AI model catalogue
Website model records let Powerfront offer the AI models your accounts can use, with catalogue information to help selection.
What it is
Models lists the AI models prepared for your website. AI accounts select from these records. A separate model catalogue provides capability, token-limit and price information to help Powerfront choose a model.
Who uses it and why
Administrators agree which models their AI features should use. Developers identify requirements such as tool use, images or structured replies. Powerfront manages the model records and checks the connection for you.
Where to find it
Powerfront manages Settings › AI › Models and selects a model for an account in Settings › AI › Accounts. These are Powerfront-managed screens.
How it works
- Powerfront chooses a Provider and a Model, then gives the record a Name.
- The model becomes available to accounts on that website.
- Powerfront assigns it to the required account. Features using that account then use its model connection.
The catalogue can describe image input, function calling, reasoning, image output and token limits, with prices where known. The Model selector can display input and output prices beside a model identifier. Other (custom) allows a model identifier that is absent from the catalogue.
Configure it
Ask Powerfront to add or change a model for your website. The provider choices in the model editor are OpenAI, AWS Bedrock, Google AI Studio, Salesforce Models API and Workers AI (Cloudflare). A new record starts with OpenAI.
Give Powerfront the intended use and any required model identifier. The catalogue selection alone does not supply provider credentials or prove that your provider account can use that model.
Fields
| Field | Where | Required | What it is for |
|---|---|---|---|
| Name | Settings › AI › Models | Yes | Gives the website model a recognisable name. |
| Provider | Settings › AI › Models | Yes | Selects the model service; the new-record form starts with OpenAI. |
| Model | Settings › AI › Models | Yes | Selects a model from the provider's catalogue. |
| Model ID | Settings › AI › Models | No | Supplies a model identifier when Other (custom) is selected. |
Limits and good to know
- Website model records and the reference catalogue are separate. A catalogue refresh does not itself switch your account to another model.
- The catalogue becomes due for refresh after one hour. The first lookup waits for a refresh; later lookups can return cached information while a refresh runs. Fetches have a 30-second timeout.
- If a refresh fails, existing catalogue information or built-in entries remain available. This does not switch failed AI requests to another provider.
- Imported catalogue entries omit expired models, fine-tuned entries and ChatGPT-prefixed entries. This filtering does not delete existing website model records.
- Missing or zero-valued catalogue prices are treated as unknown. Catalogue prices support estimates, not a confirmed invoice.
- A website cannot add a second record with the same model identifier. Editing an existing record retains its original model identifier; ask Powerfront to add the replacement model separately.
Depends on and works with
+5 more · used by
- AWS Bedrock model connectionConfigured by
The account selects a website model configured for AWS Bedrock.
- Cloudflare AI Gateway connectionConfigured by
The selected model can determine the effective upstream provider.
- AI usage and estimated costConfigured by
Model catalogue pricing determines whether token cost can be estimated and which rates apply.
- AI platform execution limitsConfigured by
Model context limits affect the available token budget.
- Salesforce Models API connectionConfigured by
Model configuration selects the Salesforce Models provider and model.
View all connections as a list
This feature uses
- AI model provider selectionRelated
Review provider compatibility when choosing a model from the website catalogue.
Used by
- AI feature settings and change historyConfigured by
For accounts that use a model, model setup determines the available AI capabilities.
- Context Compiler for AI conversationsConfigured by
Model input limits determine the adaptive budget when no override is supplied.
- AI assistance activityEnhances
Known model pricing allows estimated cost calculation.
- Images and attachments for AI agentsRequires
Image input requires a model that supports it.
- AI accounts and shared defaultsConfigured by
AiFlow accounts select a website model and use its provider and context limit.
- AI model provider selectionConfigured by
AiFlow uses the provider recorded against the selected website model.
- OpenAI API connection optionsConfigured by
The account selects a website model with an OpenAI provider.
- Google AI Studio model connectionConfigured by
Powerfront prepares a model with Google AI Studio as its provider.
- AWS Bedrock model connectionConfigured by
The account selects a website model configured for AWS Bedrock.
- Cloudflare AI Gateway connectionConfigured by
The selected model can determine the effective upstream provider.
- AI usage and estimated costConfigured by
Model catalogue pricing determines whether token cost can be estimated and which rates apply.
- AI platform execution limitsConfigured by
Model context limits affect the available token budget.
- Salesforce Models API connectionConfigured by
Model configuration selects the Salesforce Models provider and model.
FAQ and troubleshooting
Why is a model missing from my account's choices?
Your account selects website model records, rather than the entire reference catalogue. Ask Powerfront to add the required model.
Can Powerfront add a model absent from the list?
The editor offers Other (custom) and Model ID. Powerfront still needs to check provider compatibility and access.
Does a listed capability mean every AI feature supports it?
No. The selected provider connection and AI feature must also support the required input or operation.
Will a catalogue outage stop existing models working?
Catalogue refresh failure retains cached or built-in information. Whether an AI request succeeds depends on its configured model service.