|
| 1 | +--- |
| 2 | +title: "Use ARIA for autoresearch" |
| 3 | +description: "Learn how to use ARIA, W&B's AI Research and Iteration Agent, to analyze results and run experiments." |
| 4 | +--- |
| 5 | + |
| 6 | +Use ARIA, W&B's AI Research and Iteration Agent, as your research assistant to analyze experiments, explain results, identify patterns across runs, recommend next steps, build visualizations, and more. |
| 7 | + |
| 8 | +Based on your previous runs, ARIA can recommend and run new experiments. |
| 9 | + |
| 10 | +ARIA uses W&B Launch to run experiments in a sandbox environment. Before ARIA can run experiments in your project, use [ARIA to help you set up W&B Launch](/aria/autoresearch#set-up-wb-launch-with-aria). |
| 11 | + |
| 12 | +## Set up W&B Launch with ARIA |
| 13 | + |
| 14 | +ARIA can help you configure W&B Launch from the chat. For example, if you ask ARIA: |
| 15 | + |
| 16 | +```text title="User prompt" |
| 17 | +Can you help me set up W&B Launch? |
| 18 | +``` |
| 19 | + |
| 20 | +ARIA checks which [W&B Launch components](/aria/autoresearch#launch-components) are available in your project and guides you through the remaining setup steps. |
| 21 | + |
| 22 | +The following example shows part of an ARIA response to the prompt, "Can you help me set up W&B Launch?": |
| 23 | + |
| 24 | + |
| 25 | +<Accordion title="ARIA response"> |
| 26 | + |
| 27 | +```text |
| 28 | +## 1. Decide where jobs will actually run |
| 29 | +
|
| 30 | +Pick one: |
| 31 | +
|
| 32 | +### Option A — Kubernetes / cluster setup |
| 33 | +Use this if you want shared GPU/CPU infrastructure and a |
| 34 | +persistent Launch agent. |
| 35 | +
|
| 36 | +Recommended for “real” team usage. |
| 37 | +
|
| 38 | +### Option B — Local Docker setup |
| 39 | +Use this if you want jobs to run on your own machine, |
| 40 | +e.g. local workstation with Docker and possibly a GPU. |
| 41 | +
|
| 42 | +Good for testing Launch quickly. |
| 43 | +
|
| 44 | +I’d usually start with **Kubernetes** if |
| 45 | +you have cluster access. If not, start with **Local Docker**. |
| 46 | +
|
| 47 | +## 2. Create a Launch queue in the W&B UI |
| 48 | +
|
| 49 | +Go to the Launch page for the entity: https://wandb.ai/wandb/launch |
| 50 | +
|
| 51 | +Then: |
| 52 | +
|
| 53 | +1. Click **Create Queue**. |
| 54 | +2. Select entity: `wandb`. |
| 55 | +3. Give the queue a name. |
| 56 | +4. Choose the queue resource type: |
| 57 | + - **Kubernetes** for cluster execution. |
| 58 | + - **Docker** for local execution. |
| 59 | +5. Add the queue resource configuration. |
| 60 | +6. Click **Create Queue**. |
| 61 | +
|
| 62 | +## 3. Kubernetes path: install a Launch agent with Helm |
| 63 | +
|
| 64 | +If using Kubernetes, you need: |
| 65 | +
|
| 66 | +- A Kubernetes cluster. |
| 67 | +- `kubectl` access to the cluster. |
| 68 | +- `helm` installed locally. |
| 69 | +- A W&B **service account API key**. |
| 70 | +- Permission to create/update/delete Kubernetes resources in the target namespace. |
| 71 | +
|
| 72 | +Do **not** use your personal API key for the agent if this is a real |
| 73 | +setup. Use a W&B service account key. |
| 74 | +``` |
| 75 | + |
| 76 | +</Accordion> |
| 77 | + |
| 78 | + |
| 79 | +### Launch components |
| 80 | + |
| 81 | +W&B Launch has three core components: |
| 82 | + |
| 83 | +* **Launch job**: A reusable blueprint for configuring and running a task in an ML workflow. |
| 84 | +* **Launch queue**: A first-in, first-out queue that submits jobs to a compute target, such as a Kubernetes cluster. |
| 85 | +* **Launch agent**: A process that polls a queue and runs jobs on the compute target configured for that queue. |
| 86 | + |
| 87 | +ARIA can manage the W&B-side Launch workflow after a queue and active Launch agent are available. ARIA usually cannot start a durable Launch agent on your local machine, Kubernetes cluster, SageMaker environment, or Vertex AI environment. A user or administrator must provide the compute environment, credentials, and running agent. |
| 88 | + |
| 89 | +The following table summarizes the steps and who usually performs them to enable ARIA to run experiments with W&B Launch. |
| 90 | + |
| 91 | +| Step | Usually performed by | Details | |
| 92 | +| -------------------------------------- | ----------------------------------- | ---------------- | |
| 93 | +| Create a Launch queue | ARIA or team admin | ARIA can help create a queue when the required permissions and target backend are clear. Team admin permissions might be required | |
| 94 | +| Choose or provision a compute backend | Team member or infrastructure owner | ARIA needs access to a Docker host, Kubernetes cluster, SageMaker environment, or Vertex AI environment. ARIA cannot provision infrastructure unless it is already accessible through W&B Launch. | |
| 95 | +| Start the Launch agent | Team member or infrastructure owner | The Launch agent must run persistently on your machine or cluster with W&B credentials. | |
| 96 | +| Configure a service account or API key | Team admin | Create and store service account keys securely. For more information, see [Manage secrets](https://docs.wandb.ai/platform/secrets#manage-secrets). | |
| 97 | +| Submit experiments to the queue | ARIA | After a queue has an active agent, ARIA can submit jobs, relaunch runs with config overrides, monitor results, debug failures, and compare metrics. | |
0 commit comments