New on the blog: why AI belongs in the feed →
View Categories

Provider Profiles

Provider profiles require Premium or higher. They appear at the bottom of GravityOps -> AI Feed -> Credentials.

A profile is a named API key with its own settings. Use extra profiles to:

  • keep separate Anthropic or OpenAI keys for different clients or projects on one site;
  • send Anthropic or OpenAI requests through a gateway or proxy;
  • connect another service that uses the OpenAI Chat Completions format, such as OpenRouter or a local Ollama server.

The keys saved in the Provider credentials box at the top of the tab are the default profiles for Anthropic and OpenAI. They work on every plan and are listed in the same table.

Add A Profile #

  1. Select Add Anthropic (Claude) profile, Add OpenAI (ChatGPT) profile, or Add OpenAI-compatible provider profile. A profile keeps the type it was created with.
  2. Fill in the form and select Add Profile.
  3. Select Test in the profile’s row.

Profile Settings #

Name is shown wherever a feed chooses a key or a model.

API Key is encrypted before it is stored and is never shown again. Leave it blank when editing to keep the saved key. Only an OpenAI-compatible profile may be saved without a key, for a local server that needs none.

Base URL depends on the type:

  • For Anthropic and OpenAI profiles it is optional. Leave it blank to use the provider’s own address. To use a gateway, enter its address without the /v1 part.
  • For an OpenAI-compatible profile it is required. Enter the address that ends just before /chat/completions, including the version, such as https://openrouter.ai/api/v1.

Organization ID and Project ID (OpenAI profiles) are sent as the OpenAI-Organization and OpenAI-Project headers.

Extra Headers are sent with every request from the profile, one Name: value per line. Their values are encrypted and never shown again; the form lists only the saved header names. Entering anything replaces all saved headers. Headers that carry the key or describe the request, such as Authorization, x-api-key and Content-Type, cannot be set.

Models (OpenAI-compatible profiles) is the list feeds choose from, one model ID per line, written exactly as the provider names it. The first model is used by the connection test.

Status disables a profile without deleting it. Feeds cannot use a disabled profile.

Capabilities Of An OpenAI-Compatible Profile #

Services that use this format do not all support the same features, and the plugin cannot detect them reliably. Tick only what the provider supports. The connection test checks each one you tick.

Capability What it allows Without it
Streaming The “show result while generated” button shows text as it arrives. That button shows the whole answer at once.
Structured output A feed can fill several fields from one response, using a JSON schema. See JSON mode.
JSON mode A feed can fill several fields. The provider guarantees JSON but not its exact shape, so results are less reliable than structured output. With neither, the feed can use Single target field only.
Reports token usage The ledger shows input and output tokens, and token caps count them. The ledger shows usage as unavailable. Token caps cannot count this profile’s requests.

Effort, Thinking and Speed are not offered for OpenAI-compatible profiles. The provider’s defaults are used.

Base URL Safety #

A base URL is requested by your server, so it must use HTTPS and must not point to a private, loopback or link-local address. This prevents a profile from being used to reach services inside your network. The address is checked when the profile is saved and again on every request, and redirects from a custom base URL are not followed.

To use a server on your own network, such as Ollama at http://localhost:11434/v1, add this to wp-config.php:

define( 'GOAIF_ALLOW_PRIVATE_PROVIDER_HOSTS', true );

With it, plain HTTP and private addresses are allowed for every profile on the site. Use it only on sites where every administrator who can edit profiles is trusted with that access.

Use A Profile In A Feed #

  • For an Anthropic or OpenAI profile, choose it in the feed’s API Key setting. The setting appears when the provider has more than one usable key.
  • For an OpenAI-compatible profile, set Provider to OpenAI-compatible provider and choose a model. Models are grouped by profile, so the model choice also selects the profile.

See Text Feed Settings for the fallback model.

What Differs Between Compatible Providers #

  • No usage reported. The ledger records usage as unavailable instead of zero. Turn off Reports token usage if the connection test says the provider does not report it.
  • No request ID. The ledger stores the response ID when the provider sends no request ID header, and nothing when it sends neither.
  • Errors sent with a success status. Some services answer with HTTP 200 and an error object. The plugin treats that as a failure.
  • Streams that end without a final marker. A stream is accepted as complete when the provider sent a finish reason, even without the closing [DONE] line. A stream with neither is reported as ended before completion.

Deleting A Profile #

Deleting a profile removes its key and settings. Feeds that used it keep their settings but fail until they are given another key or model. Feeds are not changed automatically.

Plans #

Extra profiles, OpenAI-compatible providers and the fallback model require Premium or higher. If a site no longer has that plan, feeds that use them are paused and listed in the paused-feeds notice; their settings are kept.

Stuck? Ask in GravityOps Help #

Search past answers or post your question, tagged "ai-feed".