AI Providers
Setup Guide — Connecting AI Services to the OnCoor AI Module
Module: AI Module › Configuration | Last updated: July 2026
Contents
- Overview
- Key Concepts: Providers vs. Models
- Finding Your Way Around
- The Providers That Come Built In
- Turning On a Provider
- Adding Models to a Provider
- Choosing Your Default AI
- Quick Reference
1. Overview
The AI Module needs an AI service to do its thinking — matching similar records, answering questions, making suggestions. OnCoor doesn't have that intelligence built into itself; instead it connects to an AI service, called a provider. This page is where you choose and switch on the providers you want to use.
The good news: OnCoor arrives ready to use. Out of the box it's set up with a local, private AI that runs on your own server, so you can start using AI features straight away — no sign-ups, no accounts, and nothing to pay per use. If you later decide you want a more powerful AI, you can connect a cloud service in a few clicks.
What you can do here
- Use the built-in local AI with zero setup and zero cost.
- Connect cloud AI services — such as OpenAI, Groq, Google Gemini, or Azure OpenAI — when you want more power.
- Choose which AI handles each job, and switch it whenever you like.
- Keep your data on your own systems by default — nothing leaves your network unless you choose a cloud provider.
WHERE TO FIND IT — AI providers live under Admin → AI → Configuration, split across two tabs: Providers and Models.
2. Key Concepts: Providers vs. Models
Setting up AI comes down to two things. Understanding the difference makes everything else straightforward.
| Provider | Model | |
|---|---|---|
| What it is | The AI service OnCoor connects to | A specific AI "brain" within that service that does one kind of job |
| Think of it as | The company or service | The particular tool that service offers |
| Example | Ollama (on your own server), OpenAI | nomic-embed-text for matching, llama3.1 for answers |
| You set it up in | The Providers tab | The Models tab |
Rule of thumb: a provider is the service; a model is the specific capability you use from it. AI features only work when you have at least one provider switched on with at least one model switched on underneath it.
Models come in two kinds, depending on the job:
- Matching models compare records to find ones that mean the same thing. This is what powers Cognitive De-Duplication. (You may see these called "embedding" models on screen.)
- Answer models write responses in plain language. These power chat-style features and AI suggestions. (You may see these called "completion" models on screen.)
3. Finding Your Way Around
3.1 The Providers tab
Go to Admin → AI → Configuration → Providers. This lists every AI service OnCoor already knows about, with a status showing which ones are switched on. From here you can switch a provider on or off, or add a new one.
3.2 The Models tab
Go to Admin → AI → Configuration → Models. This lists the specific models available under each provider, what each one is used for, and which is the default for its job.
4. The Providers That Come Built In
So you don't have to set everything up from scratch, OnCoor already knows about several popular AI services. Only the local one is switched on to begin with — the rest are ready and waiting for you to add your account details and switch them on.
| Provider | What it is | Where it runs | Cost | On by default? |
|---|---|---|---|---|
| Ollama | Free, open-source AI you host yourself | Your own server or network | Free | ✅ Yes |
| OpenAI | The widely used commercial AI (maker of ChatGPT) | The cloud | Paid | No |
| Groq | A very fast cloud AI service | The cloud | Free tier available | No |
| Google Gemini | Google's AI service | The cloud | Free tier / paid | No |
| Azure OpenAI | OpenAI's models delivered through Microsoft Azure | Your Azure cloud | Paid | No |
WHY LOCAL IS THE DEFAULT — Ollama runs on your own hardware, so your data never leaves your network and there's nothing to pay for each use. It's the recommended starting point. You only need a cloud provider if a particular job calls for a more powerful model than your local one.
NOTE — Google Gemini may appear twice in the list. That's expected — OnCoor uses one entry for written answers and the other for record-matching. When you set up a Gemini model, simply switch on the entry that matches what the model is for; the setup screens guide you.
5. Turning On a Provider
The local provider is already on, so you only need this section when you want to connect a cloud service.
Steps
- Create an account with the provider (for example, OpenAI) and get an API key from them. An API key is like a password that lets OnCoor use your account with that service.
- Give the key to your system administrator to store securely (see the note below).
- Go to Admin → AI → Configuration → Providers.
- Click the provider you want, then Edit.
- Check the details are filled in. For the built-in providers, they already are — you normally just confirm them.
- Set Active to Yes, then Save.
Once the provider is on, switch on at least one of its models — see Adding Models to a Provider.
Here's what the fields on the provider screen mean, in plain terms:
| Field | What it means |
|---|---|
| Provider Code | A short name that identifies the service (for example, OPENAI). Set once when the provider is created and can't be changed afterward. |
| Provider Name | The friendly name shown in lists (for example, "OpenAI"). |
| Endpoint URL | The web address OnCoor uses to reach the service. Already filled in for the built-in providers — leave it as it is unless your administrator tells you otherwise. |
| Auth Type | How OnCoor proves it's allowed to use the service. Already set correctly for the built-in providers — you don't need to change it. |
| Secret Reference Name | Points to where your API key is safely stored — it is not the key itself. |
| Active | Whether the provider is switched on. |
KEEPING YOUR KEYS SAFE — Your actual API key is never typed into or saved on this screen. The Secret Reference Name only tells OnCoor where to find the key, which is kept somewhere secure. Setting up the key itself is a one-time job your system administrator handles.
6. Adding Models to a Provider
A provider doesn't do anything on its own until it has at least one model switched on. The model is the specific AI capability that does the actual work.
Remember the two kinds:
| Model type | What it's for | Example |
|---|---|---|
| Matching (embedding) | Comparing records to find duplicates — powers Cognitive De-Duplication | nomic-embed-text |
| Answers (completion) | Writing responses in plain language — powers chat and AI suggestions | llama3.1 |
Steps
- Go to Admin → AI → Configuration → Models.
- Click Add Model (or Edit an existing one).
- Fill in the fields (see the table below). At a minimum you need the Provider, Purpose, Model Code, and Model Name.
- Set Active to Yes, then Save.
Here's what each field on the model screen means, in plain terms:
| Field | What it means |
|---|---|
| Provider | The AI service this model belongs to. Pick it from the list of providers you've switched on. |
| Purpose | What the model is for: Embedding (matching records), Completion (writing answers), or Reranking (putting results in best-match order). |
| Model Code | A short name you give the model inside OnCoor to refer to it (for example, EMB_OPENAI_LARGE). Set once when you create it and can't be changed later. |
| Model Name | The model's real name as the provider calls it (for example, nomic-embed-text). This is the one OnCoor actually requests, so it must match the provider's spelling exactly. |
| Deployment Name | Only needed for Azure OpenAI — enter the deployment name from your Azure setup. Leave blank for every other provider. |
| Embedding Dimension | For matching (embedding) models only — a number that describes the size of the comparison data the model produces. Leave blank for answer models. If you're unsure of the value, the provider's documentation lists it. |
| Max Tokens | The most text the model will handle in one request, counted in "tokens" (roughly, pieces of words). A higher number allows longer input and longer answers. Optional — leave blank to use the model's own limit. |
| Temperature | How predictable versus creative the answers are, on a scale that starts at 0. Lower values (like 0) give consistent, focused results; higher values give more varied ones. Most providers accept values up to 2, though the exact maximum depends on the provider you're using. For data-quality work, keep it low — 0 is a good default. Optional. |
| Input Rate / Output Rate | What the provider charges per million units of text — for what you send in (input) and what it sends back (output). OnCoor uses these only to estimate cost in its usage reports. Leave at zero for free or local models. |
| Default Model | Whether this is the go-to model for its job (see the next section). |
| Active | Whether the model is switched on. |
OUT OF THE BOX — OnCoor comes with two local models already set up and switched on: one for record-matching and one for written answers. You only need to add models when you're connecting a new provider.
NOTE FOR LOCAL AI — Before a local (Ollama) model can be used, it also has to be downloaded onto your server. Your administrator does this once during setup.
7. Choosing Your Default AI
For each job — matching and answers — one model is marked as the default. The default is simply the model OnCoor uses automatically unless it's told to use a different one.
To switch the default to a different model:
- Switch on the new model (see the previous section).
- Mark it as the Default for its job.
- Remove Default from the old model, so only one remains.
ONE DEFAULT PER JOB — You can have several models switched on at once, but only one can be the default for each job: one default for matching, and one default for answers.
8. Quick Reference
A fast lookup for the most common actions.
| I want to… | Do this |
|---|---|
| Start using AI with no setup | Nothing — the local AI is already on |
| See which AI services are available | Admin → AI → Configuration → Providers |
| Connect a cloud AI service | Providers → click the provider → Edit → set Active = Yes → Save |
| Add a model to a provider | Models → Add Model → choose provider, name, and type → Active = Yes → Save |
| Change which AI handles a job | Models → switch on the new model → mark it Default → remove Default from the old one |
| Keep all my data on my own systems | Use the built-in local (Ollama) provider — it's the default |
| Find out where my API key goes | It's stored securely outside this screen — your administrator sets it up |
Source: OnCoor AI Module product documentation, written for end users. For the latest screens and options, always refer to the in-app interface.