• Introduction
• The base URL
• Create the connected service
• Create an agent model
• Use the model in automations
  • Reasoning
  • Streaming
• Using a local Ollama server
• Resources

Introduction

Ollama Cloud runs Ollama's models as a hosted service, so you reach them with an API key instead of running a server yourself. The same Cerb provider also talks to a local Ollama install – see Using a local Ollama server at the end.

Cerb speaks Ollama's own native API, not the OpenAI-compatible one. That single fact drives most of this page: it decides the base URL you enter, and it's why a base that works with other tools may not work here.

You'll need an Ollama account and an API key from it.

The base URL

Give Cerb the bare host, and nothing more:

https://ollama.com

Create the connected service

Ollama ships no package in the Library tab, so the service that holds your key is built by hand. It's a Token Bearer service with an account under it – two records, created from two different lists.

First the service:

1. Navigate to Search » Connected Services.

2. Click the (+) icon in the top right of the list.

3. Select the Build tab, beside Library in the strip at the top of the dialog.

4. Set Type to Token Bearer. The list also offers API Key, which is a different mechanism – this one is the Authorization header.

5. Token Name is already filled in with Bearer, which is what Ollama wants. Leave it alone. The field exists for services whose header uses some other word in that position.

6. Give the service a Name and a URI, then click Save Changes. The URI is the handle automations use to refer to the record, and it stays blank until you type it – it doesn't derive itself from the name the way the Library route's Alias does.

Then the account that holds the key:

1. Navigate to Search » Connected Accounts.

2. Click the (+) icon. The dialog asks which service to connect to – accounts are created service-first, so pick the Token Bearer service you just built rather than looking for an add account button on the service itself.

3. Give the account its own Name and URI. Neither fills itself in here either.

4. Paste your Ollama Cloud API key into Token, then click Save Changes. It's a text area rather than a single-line field, so a long key won't look the way it does elsewhere.

Cerb then sends Authorization: Bearer <your key> on every request, substituting Token Name for the word Bearer if you ever change it.

Create an agent model

An agent model record holds one model's configuration – its provider, endpoint and credentials – so automations reference it by name instead of repeating a provider block.

The fields below are in the order the form presents them. API endpoint URL comes before Model because it feeds both Refresh and Test.

1. Navigate to Search » Agent Models.

2. Click the (+) icon in the top right of the list.

3. Set Provider to Ollama.

4. Set API endpoint URL to https://ollama.com. This field is required – see above.

5. Set Authentication to the connected account you created above.

6. Click the refresh button beside Model to load the model list, and pick one. Refreshing requires an administrator; a non-admin gets a permission error rather than an empty list.

7. Give the record a short Name – this is what automations will use. Colons aren't allowed, since the name is referenced as cerb:agent_model:<name>.

8. Click Test to verify the connection, then Create.

Name the record after you pick the model, not before. Picking from the list rewrites Name with a sanitized version of the model id every time, including on a record you already named.

The capability fields are yours to fill in. Cerb keeps no table for this provider and reads no metadata from it, so Context window, Vision, Thinking and the four Ratings stay exactly as you set them – including for a model id beginning gpt-, which on some other providers would pick up OpenAI's own figures. The context window matters most, since compaction ratios are fractions of it.

Give the refresh a moment to land before you open the Model menu. Opening it too early shows the shipped suggestions rather than the list that was fetched, and the menu keeps showing them until you close and reopen it. Those are real model ids, so there's nothing on screen to tell you apart from a list that loaded.

Use the model in automations

Reference the model by the name you gave the record.

• automation
• policy

• start:
    llm.chat/summarize:
      output: results
      inputs:
        model: ollama-cloud
        messages:
          0:
            role: user
            content: Summarize this conversation in one sentence.
    return:
      summary@key: results:content

• commands:
    llm.chat:
      allow@bool: yes

The same record works with llm.agent: for tool-using conversations, and with llm.router: to pick between several models as data. An automation that doesn't name a model resolves a pool instead – a search across agent models rather than a named record. Omit a search entirely and you get every available model, in the priority order an admin set.

Reasoning

effort: works on this provider, but it's translated rather than passed through:

llm:
  ollama:
    model: llama3.2
    effort: medium

Ollama has no effort scale. It has a think parameter, and Cerb maps onto it:

Levels outside the scale clamp rather than fail – minimal becomes low, and xhigh or max become high. No extra_body: escape hatch is needed.

Streaming

This provider streams by default. Set stream@bool: to no if something between Cerb and Ollama buffers responses rather than passing them through.

Using a local Ollama server

The same provider reaches an Ollama install on your own machine. Nothing above changes except the base URL, and that a local server normally needs no key – in which case leave Authentication blank rather than attaching an account with an empty token.

Inside a container, localhost is the container rather than the machine, which is why a Dockerized Cerb needs the second form to reach a server running beside it. There's still no default – fill the field in either way.

Refresh means something different locally. /api/tags reports the models you've actually pulled, not a catalog, so pull a model before you reach for Refresh.

Resources

• Reference: Agent models
• Reference: Connected services
• Guide: Launch Cerb in Docker