> ## Documentation Index
> Fetch the complete documentation index at: https://langwatch.ai/docs/llms.txt
> Use this file to discover all available pages before exploring further.

# Langy Troubleshooting

> Common Langy problems and the exact check for each, from Langy not appearing in your org to a PR it cannot open, no configured model, and a stopped turn.

Each problem below lists the check that resolves it. Start with the one that matches the symptom.

<AccordionGroup>
  <Accordion title="Langy is not available in my org">
    * **Cloud:** Langy is available. Open the panel to get started.
    * **Self-hosted:** the agent pod deploys by default. Confirm it was not disabled (`langyagent.chartManaged=false`) and the pod is running. A running pod is not the whole story: with `enableForAllUsers: false` the install keeps the rollout flag authoritative, so Langy stays dark until you target the projects or organizations you want. See [Self-hosting overview](/docs/self-hosting/langy/overview) and [Setup](/docs/self-hosting/langy/setup#staged-rollout).
  </Accordion>

  <Accordion title="Langy can't open a PR">
    Two common causes.

    * **The GitHub App is not installed.** Langy opens pull requests only on repositories where its GitHub App is installed. When it reaches for `git` or `gh` and no install exists, it stops the turn and shows an Install card in chat. Use that card to install the app, then re-run the turn. See [GitHub access](/docs/langy/security/github-access).
    * **Outbound HTTPS is off (self-hosted).** A hardened install with `networkPolicy.allowExternalHttps` set to `false` blocks `github.com`, `api.github.com`, and `codeload.github.com`. Allow those FQDNs (or turn `allowExternalHttps` on) so the worker can push and open the PR. See [Networking and egress](/docs/self-hosting/langy/networking-and-egress).
    * Also check you are under the daily PR cap. Langy opens at most 20 pull requests per project per day. See [Limits](/docs/langy/reference/limits).
  </Accordion>

  <Accordion title="Langy has no model">
    Langy runs on the providers your project has enabled, and there is no default key. Configure a provider (paste an API key) or sign in with your OpenAI account for Codex. Until a model is available for the Langy role, Langy cannot run a turn. See [Supported models](/docs/langy/reference/supported-models) and [Codex](/docs/langy/codex).
  </Accordion>

  <Accordion title="A turn was stopped">
    You can stop a running turn for real, and you can continue or resume afterward. A stop is an intentional outcome, not a failure: Langy keeps the partial answer streamed so far, and nothing you did is wrong. Continue the conversation, or ask again. A refresh mid-turn rejoins the running turn rather than losing it.
  </Accordion>

  <Accordion title="Langy's worker stopped mid-reply">
    If a worker process dies mid-reply, Langy surfaces a "worker stopped" card. Your message is safe. Send it again; a fresh worker spawns for the next turn.
  </Accordion>
</AccordionGroup>

## Audit-log events to check

When Langy touches GitHub, it writes audit-log events. Use these to confirm what happened and when.

| Event                     | Meaning                                                   |
| ------------------------- | --------------------------------------------------------- |
| `langy.github.install`    | The Langy GitHub App was installed and bound to your org. |
| `langy.github.disconnect` | The Langy GitHub App was disconnected.                    |

<Info>
  **Also check:** [GitHub access](/docs/langy/security/github-access), [Limits](/docs/langy/reference/limits), [Supported models](/docs/langy/reference/supported-models).
</Info>

## Reference

* [Self-hosting overview](/docs/self-hosting/langy/overview)
* [GitHub access](/docs/langy/security/github-access)
* [Supported models](/docs/langy/reference/supported-models)
* [Limits](/docs/langy/reference/limits)