> AI agents: this is one page from PostHog's docs. Full index of Markdown docs for LLMs: https://posthog.com/llms.txt

# Linking Weights & Biases as a source - Docs

Copy page

# Linking Weights & Biases as a source - Docs

![](https://res.cloudinary.com/dmukukwp6/image/upload/texture_tan_9608fcca70)

![](https://res.cloudinary.com/dmukukwp6/image/upload/texture_tan_dark_a92b0e022d)

Let AI connect your sources for you

Skip the manual setup — run this in your project and the wizard auto-detects your databases and APIs and connects them to PostHog.

`npx @posthog/wizard warehouse`

[Learn more](/wizard.md)

![PostHog Wizard hedgehog](https://res.cloudinary.com/dmukukwp6/image/upload/wizard_3f8bb7a240.png)

![](https://res.cloudinary.com/dmukukwp6/image/upload/wizard_3f8bb7a240.png)Let AI connect your sources for you

The Weights & Biases connector syncs your ML experiment tracking data into PostHog, including projects, runs, sweeps, reports, and artifact versions across every project in your entity.

**Alpha release**

The Weights & Biases source is currently in alpha. Per-step run metric history isn't synced – runs include their final summary metrics instead.

## Prerequisites

You need a Weights & Biases account and an API key. API keys are personal and grant access to everything your user can see, so consider a service account for team syncs.

## Adding a data source

1.  In PostHog, go to the [Data pipeline page](https://app.posthog.com/data-management/sources) and select the **Sources** tab.

2.  Click **\+ New source** and then click **Link** next to Weights & Biases.

3.  Get your API key from [wandb.ai/authorize](https://wandb.ai/authorize) and paste it into the **API key** field.

4.  Enter the **Entity** whose projects you want to sync. This is your W&B username or team name, the first segment of your project URLs (`wandb.ai/<entity>/<project>`).

5.  If you use W&B Dedicated Cloud or a self-managed server, set the **Host** to your deployment's URL (for example `https://acme.wandb.io`). Leave it empty for W&B SaaS cloud.

6.  Click **Next**.

7.  On the next page, select the tables you want to sync and configure the sync method and frequency. Click **Import**.

Once the sync completes, you can start querying your Weights & Biases data in PostHog.

## Configuration

| Option | Type | Required |
| --- | --- | --- |
| API key | password | Yes |
| Entity (username or team) | text | Yes |
| Host (Dedicated Cloud or self-managed only) | text | No |

## Supported tables

| Table | Description | Sync method | Incremental field | Primary key |
| --- | --- | --- | --- | --- |
| projects | A project groups the runs, sweeps, reports and artifacts for one ML workstream within the entity. | Full refresh | — | — |
| runs | One row per run across every project in the entity, including config, tags and summary metrics. Incremental syncs filter server-side on createdAt or heartbeatAt; pick heartbeatAt to also pick up state and metric changes on recently active runs | Incremental, Full refresh | createdAt, heartbeatAt | — |
| sweeps | A hyperparameter search that launches and coordinates a set of runs. | Full refresh | — | — |
| reports | A saved report (workspace view) documenting and visualizing runs in a project. | Full refresh | — | — |
| artifacts | One row per artifact version across every project, walked per artifact type and collection. Projects with many run-history artifacts can make this table large | Full refresh | — | — |

## Sync details

-   **Incremental sync** – The `runs` table supports incremental syncs on `createdAt` or `heartbeatAt`. PostHog filters server-side, so only matching runs are fetched on each sync. Pick `heartbeatAt` to also pick up state and summary metric changes on recently active runs; a `createdAt` cursor only fetches newly created runs. The other tables are full refresh.
-   **JSON columns** – Run `config`, `summaryMetrics`, and `systemMetrics` are JSON-encoded strings, matching what the W&B API returns. Parse them in queries with `JSONExtract` functions.
-   **Partitioning** – Data is partitioned by month on the `createdAt` field.
-   **Pagination** – Uses cursor pagination. If a sync is interrupted, it resumes from the last successfully synced page.
-   **Rate limits** – The W&B API allows roughly 50 requests per minute on free plans and 200 on paid plans. PostHog backs off and retries automatically, but very large entities can take a while to sync.

## Troubleshooting

-   **Invalid API key** – The key may have been revoked. Generate a new one at [wandb.ai/authorize](https://wandb.ai/authorize) and update the source credentials.
-   **Tables sync but are empty** – Check the entity name. It must match the username or team that owns your projects exactly.

### Was this page useful?

HelpfulCould be better