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

# Connect your AI provider to RAUM AI

> Bring your own API key from any major AI provider. RAUM AI encrypts your key at rest and uses it to power the full ticket-resolution pipeline.

RAUM AI does not bundle an AI model — you connect your own provider account, which means you pay your AI provider directly at their standard rates. raum adds no markup on AI costs. Your API key is encrypted with AES-256-GCM before storage and is never exposed in logs or API responses.

## Supported providers

RAUM AI supports the following providers out of the box:

<CardGroup cols={3}>
  <Card title="OpenAI" icon="brain">
    GPT-4o, GPT-4 Turbo, and the full OpenAI model family, including `text-embedding-3-small` and `text-embedding-3-large`.
  </Card>

  <Card title="Anthropic" icon="sparkles">
    Claude 3 Opus, Sonnet, and Haiku. Strong performance on long-context reasoning and nuanced tone.
  </Card>

  <Card title="Google Gemini" icon="stars">
    Gemini 1.5 Pro and Flash. Includes native embedding models via the Gemini API.
  </Card>

  <Card title="Azure OpenAI" icon="cloud">
    OpenAI models deployed through your own Azure resource. Requires a base URL in addition to your API key.
  </Card>

  <Card title="AWS Bedrock" icon="server">
    Access Anthropic, Meta, and other models through the AWS Bedrock gateway. Requires a gateway URL.
  </Card>

  <Card title="Vertex AI" icon="database">
    Google models via the Vertex AI API. Requires a gateway URL pointing to your Vertex endpoint.
  </Card>

  <Card title="DeepSeek" icon="code">
    DeepSeek-V2 and DeepSeek-Chat. Cost-efficient option for high-volume support workflows.
  </Card>

  <Card title="Groq" icon="bolt">
    Ultra-low latency inference for Llama 3 and Mixtral models. Optimised for speed-sensitive deployments.
  </Card>

  <Card title="Mistral" icon="wind">
    Mistral Large, Mistral 8x7B, and Mistral Embed for embeddings.
  </Card>

  <Card title="Moonshot (Kimi)" icon="moon">
    Kimi long-context models. Well-suited for tickets with large attachment payloads.
  </Card>

  <Card title="Qwen" icon="globe">
    Alibaba's Qwen model family via the Dashscope API.
  </Card>

  <Card title="OpenRouter" icon="arrow-right-arrow-left">
    Route requests across multiple providers through a single OpenRouter API key.
  </Card>

  <Card title="Custom" icon="wrench">
    Any OpenAI-compatible endpoint. Supply a base URL and your credentials.
  </Card>
</CardGroup>

## Connect your provider

<Steps>
  <Step title="Open AI Engine settings">
    In the RAUM AI dashboard, click **Settings** in the left sidebar, then select the **AI Mode** tab.
  </Step>

  <Step title="Select your provider">
    Click the card that matches your AI provider. The form updates to show only the fields relevant to that provider.
  </Step>

  <Step title="Enter your API key">
    Paste your API key into the **API Key** field. The key is masked on input and encrypted with AES-256-GCM before it is saved.
  </Step>

  <Step title="Enter a base URL (if required)">
    Some providers require a base URL or gateway URL in addition to an API key. See [provider-specific notes](#provider-specific-notes) below.
  </Step>

  <Step title="Verify your key and fetch models">
    Click **Verify API Key & Fetch Models**. RAUM AI contacts your provider to confirm the key is valid and retrieves the list of available models.
  </Step>

  <Step title="Select your models">
    Choose an **All-purpose model** (used for reasoning, planning, and reply generation) and an **Embedding model** (used for semantic search over your procedures and articles). Both dropdowns are populated from your live model list.
  </Step>

  <Step title="Save">
    Click **Save AI Settings**. RAUM AI will use these settings immediately for all subsequent ticket processing.
  </Step>
</Steps>

## Provider-specific notes

<AccordionGroup>
  <Accordion title="Azure OpenAI">
    Azure OpenAI requires your deployment's base URL in addition to an API key. The URL follows the format:

    ```text theme={null}
    https://your-resource-name.openai.azure.com/
    ```

    You create this URL in the Azure portal when you deploy a model. Make sure your deployment name matches a model that supports both chat completions and embeddings, or use separate deployments for each.
  </Accordion>

  <Accordion title="AWS Bedrock">
    AWS Bedrock requires a gateway URL pointing to the Bedrock runtime in your AWS region:

    ```text theme={null}
    https://bedrock-runtime.{region}.amazonaws.com/openai/v1
    ```

    Replace `{region}` with your AWS region (e.g., `us-east-1`). Your API key should be an AWS access key or a token from a Bedrock-compatible gateway.
  </Accordion>

  <Accordion title="Vertex AI">
    Vertex AI requires a gateway URL that points to your project's OpenAI-compatible endpoint on Google Cloud:

    ```text theme={null}
    https://aiplatform.googleapis.com/v1/projects/{project}/locations/{location}/publishers/google/models/{model}:serverStreamingPredict
    ```

    RAUM AI uses the OpenAI-compatible path, so confirm your endpoint supports that interface before saving.
  </Accordion>

  <Accordion title="OpenRouter">
    OpenRouter routes requests to multiple underlying providers. Use `https://openrouter.ai/api/v1` as the base URL and your OpenRouter API key. You can then select any model available on OpenRouter from the model dropdown.
  </Accordion>

  <Accordion title="Custom (OpenAI-compatible)">
    Supply the full base URL of your endpoint, for example `https://api.yourprovider.com/v1`. RAUM AI will append standard OpenAI paths such as `/chat/completions` and `/embeddings`. Your endpoint must support these paths.
  </Accordion>
</AccordionGroup>

## About bring-your-own-key (BYOK)

RAUM AI uses a strict BYOK model:

* You create and own your API key directly with the AI provider.
* Usage is billed to your provider account at their standard rates.
* RAUM AI does not resell AI capacity or add a cost multiplier.
* You can rotate or revoke your key at any time without affecting your RAUM AI account — simply update the key in **Settings → AI Mode**.

## Failover between providers

RAUM AI includes a circuit breaker that tracks provider errors. If your primary provider returns repeated failures, RAUM AI surfaces the issue in the dashboard rather than silently degrading responses. You can configure a secondary provider key for automatic failover in **Settings → AI Mode → Failover Provider**.

<Note>
  Embedding models and all-purpose models must come from the same provider unless you configure a dedicated embedding provider separately.
</Note>