Servoy TI/NG Client Web-Service for calling Groq through the OpenAI-compatible API.

This package exposes a typed, documented API under plugins.svyGroq.* (SPM installable).

In Servoy Developer:

1. Open SPM.
2. Click Add...
3. Enter this URL:

https://hv-office.de/changelog/wpm.json

After that, svyGroq will appear in the list and you can download/install it from there.

Important: The TI/NG Client runs in the browser.

• If you set plugins.svyGroq.apiKey / call setApiKey() in the client, that key is not secret.
• Recommended production architecture is Servoy Server (or your backend) → Groq, and the client calls your server.

For that architecture:

• Set plugins.svyGroq.baseUrl to your proxy endpoint.
• Call plugins.svyGroq.setProxyMode(true) (so no Authorization header is sent from the client).

// Direct mode (dev/testing only)
plugins.svyGroq.setApiKey(scopes.globals.GROQ_API_KEY);

var text = plugins.svyGroq.chatCompletionText(
  plugins.svyGroq.models.DEFAULT,
  [ plugins.svyGroq.user('Hello from Servoy') ],
  { temperature: 0.2 }
);

If your scripting context treats calls as Promises, use .then(...):

plugins.svyGroq.chatCompletionText(plugins.svyGroq.models.DEFAULT, [plugins.svyGroq.user('Hi')])
  .then(function(text) {
    application.output(text);
  })
  .catch(function(e) {
    application.output('Groq error: ' + e);
  });

• apiKey: string | null
  • Only used if sendAuthHeader=true.
• baseUrl: string
  • Default: https://api.groq.com/openai/v1
  • For proxy mode point to your server.
• defaultModel: string
  • Used when model parameter is empty.
• models: groqModels
  • Predefined aliases for convenience:
    • DEFAULT
    • LLAMA_3_3_70B_VERSATILE
    • LLAMA_3_1_8B_INSTANT
    • LLAMA3_70B_8192
    • LLAMA3_8B_8192
    • MIXTRAL_8X7B_32768
    • GEMMA2_9B_IT
• sendAuthHeader: boolean
  • Default true. In proxy mode set false.
• timeoutMs: number
  • Default 60000. Applies to non-streaming calls.
• retry: groqRetryConfig
  • Default retries on: 408, 429, 500, 502, 503, 504.
• debug: boolean
  • When true, writes debug logs to console and fills lastRequest.
• lastRequest: groqLastRequestInfo | null
  • Contains request telemetry (status, duration, requestId, ...).

• setApiKey(apiKey)
• setProxyMode(enabled)
  • true: client does not send Authorization header.
• chatCompletion(model, messages, options) → full JSON response
• chatCompletionText(model, messages, options) → convenience text response

• chatCompletionEx(model, messages, options, requestOptions)
  • requestOptions.timeoutMs, requestOptions.retry, requestOptions.headers
• chatCompletionTextEx(model, messages, options, requestOptions)

• system(text) / user(text) / assistant(text)
• registerModel(alias, modelId) / getModels()
• extractText(resp)
• extractToolCalls(resp)
• createFunctionTool(name, description?, parameters?)

chatCompletionJson(model, messages, options?, requestOptions?) returns parsed JSON.

plugins.svyGroq.setApiKey(scopes.globals.GROQ_API_KEY);

var obj = plugins.svyGroq.chatCompletionJson(plugins.svyGroq.models.DEFAULT, [
  plugins.svyGroq.system('Return JSON only.'),
  plugins.svyGroq.user('Give me {"ok":true,"n":1}')
]);

application.output(obj.ok);

You can send tools and tool_choice through the regular options object.

var tool = plugins.svyGroq.createFunctionTool(
  'sum',
  'Sums two integers',
  {
    type: 'object',
    properties: {
      a: { type: 'integer' },
      b: { type: 'integer' }
    },
    required: ['a', 'b']
  }
);

var resp = plugins.svyGroq.chatCompletion(
  plugins.svyGroq.models.DEFAULT,
  [ plugins.svyGroq.user('Use the tool to sum 4 and 5') ],
  { tools: [tool], tool_choice: 'auto' }
);

var toolCalls = plugins.svyGroq.extractToolCalls(resp);
application.output(JSON.stringify(toolCalls));

chatCompletionStreamText(model, messages, options?, onDelta?, streamOptions?)

plugins.svyGroq.setApiKey(scopes.globals.GROQ_API_KEY);

var full = plugins.svyGroq.chatCompletionStreamText(
  plugins.svyGroq.models.DEFAULT,
  [ plugins.svyGroq.user('Write 5 short sentences.') ],
  { temperature: 0.2 },
  function(deltaText) {
    application.output(deltaText); // live chunks
  }
);

application.output('DONE: ' + full);

Note: Streaming depends on browser networking behavior. If your environment buffers the response, you may not see incremental deltas.

• models is null / missing IntelliSense
  • Ensure you installed the latest zip and reloaded solution. The package includes .spec types and a definition stub to avoid null.
• GroqService: apiKey not set
  • Call setApiKey() or set sendAuthHeader=false (proxy mode).
• 429 / rate limits
  • Tune retry and/or reduce max_tokens, lower concurrency.

• npm install
• npm run build

Output is written to dist/servoy/svy-groq.

• npm run make_release

This generates:

• svyGroq.zip (latest)
• svyGroq-<version>.zip (versioned)

Licensed under the Apache License 2.0. See LICENSE and the attribution notes in NOTICE.