Join the Waitlist
Guide

How to Connect a Local Ollama Model to DAVE

By DAVE by voolama6 stepsUpdated AUG 2026AI

Summary

DAVE supports Ollama as a first-class provider type, letting you point any agent at a locally hosted model without touching workflow logic or agent credentials. This guide walks through the full setup.

The steps

  1. Open the Providers page

    Navigate to Admin → API Management → Providers. You must have the Admin role to access this page. If the Providers item is not visible in the menu, your account does not have Admin access: contact your workspace administrator.

    The Providers page displays a card for each configured provider, showing its name, provider type badge, model count, and base URI. If this is a new workspace, the list will be empty.

    Do this
    • Log in to your DAVE workspace.
    • In the left sidebar, navigate to Admin → API Management → Providers.
    • Confirm you can see the provider list. If you see an access error, you do not have the Admin role.
    Example

    A workspace administrator logs in, clicks Admin in the sidebar, then API Management, then Providers. The page loads showing two existing provider cards: one for OpenAI and one for Anthropic. They are about to add a third for Ollama.

    Best practice

    Only users with the Admin role can view or manage providers. Keep provider management limited to the people responsible for AI infrastructure in your workspace. DAVE ships with six default tenant roles (Admin, Create, Curate, Use, Reporting, and Financial), all editable, so your administrator can align access to your governance requirements.

  2. Click Add Provider and select the Local (Ollama) type

    Click Add Provider in the top-right corner of the Providers page. The create form appears above the provider list.

    Fill in the Name field first. Choose a descriptive label that makes the provider's purpose and context immediately clear to anyone who sees it in an agent configuration dropdown. Then open the Provider Type dropdown and select Local (Ollama).

    As soon as you select Local (Ollama), DAVE auto-fills the Base URI field with the Ollama default endpoint: http://localhost:11434/v1. This is the correct value if Ollama is running on the same machine as the browser making the request. If your Ollama instance is running on a different host or port, replace this value with the correct endpoint before continuing.

    Do this
    • Click Add Provider in the top-right corner.
    • Enter a descriptive name in the Name field, for example: Ollama Local or Ollama Dev - llama3.2.
    • Open the Provider Type dropdown and select Local (Ollama).
    • Confirm the Base URI has been auto-filled as http://localhost:11434/v1. If your Ollama instance is on a different host or port, update this value now.
    Example

    An administrator names the provider Ollama Local - llama3.2, selects Local (Ollama) as the type, and sees the Base URI auto-fill to http://localhost:11434/v1. Their Ollama instance is running on the default port on the same machine, so they leave the URI unchanged.

    Best practice

    Use the provider name to record the model family and environment, for example Ollama Dev - mistral or Ollama Staging - llama3.2. If you run multiple Ollama instances pointing at different models or hosts, create a separate provider record for each one. This makes it easy to switch an agent between models by changing its provider reference, without touching the workflow graph.

  3. Enter the API key, or leave blank if Ollama runs without authentication

    The API Key field is optional for Ollama. By default, Ollama does not require authentication, and you can leave this field empty. If you have configured your Ollama instance to require a bearer token or API key, enter it here. The field is masked by default: click the eye icon to toggle visibility while you type.

    Whatever you enter is stored in DAVE's per-tenant credential vault, encrypted with AES-256-GCM authenticated encryption, and never returned through the API once saved. The key is transmitted only over HTTPS and is never logged in plain text.

    Do this
    • If your Ollama instance does not require authentication (the default), leave the API Key field empty.
    • If your Ollama instance requires a bearer token, enter it in the API Key field. Click the eye icon to verify what you have typed before saving.
    Example

    The administrator's Ollama instance is running on their local development machine with no authentication configured. They leave the API Key field blank and move on to the model catalog step.

    Best practice

    Even if your Ollama instance does not currently require authentication, consider whether it should. An unauthenticated Ollama endpoint accessible on a shared network can be called by anyone on that network. If you add authentication to Ollama later, return to this provider record, edit it, and enter the new key. Because providers are centralized, updating the key here updates every agent that references this provider.

  4. Fetch models from your Ollama instance

    With the Base URI and API Key configured, use the Model Catalog section to populate the list of models DAVE can use from this provider. The Local (Ollama) provider type supports model fetching: click the Fetch Models button to retrieve the list of models currently available in your Ollama instance.

    DAVE calls your Ollama endpoint and returns the models it finds. Use the search box to filter the list, then check the models you want to include in this provider's catalog. Use All or None to bulk-select. Only models you check here will appear in agent configuration dropdowns for this provider.

    If the fetch fails, see the FAQ section below. If you prefer to enter models manually, click the JSON button to switch to the JSON editor and paste a model array directly.

    Do this
    • In the Model Catalog section, confirm the toggle is set to Fetch from API (not JSON).
    • Click Fetch Models.
    • Wait for the model list to load. If you see an error, check that your Ollama instance is running and that the Base URI is correct.
    • Check the models you want to include. Uncheck any you do not need.
    • If you prefer to enter models manually, click JSON and paste a JSON array in the format [{ "id": "llama3.2", "name": "Llama 3.2" }].
    Example

    The administrator clicks Fetch Models. DAVE calls http://localhost:11434/v1 and returns three models: llama3.2, mistral, and nomic-embed-text. They check llama3.2 and mistral and leave nomic-embed-text unchecked, since it is an embedding model not used for agent interactions in this workspace.

    Best practice

    Only include models your team actually uses. A smaller catalog makes model selection in agent configuration faster and less error-prone. If you add a new model to your Ollama instance later, return to this provider, click the pencil icon to edit it, and run Fetch Models again to pick up the new model.

  5. Save the provider and verify connectivity

    Click Create Provider to save. DAVE stores the provider record and returns you to the provider list, where a new card for your Ollama provider appears with the provider type badge Local (Ollama), the model count from your catalog, and the base URI.

    To verify connectivity, click the pencil icon on the new provider card to open the edit form, then click Fetch Models again. A successful fetch confirms that DAVE can reach your Ollama endpoint with the stored configuration. If the fetch fails, check the FAQ section below before assigning this provider to any agent.

    Do this
    • Click Create Provider.
    • Confirm the new provider card appears in the provider list with the correct name, Local (Ollama) badge, model count, and base URI.
    • Click the pencil icon to reopen the provider, then click Fetch Models to confirm connectivity.
    • If the fetch succeeds, the provider is ready to use in agent configuration.
    Example

    After clicking Create Provider, the administrator sees a new card labeled Ollama Local - llama3.2 with a Local (Ollama) badge and a model count of 2. They click the pencil icon, click Fetch Models, and the same two models appear. Connectivity is confirmed.

    Best practice

    Always verify connectivity immediately after creating a provider. A provider that cannot reach its endpoint will cause every agent that references it to fail at runtime. If you rotate your Ollama configuration or change the host, return to this provider and run Fetch Models again to confirm the updated endpoint is reachable.

  6. Assign the Ollama provider to an agent

    With the provider saved and connectivity confirmed, the Ollama models in your catalog are now available in agent configuration. Navigate to Agents in the sidebar, open or create an agent, and create a new version. In the version's Config JSON, set the provider to reference your Ollama provider and the model to one of the models you included in the catalog.

    From this point, any workflow that uses that agent will call your local Ollama model at runtime. The workflow graph, the human review gates, the audit log, and every other DAVE feature work exactly as they do with any other provider type. The only difference is where the model inference runs: on your local machine or network, not on a third-party cloud AI service.

    Do this
    • Navigate to Agents in the sidebar.
    • Open an existing agent or click New Agent to create one.
    • Create a new agent version.
    • In the Config JSON, set the provider to your Ollama provider and the model to one of the models in its catalog.
    • Save the version and test it in a workflow to confirm the agent calls your Ollama instance correctly.
    Example

    The administrator opens an existing draft-generation agent, creates a new version, and sets the Config JSON to reference the Ollama Local - llama3.2 provider with model llama3.2. They save the version, run a test workflow instance, and confirm the agent produces output using the local model.

    Best practice

    Retire the previous agent version once you have confirmed the new Ollama-backed version works correctly. This ensures no new workflow instances can be assigned the old version by mistake. Because agent versions are immutable, the old version's configuration is preserved in the version history for reference even after it is retired.

What this guide covers

This guide explains how to connect a locally hosted Ollama model to DAVE as a provider, so that DAVE agents can call that model during workflow execution.

DAVE supports seven AI provider types: OpenAI, Anthropic, Azure OpenAI, Google AI, OpenAI-Compatible, Local (Ollama), and Custom. Local (Ollama) is the provider type for connecting to an Ollama instance running on your own machine or local network.

An important distinction: model hosting is not product hosting. Connecting Ollama to DAVE means you are hosting the AI model, not the platform. DAVE itself is fully managed and hosted by voolama LLC. There is no self-hosted, on-premises, or customer-operated deployment of DAVE. What you control with Ollama is where the model inference runs: on your hardware, not on a third-party AI cloud service. The DAVE platform, including its workflow engine, audit logs, credential vault, and human review system, continues to run on voolama's managed infrastructure regardless of which provider type you use.

This distinction matters for architecture decisions, compliance scope, and data flow design. When an agent calls an Ollama model, the prompt and response travel between DAVE's managed infrastructure and your local Ollama endpoint. Plan your network configuration accordingly.

When to use a local Ollama model

Ollama is the right provider choice in several situations:

  • Data sensitivity. If your workflow processes content that must not be sent to a third-party AI service, routing agent calls through a local Ollama model keeps inference on your infrastructure. Review your data flow end-to-end: the prompt originates in DAVE's managed infrastructure and travels to your Ollama endpoint, so network path and endpoint security are your responsibility.
  • Cost control. Local model inference has no per-token cost from an AI provider. If you are running high-volume workflows where third-party API costs are a concern, a local model can reduce that spend. You bear the infrastructure cost of running the model instead.
  • Model experimentation. Ollama makes it easy to run and switch between open-weight models locally. If you want to evaluate a model before committing to a production configuration, connecting it as an Ollama provider in DAVE lets you test it in real workflows without a third-party API account.
  • Restricted environments. If your development environment has restricted outbound internet access, a local Ollama instance may be the only AI option available.

Ollama is not the right choice when you need the highest-capability frontier models (available only through their respective cloud APIs), when your Ollama instance is not reliably reachable from DAVE's infrastructure, or when your team does not have the capacity to manage a local model runtime.

Before you start: what you need

Confirm the following before following the steps in this guide:

  • Ollama is installed and running. DAVE does not install or manage Ollama. You are responsible for installing, running, and maintaining your Ollama instance. Ollama must be running and reachable at a network address that DAVE's infrastructure can reach before you attempt to connect it.
  • At least one model is pulled in Ollama. DAVE fetches the model list from your Ollama instance. If no models are pulled, the fetch will return an empty list. Pull at least one model in Ollama before configuring the provider in DAVE.
  • The Ollama endpoint is network-accessible from DAVE. The default Ollama endpoint is http://localhost:11434/v1. This address is only reachable if DAVE's infrastructure and your Ollama instance are on the same machine or network. If they are not, configure your Ollama instance to listen on a network-accessible address and ensure any firewalls or routing rules allow the connection.
  • You have the Admin role in your DAVE workspace. Provider management is restricted to the Admin role. Only users with Admin access can create, edit, or delete providers.

Frequently asked questions

Does DAVE install or manage Ollama for me?

No. DAVE is a fully managed platform hosted by voolama LLC. Ollama is model-hosting software that you install and run on your own machine or network. DAVE connects to your Ollama instance as a provider but does not install, configure, or manage the Ollama runtime. You are responsible for keeping Ollama running and reachable.

Can I use the default base URI http://localhost:11434/v1?

Only if your Ollama instance is running on a host that DAVE's infrastructure can reach at that address. In most configurations, DAVE's managed infrastructure and your local machine are on different networks, so localhost from DAVE's perspective refers to voolama's servers, not your machine. You will need to configure Ollama to listen on a network-accessible address and use that address as the base URI in DAVE.

What if Fetch Models returns an empty list?

This usually means no models are pulled in your Ollama instance, or the base URI or API key is incorrect. Confirm Ollama is running, that at least one model is pulled, and that the base URI in DAVE matches your Ollama endpoint exactly. You can also switch to JSON mode and enter the model list manually using the format [{ "id": "llama3.2", "name": "Llama 3.2" }].

Can I have multiple Ollama providers pointing at different models or hosts?

Yes. Create a separate provider record for each Ollama instance or model configuration you want to use. Each provider has its own base URI, API key, and model catalog. Agents reference a specific provider, so you can route different agents to different Ollama instances without changing workflow logic.

Is the Ollama API key stored securely?

Yes. Any value entered in the API Key field is stored in DAVE's per-tenant credential vault, encrypted with AES-256-GCM authenticated encryption. The key is never returned through the API once stored and is never logged in plain text. If you leave the field blank (the default for Ollama with no authentication), nothing is stored.

What happens if my Ollama instance goes offline?

Any workflow instance that reaches an Agent Interaction node referencing your Ollama provider will fail or pause. DAVE cannot call a provider whose endpoint is unreachable. Ensure your Ollama instance is running and reachable before executing workflows that depend on it. If the instance goes offline during a running workflow, the instance may fail and generate a task for human review and intervention.