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

# IBM watsonx

> Connect IBM watsonx.ai to Rootly to power incident summaries, alert summaries, and AI text editing with foundation models in your IBM Cloud region.

## Overview

IBM watsonx.ai hosts foundation models in your IBM Cloud account, in a region you control, with enterprise auth and audit. Connect a watsonx project to Rootly and the same AI features that run on OpenAI or Anthropic — incident summaries, alert summaries, text editor, title generation — route through your watsonx project instead. Your incident data and AI prompts stay inside IBM Cloud's regional boundary the entire time.

The integration uses an IBM Cloud API key plus a regional inference endpoint. Rootly exchanges the API key for IAM bearer tokens automatically and refreshes them as needed, so the customer-side configuration is just two fields: region and key.

<CardGroup cols={2}>
  <Card title="Pick Your Region" icon="globe">
    Six IBM Cloud regions available (Dallas, Frankfurt, London, Tokyo, Sydney, Toronto) so inference happens in the jurisdiction your compliance team approved.
  </Card>

  <Card title="Pick Your Model" icon="cube">
    Choose any foundation model your watsonx project has access to — Granite, Llama variants, and others — pulled live from your project at configuration time.
  </Card>

  <Card title="AI-Powered Summaries" icon="sparkles">
    Drives incident summarization, alert summarization, the in-app text editor, and incident title generation when WatsonX is the active provider.
  </Card>

  <Card title="Enterprise Auth" icon="shield-keyhole">
    IBM Cloud API keys exchanged for short-lived IAM bearer tokens. Rootly handles the token refresh dance; you never paste a long-lived bearer.
  </Card>
</CardGroup>

***

## Before You Begin

<Info>
  **You'll need access to both sides of the connection.**

  * **In IBM Cloud** — permission to create an API key under your IBM Cloud account and a watsonx.ai project the API key has access to. Most enterprises require IAM access via a watsonx-specific role.
  * **In Rootly** — an admin role so you can reach **Configuration → Integrations** and complete the WatsonX setup.

  If you're evaluating IBM watsonx for the first time, IBM's [Quick start guide](https://www.ibm.com/watsonx/developer/get-started/quick-start/) walks through provisioning a watsonx instance and creating your first project before you reach the credential step.
</Info>

***

## Generate Your IBM watsonx API Key

The vendor-side setup happens in IBM Cloud. Refer to [IBM's API key documentation](https://dataplatform.cloud.ibm.com/docs/content/wsj/analyze-data/fm-credentials.html) for the latest UI specifics; the high-level flow is below.

<Steps>
  <Step title="Open IBM Cloud API Keys">
    Sign into IBM Cloud and navigate to **Manage → Access (IAM) → API keys**. Click **Create**.
  </Step>

  <Step title="Create The Key">
    Name the key something descriptive (`rootly-watsonx-production` works well). Choose an account-level or service-ID-scoped key based on your security policy. Click **Create**.

    <Warning>
      IBM Cloud shows the API key value **once**. Copy it now and store it in your secrets manager — you cannot retrieve it later, only revoke and reissue.
    </Warning>
  </Step>

  <Step title="Confirm Project Access">
    Open the watsonx.ai project Rootly should use, go to **Manage → Access control**, and confirm the IBM Cloud identity tied to the API key has at least **Editor** access to the project (Viewer is insufficient for inference calls).
  </Step>

  <Step title="Note Your Region">
    Confirm which IBM Cloud region hosts your watsonx project. Rootly supports the six watsonx.ai regions — Dallas (`us-south`), Frankfurt (`eu-de`), London (`eu-gb`), Tokyo (`jp-tok`), Sydney (`au-syd`), and Toronto (`ca-tor`). The region you pick in Rootly must match the project's region exactly.
  </Step>
</Steps>

***

## Connect WatsonX to Rootly

With the API key in hand, the Rootly side is two fields and a save.

<Steps>
  <Step title="Open The WatsonX Settings">
    In Rootly, go to **Configuration → Integrations** and locate **WatsonX**. Click **Setup**.
  </Step>

  <Step title="Pick The Region">
    Select the IBM Cloud region from the **Host** dropdown:

    <Tabs>
      <Tab title="Americas">
        * `us-south.ml.cloud.ibm.com` — Dallas
        * `ca-tor.ml.cloud.ibm.com` — Toronto
      </Tab>

      <Tab title="Europe">
        * `eu-de.ml.cloud.ibm.com` — Frankfurt
        * `eu-gb.ml.cloud.ibm.com` — London
      </Tab>

      <Tab title="Asia-Pacific">
        * `jp-tok.ml.cloud.ibm.com` — Tokyo
        * `au-syd.ml.cloud.ibm.com` — Sydney
      </Tab>
    </Tabs>

    Pick the region that matches your watsonx.ai project. Inference requests route to this regional endpoint; data residency follows.
  </Step>

  <Step title="Paste The API Key">
    Paste the IBM Cloud API key into the **API Key** field. The value is stored encrypted at rest; Rootly exchanges it for short-lived IAM bearer tokens on each request and refreshes them automatically.
  </Step>

  <Step title="Save And Verify">
    Click **Save**. Rootly tests the credentials against the regional endpoint immediately — a successful save confirms the API key authenticates against IAM and the watsonx project is reachable.
  </Step>
</Steps>

***

## Configuration Reference

<ParamField path="Host" type="enum" required>
  IBM Cloud region for inference. Must match the region of the watsonx project the API key has access to.

  Valid values: `us-south.ml.cloud.ibm.com`, `eu-de.ml.cloud.ibm.com`, `eu-gb.ml.cloud.ibm.com`, `jp-tok.ml.cloud.ibm.com`, `au-syd.ml.cloud.ibm.com`, `ca-tor.ml.cloud.ibm.com`.
</ParamField>

<ParamField path="API Key" type="string" required>
  IBM Cloud API key with at least Editor access to the watsonx.ai project. Stored encrypted at rest. Rotated by revoking the key in IBM Cloud and pasting a new one here.
</ParamField>

<ParamField path="Foundation Model" type="string">
  The watsonx model Rootly invokes for AI tasks. Rootly defaults to a Llama variant but exposes any foundation model your watsonx project lists. Configurable per AI feature; the model list updates from your project on each Rootly configuration load.
</ParamField>

***

## AI Features Powered by WatsonX

Once WatsonX is connected and enabled as the active AI provider, it drives the following Rootly features:

<CardGroup cols={2}>
  <Card title="Incident Summaries" icon="file-lines">
    Auto-generated summaries of incident channels, timelines, and key events for retrospectives and stakeholder updates.
  </Card>

  <Card title="Alert Summaries" icon="bell">
    Condensed, plain-English versions of monitoring alert payloads so responders triage faster.
  </Card>

  <Card title="Text Editor" icon="pen-to-square">
    The in-app text editor's AI actions (rewrite, expand, simplify, tone-shift) call watsonx instead of the default provider.
  </Card>

  <Card title="Incident Title Generation" icon="heading">
    Suggested incident titles based on the triggering alert and channel context.
  </Card>
</CardGroup>

<Note>
  Switching the active AI provider — between watsonx, OpenAI, Anthropic, and others — is a per-team configuration. WatsonX being connected doesn't automatically route every team's AI features through it. Contact Rootly support or your account team to activate WatsonX for specific teams or organization-wide.
</Note>

***

## Selecting a Foundation Model

Rootly pulls the list of available foundation models live from your watsonx project. Whatever models your IBM Cloud identity has access to — IBM Granite, Meta Llama variants, third-party models published in watsonx, custom-tuned models — appear in the selector when configuring AI features.

Default model behavior:

* **Out of the box**, Rootly invokes a recent instruction-tuned Llama variant suitable for summarization and short-form generation
* **To override**, contact your Rootly account team — model selection is managed via a server-side configuration that maps each AI task type to a specific model ID
* **Model availability** depends on your IBM Cloud region; not all foundation models are available in every region

Refer to IBM's [foundation models documentation](https://www.ibm.com/products/watsonx-ai/foundation-models) for the current list of supported models and their regional availability.

***

## Test the Integration

After saving the WatsonX settings, verify end-to-end inference works.

<Steps>
  <Step title="Trigger A Test AI Feature">
    Open any active incident and use the AI text editor (rewrite or summarize), or open the AI summary panel. If WatsonX is the active provider for your team, the request routes through your IBM Cloud project.
  </Step>

  <Step title="Confirm Inference Succeeds">
    The AI response should appear within a few seconds. If it doesn't, see [Troubleshooting](#troubleshooting) below.
  </Step>

  <Step title="Check IBM Cloud Activity">
    Optional — open IBM Cloud's Activity Tracker for your watsonx project to confirm the inference request arrived from Rootly and which foundation model it invoked. Useful for capacity planning and cost attribution.
  </Step>
</Steps>

<Check>
  A successful AI response in Rootly plus the corresponding inference event in IBM Cloud's Activity Tracker confirms WatsonX is wired correctly. From here, broaden the WatsonX rollout to additional teams as needed.
</Check>

***

## Troubleshooting

<AccordionGroup>
  <Accordion title="Save fails with an authentication error" icon="lock">
    The API key is invalid, expired, or doesn't have access to the chosen region's watsonx project. Confirm:

    * The key value was copied without surrounding whitespace
    * The IBM Cloud identity attached to the key has at least Editor access to the watsonx project in the selected region
    * The key hasn't been revoked in IBM Cloud's API keys panel
  </Accordion>

  <Accordion title="AI responses time out or fail silently" icon="ghost">
    Most common causes:

    * The selected region's watsonx endpoint is throttling — check IBM Cloud's status page for region-specific incidents
    * The foundation model selected isn't available in your region — switch to a model that lists your region in IBM's availability matrix
    * Your IBM Cloud plan's monthly tokens-per-minute quota is exhausted — review usage in IBM Cloud and consider upgrading
  </Accordion>

  <Accordion title="The model list is empty in Rootly" icon="square-question">
    The API key authenticates but the project has no models accessible to that identity. Open the watsonx project in IBM Cloud and confirm at least one foundation model is associated with the project under **Manage → Foundation models**. Refresh the Rootly WatsonX settings page after granting access.
  </Accordion>

  <Accordion title="AI features still route to OpenAI or Anthropic after saving WatsonX" icon="shuffle">
    Provider routing is per-team and managed server-side. Connecting WatsonX makes it available; activating it for specific teams or org-wide requires a Rootly account team change. Reach out to support to flip the routing.
  </Accordion>

  <Accordion title="403 errors from IBM Cloud" icon="user-shield">
    The API key authenticated, but the IBM Cloud identity lacks permission for the specific watsonx operation. In IBM Cloud IAM, confirm the identity has the **watsonx.ai Service** access policy plus project-level Editor access. IBM's IAM separates platform-level and service-level permissions; both are required.
  </Accordion>

  <Accordion title="Rotating the API key" icon="rotate">
    Revoke the existing key in IBM Cloud's API keys panel, create a new one with the same scope, and paste the new value into Rootly's WatsonX settings. The previous key continues to work in Rootly until you save the new value, then it's immediately replaced.
  </Accordion>
</AccordionGroup>

***

## Frequently Asked Questions

<AccordionGroup>
  <Accordion title="How does WatsonX compare to OpenAI or Anthropic in Rootly?" icon="circle-info">
    All three power the same AI features (incident summaries, alert summaries, text editor, title generation). WatsonX is the right choice when your security or compliance posture requires inference inside IBM Cloud's regional boundary, or when you've already committed to IBM Cloud for other workloads. OpenAI and Anthropic are simpler to set up but route data through their own hosted endpoints.
  </Accordion>

  <Accordion title="Can I connect multiple watsonx regions?" icon="circle-question">
    One watsonx project per Rootly organization. To use multiple regions, pick the one your primary compliance requirements demand and use that project across your watsonx workloads.
  </Accordion>

  <Accordion title="Does my incident data leave IBM Cloud?" icon="triangle-exclamation">
    Inference requests from Rootly to watsonx travel over TLS to the regional endpoint you chose. The prompt and response are processed by IBM Cloud's watsonx service in that region; no data routes outside IBM Cloud during inference. Rootly stores the response back into the incident or alert record.
  </Accordion>

  <Accordion title="Can I use a custom-tuned foundation model?" icon="list">
    Yes, provided the model is published in your watsonx project and accessible to the IBM Cloud identity behind the API key. Contact your Rootly account team to set the custom model as the active model for specific AI tasks.
  </Accordion>

  <Accordion title="What's the cost model?" icon="bookmark">
    Inference cost is billed by IBM Cloud based on tokens consumed per foundation model. Rootly doesn't add a markup. Review IBM Cloud's watsonx.ai pricing for current per-model rates.
  </Accordion>
</AccordionGroup>

***

## Next Steps

<CardGroup cols={2}>
  <Card title="Anthropic" icon="brain" href="/integrations/anthropic">
    Alternate AI provider — Claude models hosted by Anthropic, no IBM Cloud account required.
  </Card>

  <Card title="OpenAI" icon="sparkles" href="/integrations/openai">
    Alternate AI provider — GPT models hosted by OpenAI, simpler setup but no regional control.
  </Card>

  <Card title="Azure OpenAI" icon="microsoft" href="/integrations/azure-openai">
    Enterprise-hosted GPT models inside your Azure tenancy if you're already on Microsoft Cloud.
  </Card>

  <Card title="User Permissions" icon="user-shield" href="/managing-users/user-permissions">
    Control which users and roles can configure and trigger AI features.
  </Card>
</CardGroup>