Skip to main content

​
Overview

The Groq Microservice is designed to be simple yet flexible. This guide covers all configuration options, from environment variables to customizing the AI system prompt for your specific institutional needs.

​
Environment Variables

The service uses environment variables stored in a .env file at the project root. These variables control authentication, server behavior, and operational settings.
Never commit your .env file to version control. The .gitignore file already excludes it by default.

​
Required Variables

Type: string (required)Your Groq API authentication key for accessing the LLM API.
.env
GROQ_API_KEY=gsk_your_actual_api_key_here
Where to get it:
  1. Sign up at console.groq.com
  2. Navigate to API Keys section
  3. Create a new API key
  4. Copy the key immediately (it won’t be shown again)
Security Notes:
  • Store this key securely
  • Rotate keys periodically
  • Use different keys for development and production
  • Never expose in client-side code or logs
Implementation in code:
server.js:33
headers: {
  Authorization: `Bearer ${process.env.GROQ_API_KEY}`,
  "Content-Type": "application/json",
}

​
Optional Variables

Type: number (optional)Default: 5055The port number on which the Express server will listen for incoming requests.
.env
PORT=5055
Common port configurations:
  • Development: 5055 (default)
  • Production: 80 (HTTP), 443 (HTTPS), or 3000
  • Docker: Use environment variable passed at runtime
Implementation in code:
server.js:56-58
app.listen(process.env.PORT || 5055, () => {
  console.log(`Groq micro corriendo en http://localhost:${process.env.PORT || 5055}`);
});
For production deployments behind a reverse proxy (nginx, Apache), use a non-privileged port like 5055 or 3000.

​
System Prompt Customization

The buildSystem() function defines the AI’s behavior and output format. This is the core of the document generation logic and is highly customizable.

​
Default System Prompt

The default prompt is optimized for Mexican government institutional documents:
server.js:9-16
function buildSystem() {
  return [
    "Eres un asistente de redacción institucional para un Ayuntamiento en México.",
    "Devuelve SOLO el texto final del machote (sin explicación).",
    "Incluye: encabezado institucional, ASUNTO, lugar y fecha, cuerpo, cierre, ATENTAMENTE, firma y C.c.p.",
    "Usa variables en formato {{Variable}} (doble llaves). No uses corchetes [] ni llaves simples.",
  ].join(" ");
}

​
Understanding the Prompt Components

1

Identity and Role

"Eres un asistente de redacción institucional para un Ayuntamiento en México."
Defines the AI’s role as an institutional writing assistant for Mexican municipal governments. This sets the tone, formality level, and cultural context.Customization options:
  • Change “Ayuntamiento” to your specific institution type (Secretaría, Dirección, etc.)
  • Modify country/region for different administrative contexts
  • Adjust formality level (formal, semi-formal)
2

Output Format Control

"Devuelve SOLO el texto final del machote (sin explicación)."
Ensures the AI returns only the template text without explanations or meta-commentary.
This prevents the AI from adding “Here’s the template:” or “I’ve created…” prefixes.
3

Document Structure Requirements

"Incluye: encabezado institucional, ASUNTO, lugar y fecha, cuerpo, cierre, ATENTAMENTE, firma y C.c.p."
Specifies the mandatory sections of the document template. This ensures consistency across all generated documents.Required sections:
  • encabezado institucional: Institutional header
  • ASUNTO: Subject line
  • lugar y fecha: Location and date
  • cuerpo: Main body content
  • cierre: Closing statement
  • ATENTAMENTE: Formal signature block
  • firma: Signature line
  • C.c.p.: Carbon copy recipients
4

Variable Formatting Rules

"Usa variables en formato {{Variable}} (doble llaves). No uses corchetes [] ni llaves simples."
Enforces consistent variable syntax using double curly braces {{Variable}} instead of other formats.Why this matters:
  • Enables easy template variable replacement
  • Compatible with common templating engines (Handlebars, Mustache)
  • Prevents ambiguity with other bracket types
Example variables in generated templates:
  • {{NombreCompleto}}
  • {{Cargo}}
  • {{Dependencia}}
  • {{FechaActual}}

​
Customization Examples

function buildSystem() {
  return [
    "Eres un asistente de redacción institucional para una Secretaría Federal en México.",
    "Devuelve SOLO el texto final del machote (sin explicación).",
    "Incluye: membrete federal, ASUNTO, lugar y fecha, ANTECEDENTES, cuerpo, CONSIDERANDOS, resolutivos, ATENTAMENTE, SUFRAGIO EFECTIVO. NO REELECCIÓN, firma y C.c.p.",
    "Usa variables en formato {{Variable}} (doble llaves). No uses corchetes [] ni llaves simples.",
  ].join(" ");
}

​
LLM Model Configuration

The service uses Groq’s LLM API with configurable model and parameters.

​
Current Configuration

server.js:36-40
body: JSON.stringify({
  model: "llama-3.1-8b-instant", // "llama-3.3-70b-versatile"
  messages,
  temperature: 0.4,
})

​
Model Options

Characteristics:
  • Fast response times (< 1 second)
  • Lower cost per request
  • Good for straightforward template generation
  • 8 billion parameters
Best for:
  • High-volume production environments
  • Standard document templates
  • Cost-sensitive deployments
  • Real-time user-facing applications
Characteristics:
  • Higher quality outputs
  • Better understanding of complex requests
  • Slightly slower response times
  • 70 billion parameters
Best for:
  • Complex document requirements
  • Higher quality expectations
  • Lower request volumes
  • When response time is less critical
How to switch:
server.js:36
model: "llama-3.3-70b-versatile",

​
Temperature Setting

server.js:39
temperature: 0.4,
Temperature: 0.4 (range: 0.0 - 2.0)
  • Current (0.4): Balanced between consistency and creativity
  • Lower (0.1-0.3): More deterministic, conservative outputs
  • Higher (0.6-0.9): More creative, varied responses
For institutional documents, we recommend keeping temperature between 0.2-0.5 to maintain professional consistency.

​
Request Configuration

The API endpoint accepts requests with specific structure:
server.js:25-28
const messages = [
  { role: "system", content: buildSystem() },
  { role: "user", content: `Área: ${area || "No especificada"}\nSolicitud: ${prompt}` },
];

​
Customizing User Message Format

You can modify how user input is formatted before sending to the LLM: 
const messages = [
  { role: "system", content: buildSystem() },
  { 
    role: "user", 
    content: `Área: ${area || "No especificada"}
Institución: ${institucion || "Ayuntamiento"}
Nivel de urgencia: ${urgencia || "Normal"}
Solicitud: ${prompt}` 
  },
];

​
Middleware Configuration

The service includes middleware for CORS and JSON parsing:

​
CORS Configuration

server.js:6
app.use(cors());
The default configuration allows all origins. For production, restrict to specific domains: 
import cors from "cors";

app.use(cors({
  origin: [
    'https://yourdomain.com',
    'https://app.yourdomain.com'
  ],
  methods: ['POST'],
  credentials: true
}));

​
JSON Body Parsing

server.js:7
app.use(express.json({ limit: "1mb" }));
Limit: 1mb - Maximum request body size
Increase the limit if you expect larger prompts:
app.use(express.json({ limit: "5mb" }));

​
Error Handling Configuration

The endpoint includes comprehensive error handling:
server.js:19-54
try {
  const { prompt, area } = req.body || {};
  if (!prompt || typeof prompt !== "string") {
    return res.status(400).json({ error: "prompt es requerido" });
  }
  
  // API call...
  
  if (!r.ok) {
    const details = await r.text();
    return res.status(502).json({ error: "Groq error", details });
  }
  
  // Success...
} catch (e) {
  return res.status(500).json({ error: "server error", details: String(e) });
}

​
Custom Error Messages

You can customize error responses: 
if (!prompt || typeof prompt !== "string") {
  return res.status(400).json({ 
    error: "Solicitud inválida",
    message: "El campo 'prompt' es requerido y debe ser texto",
    code: "MISSING_PROMPT"
  });
}

​
Environment-Specific Configuration

Create multiple environment files for different deployments: 
.env.development    # Local development
.env.staging        # Staging environment
.env.production     # Production environment
Load the appropriate file based on NODE_ENV: 
import dotenv from 'dotenv';

const envFile = process.env.NODE_ENV === 'production' 
  ? '.env.production' 
  : '.env.development';

dotenv.config({ path: envFile });

​
Next Steps

Build docs developers (and LLMs) love

Get started for freeTalk to us