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

# LLM configuration

> Connect LLM providers to Hadiq.io and set the default model for your organization.

Hadiq.io supports a wide range of LLM providers — both hosted APIs and self-hosted models. You add providers in **Admin** > **Configuration** > **LLM**. Once a provider is connected, its models become available for chat, agents, and search summarization across your organization.

## Supported providers

<AccordionGroup>
  <Accordion title="OpenAI">
    Models: GPT-4o, GPT-4o mini, GPT-4, GPT-4 Turbo, o1, o3-mini, and others.

    Requires an OpenAI API key from [platform.openai.com](https://platform.openai.com).
  </Accordion>

  <Accordion title="Anthropic">
    Models: Claude 3.5 Sonnet, Claude 3.5 Haiku, Claude 3 Opus, and others.

    Requires an Anthropic API key from [console.anthropic.com](https://console.anthropic.com).
  </Accordion>

  <Accordion title="Google">
    Models: Gemini 2.5 Pro, Gemini 2.0 Flash, Gemini 1.5 Pro, and others.

    Requires a Google AI API key or Vertex AI credentials.
  </Accordion>

  <Accordion title="Mistral">
    Models: Mistral Large, Mistral Small, and others.

    Requires a Mistral API key from [console.mistral.ai](https://console.mistral.ai).
  </Accordion>

  <Accordion title="Cohere">
    Models: Command R+, Command R, and others.

    Requires a Cohere API key from [dashboard.cohere.com](https://dashboard.cohere.com).
  </Accordion>

  <Accordion title="AWS Bedrock">
    Access Anthropic, Meta, Mistral, and Amazon models through AWS Bedrock. Requires AWS credentials (access key + secret, or IAM role) and a Bedrock-enabled region.
  </Accordion>

  <Accordion title="Ollama (self-hosted)">
    Run any model locally or on your own infrastructure using [Ollama](https://ollama.com). Hadiq.io auto-discovers the models available on your Ollama server.
  </Accordion>

  <Accordion title="LiteLLM proxy">
    Route requests through a [LiteLLM](https://github.com/BerriAI/litellm) proxy to access any model it supports, including custom deployments and load-balanced endpoints.
  </Accordion>

  <Accordion title="OpenRouter">
    Access hundreds of models through the OpenRouter API. Requires an OpenRouter API key and base URL.
  </Accordion>

  <Accordion title="LM Studio">
    Connect to a locally running [LM Studio](https://lmstudio.ai) server. Hadiq.io fetches the available models and their capabilities automatically.
  </Accordion>

  <Accordion title="vLLM and other OpenAI-compatible servers">
    Any server that exposes an OpenAI-compatible `/v1/chat/completions` endpoint can be connected using the **OpenAI-compatible** provider option.
  </Accordion>
</AccordionGroup>

## Adding a provider

<Tabs>
  <Tab title="OpenAI">
    <Steps>
      <Step title="Open LLM settings">
        Go to **Admin** > **Configuration** > **LLM** and click **Add LLM Provider**.
      </Step>

      <Step title="Select OpenAI">
        Choose **OpenAI** from the list of built-in providers.
      </Step>

      <Step title="Enter your API key">
        Paste your OpenAI API key. Hadiq.io encrypts and stores the key; it is never exposed in full after saving.
      </Step>

      <Step title="Choose models">
        Select which models to enable. Toggle visibility on or off for each model to control which ones appear in the model picker.
      </Step>

      <Step title="Save">
        Click **Save**. Hadiq.io tests the connection before saving. If the test fails, verify your API key and try again.
      </Step>
    </Steps>
  </Tab>

  <Tab title="Anthropic">
    <Steps>
      <Step title="Open LLM settings">
        Go to **Admin** > **Configuration** > **LLM** and click **Add LLM Provider**.
      </Step>

      <Step title="Select Anthropic">
        Choose **Anthropic** from the list of built-in providers.
      </Step>

      <Step title="Enter your API key">
        Paste your Anthropic API key.
      </Step>

      <Step title="Choose models">
        Select and toggle visibility for the Claude models you want to make available.
      </Step>

      <Step title="Save">
        Click **Save**. Hadiq.io verifies the connection before storing the configuration.
      </Step>
    </Steps>
  </Tab>

  <Tab title="Ollama (self-hosted)">
    <Steps>
      <Step title="Start Ollama">
        Make sure your Ollama server is running and reachable from the Hadiq.io backend. Pull the models you want with `ollama pull <model-name>`.
      </Step>

      <Step title="Open LLM settings">
        Go to **Admin** > **Configuration** > **LLM** and click **Add LLM Provider**.
      </Step>

      <Step title="Select Ollama">
        Choose **Ollama** from the list of providers.
      </Step>

      <Step title="Enter the API base URL">
        Enter the URL of your Ollama server, for example `http://localhost:11434`. Hadiq.io queries the `/api/tags` endpoint to discover available models.
      </Step>

      <Step title="Select models">
        Hadiq.io lists the models found on your server. Select the ones you want to enable.
      </Step>

      <Step title="Save">
        Click **Save**.
      </Step>
    </Steps>
  </Tab>

  <Tab title="AWS Bedrock">
    <Steps>
      <Step title="Open LLM settings">
        Go to **Admin** > **Configuration** > **LLM** and click **Add LLM Provider**.
      </Step>

      <Step title="Select Bedrock">
        Choose **Bedrock** from the list of providers.
      </Step>

      <Step title="Enter credentials">
        Provide your AWS region and either:

        * An access key ID and secret access key, or
        * A Bedrock bearer token

        If neither is provided, Hadiq.io falls back to the instance's IAM role credentials.
      </Step>

      <Step title="Fetch models">
        Click **Fetch Models**. Hadiq.io queries the Bedrock API and lists the foundation models and inference profiles available in your account and region.
      </Step>

      <Step title="Select models and save">
        Choose the models to enable and click **Save**.
      </Step>
    </Steps>
  </Tab>

  <Tab title="OpenAI-compatible (vLLM, custom)">
    <Steps>
      <Step title="Open LLM settings">
        Go to **Admin** > **Configuration** > **LLM** and click **Add LLM Provider**.
      </Step>

      <Step title="Select OpenAI-compatible">
        Choose **OpenAI-compatible** from the provider list.
      </Step>

      <Step title="Enter connection details">
        * **API base URL**: The base URL of your server, e.g. `http://your-vllm-host:8000/v1`
        * **API key**: If your server requires authentication, enter the key here. Leave blank for unauthenticated servers.
      </Step>

      <Step title="Fetch and select models">
        Click **Fetch Models**. Hadiq.io queries the `/v1/models` endpoint on your server. Select the models you want to make available.
      </Step>

      <Step title="Save">
        Click **Save**.
      </Step>
    </Steps>
  </Tab>
</Tabs>

## Setting the default LLM

The default LLM is used for all conversations and agents unless a user or agent specifies a different model.

<Steps>
  <Step title="Open LLM settings">
    Go to **Admin** > **Configuration** > **LLM**.
  </Step>

  <Step title="Set the default">
    Find the provider containing your preferred model and click **Set as default**. Then select the specific model from that provider to use as the organization default.
  </Step>
</Steps>

You can also set a separate **default vision model** for tasks that involve processing images. Use the **Set as default vision** option on any provider that includes a vision-capable model.

## Restricting provider access to groups

By default, all LLM providers are public — every user can select any enabled model. To restrict a provider to specific groups:

1. Open the provider in **Admin** > **Configuration** > **LLM**.
2. Toggle **Public** off.
3. Select the groups that should have access to this provider.
4. Save.

Users outside the selected groups will not see models from that provider in their model picker.

## Testing a provider

After saving a provider, click **Test** to send a short request and verify the connection. If the test fails, Hadiq.io shows the error message returned by the provider's API.

## Auto mode

Some providers support **Auto mode**, which automatically syncs the list of available models from the provider's API on a regular basis. When a new model becomes available (for example, a new GPT-4o variant), it appears in Hadiq.io without manual updates. Enable Auto mode in the provider settings.

<Note>
  Auto mode is available for providers where Hadiq.io maintains a model registry, including OpenAI and Anthropic. For self-hosted providers like Ollama, models are fetched on demand when you open the provider settings.
</Note>