Local Model Readiness Checklist for DAVE

In DAVE, a Local (Ollama) provider connects DAVE's hosted platform to an Ollama model server. Ollama is an open-source tool that lets you run large language models on your own hardware. DAVE treats the Ollama server as one of its seven supported AI provider types, alongside OpenAI, Anthropic, Azure OpenAI, Google AI, OpenAI-Compatible, and Custom.

The key distinction to hold clearly: DAVE is the platform, and it is fully managed and hosted by voolama LLC. You do not run DAVE on your own infrastructure. What you run locally is the Ollama model server, which DAVE's hosted infrastructure calls over the network using the base URI you configure in the provider record. The default base URI for a Local (Ollama) provider is http://localhost:11434/v1, but because DAVE is hosted remotely, localhost in that URI refers to the machine running the Ollama server, not to DAVE's servers. For DAVE to reach your Ollama server, the server must be accessible at a network address that DAVE's infrastructure can route to.

This is the most common point of confusion when setting up a Local (Ollama) provider, and it is the first thing to resolve before creating the provider record in DAVE. If your Ollama server is running on a machine that is only reachable on a private network, you will need to expose it at a publicly reachable address or use a tunneling solution before DAVE can connect to it.

A Local (Ollama) provider is the right choice in these scenarios:

• Data sensitivity requirements. Some workflows process data that must not leave a controlled environment. Running the model on your own hardware means the prompt and response never travel to a third-party AI service's infrastructure. DAVE still processes the workflow orchestration, but the AI inference happens on your server.

• Cost control for high-volume, lower-complexity tasks. Cloud AI APIs charge per token. For workflows that run frequently and use models where a smaller open-weight model is sufficient, running that model locally can reduce per-run costs significantly.

• Specific model availability. Some open-weight models are available through Ollama but not through any of the cloud provider APIs. If a workflow requires a specific model that is only available as an Ollama-compatible model, a Local (Ollama) provider is the only way to use it in DAVE.

• Mixed-provider architectures. DAVE supports multiple provider records simultaneously. You can run some agents on OpenAI or Anthropic for tasks that benefit from frontier model capability, while routing other agents through a Local (Ollama) provider for tasks where a smaller model is sufficient. Each agent independently selects its provider and model.

A Local (Ollama) provider is not the right choice if your Ollama server cannot be made reachable from DAVE's hosted infrastructure, or if the latency of routing requests from DAVE's servers to your Ollama server is unacceptable for your workflow's performance requirements.

Every item in this section must be confirmed before you create the provider record in DAVE. Creating the provider record before these prerequisites are met will result in a failed Fetch Models attempt, which tells you nothing you could not have confirmed in advance.

1. Ollama is installed and running on your model server. The Ollama process must be active and listening on its API port (default 11434) on the machine you intend to use as the model server. Confirm this by running ollama list on that machine and verifying the command returns without error.

2. At least one model is pulled and available. Run ollama list on the model server and confirm at least one model appears in the output. If the list is empty, pull the model you need with ollama pull <model-name> before proceeding. DAVE's Fetch Models function retrieves the list of models currently available on the Ollama server: if no models are pulled, the catalog will be empty after the fetch.

3. The Ollama server is reachable from DAVE's hosted infrastructure. This is the critical network prerequisite. DAVE is hosted by voolama LLC and calls your Ollama server over the network using the base URI you configure. The server must be accessible at a network address that DAVE's infrastructure can reach. Options include: a server with a public IP address and port 11434 open in the firewall; a reverse proxy that exposes the Ollama API at a public HTTPS endpoint; or a tunneling solution such as ngrok or Cloudflare Tunnel that creates a publicly reachable URL for a locally running server. Confirm reachability by making an HTTP GET request to <your-base-uri>/api/tags from a machine outside your local network and verifying you receive a JSON response listing your models.

4. You hold the Admin role in DAVE. Provider management is restricted to the Admin role. Only Admins can view, create, edit, or delete providers at Admin, then API Management, then Providers. If you do not have the Admin role, ask an existing Admin to assign it to your account.

5. You have the correct base URI for your Ollama server. This is the network address at which DAVE will reach your Ollama server. It must be a URL that DAVE's hosted infrastructure can resolve and connect to. If you are using the default Ollama port, the URI will end in :11434/v1. If you are using a reverse proxy or tunnel, use the HTTPS URL provided by that service. Write this URI down before opening DAVE: you will paste it into the provider form.

Work through these items in order. Each item must be confirmed before moving to the next.

Ollama server readiness

• Ollama is installed on the model server machine.
• The Ollama process is running and listening on port 11434 (or your configured port).
• At least one model is pulled: ollama list returns one or more model entries.
• The model you intend to use in DAVE is confirmed in the ollama list output.

Network reachability

• The Ollama server is accessible at a network address reachable from DAVE's hosted infrastructure (not only on a private LAN).
• A GET request to <your-base-uri>/api/tags from outside your local network returns a JSON response listing your models.
• If using a reverse proxy or tunnel: the proxy or tunnel is running and the public URL is confirmed.
• If using a firewall: port 11434 (or your configured port) is open for inbound connections from external addresses.
• The base URI you will enter in DAVE is written down and confirmed reachable.

DAVE provider setup

• You are logged in to DAVE with the Admin role.
• You have navigated to Admin, then API Management, then Providers.
• You have clicked Add Provider and the create form is open.
• Provider Name entered: descriptive and environment-specific (e.g., "Ollama Local llama3.2" or "Ollama Production mistral").
• Provider Type set to Local (Ollama).
• Base URI field updated to your confirmed reachable address (replacing the default http://localhost:11434/v1 if your server is not on localhost relative to DAVE).
• API Key field: left blank if your Ollama server does not require authentication, or populated if you have configured Ollama to require a key.

Model catalog and connectivity verification

• Fetch Models clicked and returned a non-empty model list.
• The model you intend to use is visible in the fetched list and has been selected.
• Create Provider clicked and the provider card appears in the provider list.
• Provider card expanded with the chevron: model catalog is populated correctly.

Agent and workflow verification

• At least one agent has been configured to use the new Local (Ollama) provider and the target model.
• A test workflow instance has been run that exercises the Ollama-backed agent.
• The test instance completed successfully with the expected output.
• Latency of the Ollama agent call is acceptable for the workflow's performance requirements.

Ongoing readiness

• A process is in place to monitor the Ollama server's availability (process health, port reachability).
• A process is in place to pull model updates when new versions are released.
• If using a tunnel or reverse proxy: a process is in place to restart it automatically if it goes down.
• The DAVE provider record will be updated if the Ollama server's base URI changes.

Does running a local model mean I am self-hosting DAVE?
No. DAVE is fully managed and hosted by voolama LLC. There is no self-hosted DAVE option. What you are running locally is the Ollama model server, which provides the AI inference endpoint. DAVE's platform calls that server over the network using the base URI you configure. The model hosting is yours; the platform hosting is voolama's.

Why does Fetch Models fail when I use the default base URI?
The default base URI for a Local (Ollama) provider is http://localhost:11434/v1. Because DAVE is hosted remotely, localhost in that URI refers to DAVE's own servers, not your Ollama machine. DAVE cannot reach your Ollama server at localhost. Replace the base URI with the network address at which your Ollama server is publicly reachable before clicking Fetch Models.

Does my Ollama server need to be publicly accessible on the internet?
It needs to be reachable from DAVE's hosted infrastructure. In practice, this means it must have a network address that DAVE's servers can route to. A server on a private LAN with no external access will not work without a tunneling solution or a reverse proxy that exposes it at a public address. A server with a public IP and an open firewall port will work directly.

Do I need an API key for a Local (Ollama) provider?
By default, Ollama does not require an API key. The API Key field in the DAVE provider form can be left blank for a default Ollama installation. If you have configured Ollama to require authentication (using the OLLAMA_API_KEY environment variable or a reverse proxy that adds authentication), enter the required key in the API Key field. The key is stored using AES-256-GCM authenticated encryption and is never returned through the API after saving.

What happens if my Ollama server goes down while a workflow is running?
If the Ollama server is unreachable when an agent tries to call it, the workflow instance may fail or pause at that node. A task may be created for human review and intervention. Restoring the Ollama server and restarting the instance (or completing the human review task) will allow the workflow to continue. This is the same behaviour as any provider outage in DAVE.

Can I use multiple Local (Ollama) providers pointing at different servers?
Yes. DAVE supports multiple provider records of the same type. You can create separate Local (Ollama) provider records for different Ollama servers, each with a different base URI and model catalog. Individual agents can then be configured to use whichever server is appropriate for their task.

Can I run more than one model on the same Ollama server?
Yes. Ollama supports multiple pulled models on a single server. When you click Fetch Models in DAVE, all models currently pulled on the server appear in the catalog. You can select all of them or only the ones your agents will use. Each agent independently selects which model from the provider's catalog it will call.

Who can create or edit a Local (Ollama) provider in DAVE?
Only users with the Admin role. The Providers page at Admin, then API Management, then Providers is not accessible to any other role.