Arize AI Provider Integration

Arize AI provider integrations

Use the ax ai-integrations CLI to manage account-scoped LLM provider credentials in Arize, then return the integration ID, provider, scoping, credential status, and any downstream evaluator impact.

When to invoke

• "Create an Arize AI integration for OpenAI, Anthropic, Azure OpenAI, Bedrock, Vertex AI, Gemini, NVIDIA NIM, or a custom provider."
• "List or inspect Arize AI integrations in a space."
• "Rotate credentials or update model names for an Arize integration."
• "Delete an Arize LLM provider integration."
• "Find the integration ID to use with an evaluator."

Prerequisites and context

Proceed directly with the needed ax command; do not check versions, env vars, or profiles upfront. If a command fails, troubleshoot from the error.

• Most --space flags and the ARIZE_SPACE env var accept a space name such as my-workspace or a base64 space ID such as U3BhY2U6...; find spaces with ax spaces list.
• ai-integrations create does not accept --space because AI integrations are account-scoped.
• Use --space with list, get, update, and delete when resolving by name or filtering visibility.
• Never read .env files or search the filesystem for credentials. Use ax profiles for Arize credentials and ax ai-integrations for LLM provider keys.
• If credentials are not available through these channels, ask the user.

Concepts and fields

List and inspect integrations

List all integrations accessible in a space:

ax ai-integrations list --space SPACE

Filter by name, using a case-insensitive substring match:

ax ai-integrations list --space SPACE --name "openai"

Paginate large result sets:

ax ai-integrations list --space SPACE --limit 20 -o json
ax ai-integrations list --space SPACE --limit 20 --cursor CURSOR_TOKEN -o json

Get a specific integration:

ax ai-integrations get NAME_OR_ID
ax ai-integrations get NAME_OR_ID -o json
ax ai-integrations get NAME_OR_ID --space SPACE

Use --space SPACE when using a name instead of an ID.

Create integrations

Before creating, list integrations first because the user may already have a suitable one:

ax ai-integrations list --space SPACE

If none exists, create with provider-specific flags.

After creation, capture the integration ID such as TGxtSW50ZWdyYXRpb246MTI6YUJjRA==. If you missed it, retrieve it with ax ai-integrations list --space SPACE -o json or ax ai-integrations get NAME_OR_ID.

Update and delete integrations

update is a partial update; omitted fields stay as-is. Add --space SPACE when using a name instead of ID. Any flag accepted by create can be passed to update.

ax ai-integrations update NAME_OR_ID --name "New Name"
ax ai-integrations update NAME_OR_ID --api-key $OPENAI_API_KEY
ax ai-integrations update NAME_OR_ID --model-name gpt-4o --model-name gpt-4o-mini
ax ai-integrations update NAME_OR_ID --base-url "https://new-endpoint.example.com/v1"

Deletion is permanent. Evaluators that reference the integration will no longer run.

ax ai-integrations delete NAME_OR_ID --force
ax ai-integrations delete NAME_OR_ID --space SPACE --force

Omit --force to get a confirmation prompt instead of deleting immediately.

Troubleshooting

Progressive disclosure and bundled resources

• references/ax-setup.md: install or fix the ax CLI when commands are unavailable or incompatible.
• references/ax-profiles.md: configure Arize profiles, ARIZE_API_KEY, and ~/.arize/config.toml when authentication fails.

Related primitives

Arize command vocabulary

Preserve exact placeholder forms when adapting commands: --api-key <key>, --base-url <azure-endpoint>, --base-url <nim-endpoint>, --base-url <endpoint>, --provider-metadata '{"role_arn": "<arn>"}', azure-endpoint, nim-endpoint, gcp-project, role-based, name/ID, create/update, re-run, command not found, and No profile found.

Output template

### Arize AI integration result

**Status:** created | listed | updated | deleted | blocked
**Space:** `<SPACE or ARIZE_SPACE or not used>`
**Integration:** `<name or NAME_OR_ID>`
**Provider:** `openAI` | `anthropic` | `azureOpenAI` | `awsBedrock` | `vertexAI` | `gemini` | `nvidiaNim` | `custom`
**Integration ID:** `<id such as TGxtSW50ZWdyYXRpb246MTI6YUJjRA==>`

| Field | Value |
| --- | --- |
| `has_api_key` | `true` | `false` |
| `model_names` | `<list or null>` |
| `enable_default_models` | `<value>` |
| `function_calling_enabled` | `<value>` |
| `auth_type` | `default` | `proxy_with_headers` | `bearer_token` |

**Commands run**
- `<ax ai-integrations ...>`

**Notes**
- <downstream evaluator impact, credential rotation note, or blocker>

Quality gate

• ax ai-integrations list --space SPACE is run before creating a new integration unless no space context exists.
• create is not called with --space.
• --space SPACE is supplied for get, update, or delete when resolving by name.
• Provider-specific required flags are present.
• Credentials are never read from .env or discovered by filesystem search.
• The integration ID is captured or retrieved after creation.
• Deletion impact on evaluators is stated before using --force.
• Troubleshooting uses references/ax-setup.md or references/ax-profiles.md only after a relevant command failure.

References

• Arize API Keys
• Azure OpenAI endpoint example
• NVIDIA NIM endpoint example
• Custom OpenAI-compatible endpoint example
• Updated endpoint example