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

# Reinforcement Learning

> Kody gets smarter with every interaction. By learning from your team's feedback, she continuously adapts her code suggestions to match your team's preferences and standards.

## Quick Overview

* 👍 Thumbs up: Kody learns what your team likes
* 👎 Thumbs down: Kody learns what to avoid
* ✨ Auto-learning: Kody tracks which suggestions get implemented

## How the Learning Works

<img src="https://mintcdn.com/kodus/AvHTi3FgfIMZisRO/images/finetuning.png?fit=max&auto=format&n=AvHTi3FgfIMZisRO&q=85&s=2f666c1cd3b62fe8f0f7b95e048576c8" alt="Kody Learnings" width="2520" height="1734" data-path="images/finetuning.png" />

### 1. Collecting Feedback

Kody learns from three main sources:

* **Direct Reactions**: Team members can react to Kody's suggestions on any code management tool:

  * 👍 = "This is what we want"
  * 👎 = "This doesn't match our style"

* **Implementation Tracking**: When your team implements Kody's suggestions, she marks it as positive feedback

* **Pattern Recognition**: Kody builds knowledge clusters based on your team's preferences

### 2. Smart Filtering

Once Kody has enough data, she:

1. Analyzes new code suggestions
2. Compares them against your team's preference clusters
3. Only shows suggestions that match your team's style
4. Filters out suggestions similar to previously rejected ones

## Why Your Feedback Matters

The more feedback you provide, the better Kody becomes at:

* ✅ Matching your coding style
* 🎯 Making relevant suggestions
* 🚫 Avoiding unwanted patterns
* 💡 Understanding team preferences

## Best Practices

1. **Be Consistent**: Encourage the team to provide feedback regularly
2. **React Promptly**: Give feedback while the context is fresh
3. **Team Alignment**: Ensure the team agrees on what makes a good suggestion