Atlas OpenAI API Integration

This guide will help you configure Atlas to authenticate and proxy requests to the OpenAI API.

OpenAI is an artificial intelligence (AI) research laboratory. It provides a variety of AI services, products, and APIs, including and especially related to text and image synthesis.

The OpenAI API provides developers with programmatic access to much of this functionality.

At the end of this guide, your running instance of Atlas will be configured to:

• Proxy HTTP requests to the OpenAI API.

• Authenticate these requests using one or more OpenAI API tokens.

Availability

Public beta. This integration is available to all Atlas users, but the API may change.

Prerequisites

• A running instance of Atlas. See installation guides for more details.

• A user account with OpenAI, or admin permissions in an OpenAI Organization.

Provision a OpenAI API token

1. Click your profile picture in the upper right-hand side of the OpenAI UI. Choose the View API Keys button in the dropdown menu.

   

   settings-button
2. Click the Create new secret key button.

   

   api-tokens-tab
3. Input a description for the API key, and click the Create secret key button.

   

   api-tokens-tab
4. Copy the API key to your clipboard and save it.

   

   api-tokens-tab

[Optional] Get ID of OpenAI Organization you'd like to use, if you have more than one

If you have more than one OpenAI Organization, you will need to specify which one to use. The OpenAI Organization ID is in the Organization settings page.

org-settings

Add OpenAI Integration to Atlas

Step 1: Add OpenAI API token to Atlas Deployment as an Environment Variable

• Choose an environment variable name for the OpenAI API token. Generally this is something like OPENAI_API_TOKEN.

• Add the OpenAI API token you provisioned as an environment variables to your Atlas deployment. The install guides have instructions for how to do this for each deployment method. For example, if you deployed Atlas using ECS, you might add an environment variable OPENAI_API_TOKEN to the Pulumi configuration. If you deployed using Kubernetes, you might add the OPENAI_API_TOKEN environment variable to a .env file.

• Note the name of the environment variable you chose. We will use this in the next step to configure the HTTP adapter.

[Optional] Step 2: Add OpenAI Organization ID to Atlas Deployment as an Environment Variable

If you have more than one OpenAI Organization, you will need to specify which one to use by adding the OpenAI Organization ID as an environment variable to your Atlas deployment.

• Choose an environment variable name for the OpenAI API Organization. Generally this is something like OPENAI_ORG_ID.

• Add the OpenAI API token you provisioned as an environment variables to your Atlas deployment. The install guides have instructions for how to do this for each deployment method. For example, if you deployed Atlas using ECS, you might add an environment variable OPENAI_ORG_ID to the Pulumi configuration. If you deployed using Kubernetes, you might add the OPENAI_ORG_ID environment variable to a .env file.

• Note the name of the environment variable you chose. We will use this in the next step to configure the HTTP adapter.

Step 3: Add OpenAI API token to Atlas Configuration

We can use the mom CLI to add the OpenAI API token to the Atlas configuration. Run this command, changing

• YOUR_ATLAS_CONFIG.yml with the path to your Atlas configuration file

• OPENAI_API_TOKEN to the name of the environment variable you chose in the previous step

• YOUR_ADAPTER_NAME to the name you want to use for the HTTP adapter in Atlas, e.g., openai

mom atlas config add-http-adapter \
    -f YOUR_ATLAS_CONFIG.yml \
    --adapter-name YOUR_ADAPTER_NAME \
    --base-url https://api.openai.com \
    -H 'Authorization: Bearer ${{ OPENAI_API_TOKEN }}'

If you used the OPENAI_ORG_ID environment variable, you can add it to the command above with the -H flag:

mom atlas config add-http-adapter \
    -f YOUR_ATLAS_CONFIG.yml \
    --adapter-name YOUR_ADAPTER_NAME \
    --base-url https://api.openai.com \
    -H 'Authorization: Bearer ${{ OPENAI_API_TOKEN }}' \
    -H 'OpenAI-Organization: ${{ OPENAI_ORG_ID }}'

The diff in your version control system should look something like this:

diff --git a/YOUR_ATLAS_CONFIG.yml b/YOUR_ATLAS_CONFIG.yml
index 1434ece..204ebde 100644
--- a/YOUR_ATLAS_CONFIG.yml
+++ b/YOUR_ATLAS_CONFIG.yml
@@ -12,6 +12,10 @@ spec:
       apiVersion: moment.dev/adapters/v1alpha1
       kind: AWS
       name: aws
+  - adapterRef:
+      apiVersion: moment.dev/adapters/v1alpha1
+      kind: HTTP
+      name: YOUR_ADAPTER_NAME
   exposedPorts: {}
   gatewayRegistration:
     backoff:
@@ -50,3 +54,13 @@ spec:
   headers:
   - name: Authorization
     value: '"token ${{ GITHUB_TOKEN }}"'
+---
+apiVersion: moment.dev/adapters/v1alpha1
+kind: HTTP
+metadata:
+  name: YOUR_ADAPTER_NAME
+spec:
+  baseUrl: https://api.openai.com
+  headers:
+  - name: Authorization
+    value: Bearer ${{ OPENAI_API_TOKEN }}

Step 4: Deploy the Updated Atlas Config

The install guides have instructions for how to deploy Atlas into a variety of environments, including Kubernetes and ECS.

Step 5: Test the Integration

Once deployed, we can use the mom curl command to test the integration. Be sure to replace openai with the name you chose in the previous step if it is different.

mom curl /v1/apis/http/openai/v1/models

Using the integration in a canvas

This integration can be used in Moment by creating a new cell in a Moment canvas, and pasting the following code. Note that you will need to assign httpAdapterName to the name you chose for the HTTP adapter in the previous step, e.g., openai

const httpAdapterName = "openai";
const response = await atlasProxyFetch(`/v1/apis/http/${httpAdapterName}/v1/models`);
return await response.json();

If the integration is working, you should see a JSON object with a list of OpenAI users.

The API will likely time out, so try bumping the Envoy timeout (connect_timeout in the envoy config).