Sources: Add, Sync, and Troubleshoot

The sources page is where data stops being a plan and starts becoming searchable context.

Use /dashboard/sources when you want to:

• add a connector-backed source to a project
• check whether it is connecting, indexing, ready, or failed
• retry or debug a source that never finished syncing

What the page supports

The current dashboard groups source connectors such as:

• GitHub
• GitLab
• Notion
• Confluence
• Slack
• Discord
• URL
• Web Crawler
• Sitemap
• PDF
• Text
• Database
• API Spec
• NPM Package
• PyPI Package
• Hugging Face
• ArXiv
• AI Browser

Connectors are shown with category groupings and a per-source status badge.

The shortest path to a working source

1. Make sure you already have the target project.
2. Open /dashboard/sources.
3. Choose the connector you want.
4. Fill in the minimum connector-specific configuration.
5. Create the source.
6. Wait for the status to move toward READY.
7. Only then test retrieval.

Understanding status

The dashboard currently surfaces statuses like:

• PENDING
• CONNECTING
• INDEXING
• READY
• ERROR
• DISABLED

The only status that should feel “done” for first-run testing is READY.

PENDING, CONNECTING, and INDEXING mean the source exists, but the content may not be ready for retrieval yet.

Two important distinctions

Integration vs source

GitHub or Slack can be connected at the org level without any project source actually being created yet.

If the OAuth connection succeeded but your project still has no searchable data, you are missing the source step.

Source created vs source indexed

A source record appearing in the UI does not guarantee the sync completed.

The real confirmation is a healthy status transition and then successful retrieval from the same project.

Troubleshooting the common failures

The source never becomes ready

Check:

1. whether the connector was configured against the right project
2. whether OAuth-backed setup was finished in Integrations first
3. whether the source is stuck in ERROR with a connector-specific failure

The source is ready but search is empty

Check:

1. the query is running against the same project
2. the source actually contains the content you expect
3. you are using the right retrieval surface for the data you ingested

The wrong connector was chosen

Use:

• URL for one page
• Web Crawler for multi-page crawl
• Sitemap when the site already exposes a stable sitemap
• AI Browser when the target content is JS-rendered and plain fetch is not enough

What a good verification loop looks like

After the source is ready:

1. run one known retrieval query against the same project
2. confirm the answer or hit references the newly synced data
3. only then add more sources or larger sync jobs

Next step

If you need to connect GitHub or Slack before adding a source, read integrations: OAuth, GitHub, and Slack. If you want the API equivalent of this workflow, go to projects and sources.