> ## Documentation Index
> Fetch the complete documentation index at: https://docs.enconvo.ai/llms.txt
> Use this file to discover all available pages before exploring further.

# Groq

> Ultra-fast inference with Groq in EnConvo

## Overview

Groq provides extremely fast inference using custom LPU hardware. Perfect for real-time applications requiring instant responses.

## Supported Models

| Model             | Description        | Speed       |
| ----------------- | ------------------ | ----------- |
| **Gemma2 9B IT**  | Google's model     | \~700 tok/s |
| **Llama 3.1 70B** | Meta's large model | \~300 tok/s |
| **Llama 3.1 8B**  | Meta's small model | \~750 tok/s |
| **Mixtral 8x7B**  | Mistral MoE        | \~500 tok/s |

## Setup

<Steps>
  <Step title="Get API Key">
    1. Go to [Groq Console](https://console.groq.com)
    2. Sign in or create an account
    3. Navigate to **API Keys**
    4. Create a new API key
  </Step>

  <Step title="Configure in EnConvo">
    1. Open **Settings** → **AI Provider**
    2. Select **Groq AI**
    3. Go to **Credentials** module
    4. Enter your API key
  </Step>

  <Step title="Select Model">
    Choose your preferred model from the dropdown
  </Step>
</Steps>

## Configuration

| Setting         | Description           | Default      |
| --------------- | --------------------- | ------------ |
| **Credentials** | API key configuration | Required     |
| **Model Name**  | Model to use          | Gemma2 9B IT |
| **Temperature** | Creativity (0-2)      | Medium (1)   |

## Validate and Use

<Steps>
  <Step title="Validate credentials">
    Click **Validate** in the Groq credential settings. If validation fails, confirm your API key is active and your Groq account has quota available.
  </Step>

  <Step title="Pick for speed">
    Groq is best when low latency matters. Start with a small or medium model for quick interactive chat.
  </Step>

  <Step title="Watch rate limits">
    If responses fail during bursts, wait briefly or switch to a smaller model. Free-tier rate limits can be strict.
  </Step>
</Steps>

## Reasoning Effort

For reasoning-capable models (GPT-OSS):

| Level      | Description    |
| ---------- | -------------- |
| **Low**    | Fast reasoning |
| **Medium** | Balanced       |
| **High**   | Thorough       |

## Pricing

Groq offers generous free tiers. Check [Groq](https://groq.com) for current pricing.

<Note>
  Groq's free tier is great for trying ultra-fast inference!
</Note>

## Why Groq?

<CardGroup cols={2}>
  <Card title="Speed" icon="bolt">
    Fastest inference available - 300-750+ tokens/second
  </Card>

  <Card title="Free Tier" icon="gift">
    Generous free usage for development
  </Card>

  <Card title="Quality Models" icon="star">
    Access to Llama, Gemma, Mixtral
  </Card>

  <Card title="Low Latency" icon="clock">
    Near-instant responses
  </Card>
</CardGroup>

## Best Practices

<AccordionGroup>
  <Accordion title="When to use Groq">
    * Real-time chat applications
    * Quick iterations during development
    * Time-sensitive tasks
  </Accordion>

  <Accordion title="Model Selection">
    * **Llama 3.1 70B**: Best quality
    * **Llama 3.1 8B**: Fastest
    * **Gemma2 9B**: Good balance
  </Accordion>
</AccordionGroup>

## Troubleshooting

<AccordionGroup>
  <Accordion title="Rate limits">
    * Free tier has rate limits
    * Wait and retry, or upgrade
  </Accordion>
</AccordionGroup>