Groq integration – Appmixer connector library

Integration with Groq's API for multimodal content analysis, chat completions, audio transcription, translation, and more.

Changelog

3 versions

v2.0.1 Latest

Fixed a crash when a Find/List component saved an EMPTY result set to a CSV file — the header row was built from the first record without checking there was one, so a search that legitimately matched nothing failed the flow instead of writing an empty file.

v2.0.0

BREAKING: renamed the MakeAPICall component to MakeApiCall. Existing flows keep running on the old component but it is no longer offered in the designer — replace it with Make API Call and re-map the inputs. The inputs changed too: Headers and Query Parameters are now key-value fields and the endpoint path is pinned to https://api.groq.com.
BREAKING: Send Prompt now emits 'answer' (plus 'prompt' and the response metadata) instead of the raw 'choices.[0].message.content' API path, matching every other AI connector. Re-bind downstream fields that read the model reply.
Send Prompt: relabelled 'Search Prompt' to 'Prompt' and 'Model ID' to 'Model', and reordered the inspector to Prompt, System Instructions, Model, Conversation ID, Advanced.
List Models: added the standard Output Type input and a dynamic output port schema, so the variable picker is populated instead of falling back to Raw Output.
Create Transcription and Create Translation: the Model input is now a typeahead backed by List Models instead of a hardcoded whisper-* list, so newly released Groq models are reachable.
Added lib.js with the shared getBaseUrl/request/resolveApiUrl and array-output helpers; every component now goes through it.
Replaced the JPEG component icons with an SVG and added examples to the output port options.
List Models: the inspector model dropdowns now pass isSource, so the model list is cached per account (context.config.listCacheTTL, default 120 s) and a failing lookup returns an empty list instead of an error popup; the item schema is exported as ITEM_SCHEMA for the offline tooling.
Create Transcription and Create Translation: the file read stream is closed when the upload fails, and a null Temperature no longer crashes the request.