Skip to main content

Connect Make with HTTP v4

Use Make's HTTP → Make a request, Version 4, to send a text prompt to Primex Chat Completions. Choose a compatible model in Make Setup.

Use the HTTP module to call Primex directly. The JSON below is request data; apply the module settings manually rather than importing it as a blueprint.

Create a scenario and keychain​

Create a new scenario, add HTTP → Make a request, and choose Version 4. Keep scheduling off while you test a single request. Existing HTTP (legacy) v3 modules have different settings; use v4 for this guide.

Create a dedicated Primex API key. In the HTTP module:

  1. Set Authentication type to API key.
  2. Under Credentials, add a keychain with a descriptive name such as Primex.
  3. Set Key to Bearer YOUR_API_KEY, including the space after Bearer.
  4. Set API key placement to In the header.
  5. Set API key parameter name to Authorization, then create and select the keychain.

Keep the key in the keychain. The JSON body and URL contain no secret, and you do not need a second Authorization entry in the ordinary Headers section.

Configure the request​

FieldValue
URLhttps://gateway.primexos.com/v1/chat/completions
MethodPOST
Body content typeapplication/JSON
Body input methodData structure
Parse responseYes
Pagination typeLeave empty
Return error if HTTP request failsYes
Timeout60 seconds
Allow redirectsNo

Under Body structure, click Add and use the data structure generator with this sample. Replace the model placeholder with your exact selected ID; Setup copies the same body with that ID already filled in:

JSON
{
"model": "YOUR_LIVE_CHAT_MODEL",
"messages": [
{
"role": "user",
"content": "Hello from Primex"
}
],
"stream": false
}

Save the structure, then fill Body content to match the sample: model is text, messages is an array with one collection, role and content are text, and stream is the Boolean false. Generating a structure establishes field types; confirm and enter the values as well.

For dynamic prompts, map the preceding module's text into messages' content field. Data structure input escapes quotes, newlines and other JSON characters. Avoid inserting unescaped dynamic text into a JSON string. If an upstream module produces multiple bundles, each bundle can cause a separate model request.

Run once and map the answer​

Save and click Run once. Inspect the HTTP module's output bundle before connecting any downstream action. With Parse response enabled, Make exposes mappable response fields after the first run. Expect HTTP 200 and a nonempty text answer at:

Data → choices → first item → message → content

The underlying Chat Completions JSON path is choices[0].message.content. Select the first choice in Make's mapping interface rather than assuming that its array-index conventions match JSON notation. The response's model and usage describe the model result and reported usage; they are not a price quote.

Inspect response headers for x-primex-request-id when available. Match the request ID, dedicated key and run time with Primex Usage. Make's scenario history and credits are separate from Primex's usage and charges.

Before a downstream action, add a filter requiring a successful status and a present, nonempty text answer. A filter that rejects a bundle stops processing that bundle; it does not raise an error or create a Primex refund. Do not treat filtered-out or tool-only replies as completed text tasks. Inspect unexpected response types and whitespace-only answers before relying on an automated action; a simple Exists condition alone is insufficient validation.

Failures, retries and data​

For the initial manual test, leave Store incomplete executions disabled in Scenario settings and do not add a Retry error handler. Enabling incomplete execution storage can allow Make to automatically retry supported errors such as ConnectionError and RateLimitError. Review the actual scenario settings and history before enabling schedules or automatic recovery. This configuration review does not establish the platform's complete retry behavior.

SymptomNext action
401Check the selected keychain, active key and Bearer prefix.
404Check the full endpoint and exact live model ID from Setup.
402Inspect Primex credit balance and key spend controls before another run.
429Identify the limit and reduce request volume; waiting does not reset a monthly spend cap.
3xx redirectLeave redirects disabled and investigate the URL; a redirect is not a model answer.
Empty, missing or tool-only answerInspect the response body and selected model; this path handles ordinary text.
Timeout or connection failureCheck Make history and Primex Usage before retrying. The first request may already have completed.

HTTP v4 documents a timeout range of 1–300 seconds; the example uses 60. A later filter or downstream module failure does not reverse a completed Primex request. Review Errors & limits and Usage.

Make processes your prompt and response, and scenario history may retain them. Review its workspace access, storage and sharing settings separately from Primex's privacy policy. A blueprint includes module settings and mapped values, even though connections need to be configured separately after import. Review an export before sharing it, especially values entered directly into module parameters. This guide's copied JSON contains only request data.

To disconnect, stop the scenario, remove its keychain selection and unused keychain as appropriate, and revoke the dedicated Primex key.

Need help connecting? Report a setup issue.