIDAI_API_KEY"))
reply = client.chat.completions.create(
model="gpt-4o-mini",
messages=[{"role": "user", "content": "Reply with exactly: ok"}],
)
print(reply.choices[0].message.content)
!pip install -q anthropic
from google.colab import userdata
import anthropic
client = anthropic.Anthropic(base_url="https://<your-gateway>", api_key=userdata.get("VIDAI_API_KEY"))
reply = client.messages.create(
model="claude-sonnet-4.5",
max_tokens=50,
messages=[{"role": "user", "content": "Reply with exactly: ok"}],
)
print(reply.content[0].text)
from google.colab import userdata
from google import genai
from google.genai import types
client = genai.Client(
api_key=userdata.get("VIDAI_API_KEY"),
http_options=types.HttpOptions(base_url="https://<your-gateway>"),
)
reply = client.models.generate_content(model="gemini-3.5-flash", contents="Reply with exactly: ok")
print(reply.text)
Any of the three reaches any provider: the OpenAI SDK can call claude-sonnet-4.5, the Anthropic SDK gpt-4o-mini, the
Google GenAI SDK either. The gateway translates the request and the answer, so you keep the SDK you know. Install a
package Colab doesn't have with !pip install -q ….
Streaming¶
Streaming works as in a script, and prints as it arrives in the cell's output. See OpenAI SDK for Python, Anthropic SDK for Python and Google GenAI SDK for Python.
Tool calls¶
Tools work on every provider the gateway translates to, as on each SDK's page. In Request logs, the Tool filter under More filters finds the calls that offered or called a tool by its name.
Reading what the gateway did¶
raw = client.chat.completions.with_raw_response.create(
model="gpt-4o-mini", messages=[{"role": "user", "content": "Hi"}]
)
print(raw.headers.get("x-vidai-request-id")) # paste into Request logs › Request reference
print(raw.headers.get("x-vidai-model")) # the model that answered
Cost is never in the answer: find the call in Request logs.
Errors¶
The SDKs raise their usual exceptions; match on the x-vidai-error-code header. A notebook's typical ones:
| Status | Gateway code | What it means |
|---|---|---|
| 401 | invalid_api_key, missing_api_key |
The secret is missing, misspelt, or Notebook access is off |
| 403 | content_guardrail_blocked |
A guardrail stopped the content, for example personal data from a dataset |
| 403 | model_not_allowed |
Your key may not call that model |
| 429 | rate_limit_exceeded, token_limit_exceeded |
A loop over many rows went too fast; slow it down |
The full list is on the OpenAI SDK for Python page.
Model names¶
Use the names your organisation has approved, as listed under Models in the Control room.
client.models.list() (OpenAI SDK) returns exactly the models your key may call.
What goes through the gateway¶
| In Colab | Through the gateway |
|---|---|
| Your code calling an SDK pointed at the gateway | Yes, to every provider your organisation has set up |
| Gemini in Colab (the built-in assistant, code completion) | No: it uses Google's own models |
| Libraries built on these SDKs (LangChain and others) | Yes, when you give them the same address and key |
Check it works in a minute¶
Run the OpenAI cell above. ok means the address and key are right. The call shows in Request logs within
seconds.
Good to know¶
- Colab runs in Google's cloud: the gateway must be reachable from the internet. A gateway only on your private network can't be reached from Colab.
- Secrets belong to your Google account. A notebook you share doesn't share the secret: each person adds their own key, so each person's calls show under their name.
- Never print the key or leave it in a cell's output; a shared notebook keeps its outputs.
- A loop over a dataframe can send thousands of calls; your key's rate and token limits apply, and the gateway's own
429 has no
Retry-Afterheader, so pause between batches.