Skip to content

FEATURE_GUIDE.md

Latest commit

 

History

History
140 lines (104 loc) · 3.81 KB

FEATURE_GUIDE.md

File metadata and controls

140 lines (104 loc) · 3.81 KB
Raw

AI Agent Chatbot - Feature Guide

Overview

Your AI chatbot has 4 main components. Here's what each does:

1. Provider & Model Selection (Required ⚠️)

Provider

  • Currently: Only Groq available
  • What it is: The AI service that powers the responses
  • Future: Can add OpenAI, Anthropic, etc.

Model Selection

  • Llama 4 Scout (17B): Balanced speed & quality, good for general questions
  • Qwen 3 (32B): More capable, better for complex reasoning & code

Use Case: Choose based on your needs - fast answers vs. better quality

2. Your Message (Required ⚠️)

The question you want to ask the AI.

Examples:

  • "What is machine learning?"
  • "Write a Python function to calculate fibonacci"
  • "Explain quantum computing"

3. System Prompt (Optional - Leave blank if not needed)

Give the AI specific instructions on HOW to respond.

What it does:

  • Changes the AI's personality and response style
  • Does NOT change the topic (doesn't help find information)
  • Executed immediately when you ask a question

Real-World Examples:

Use Case System Prompt
For Beginners "Explain everything in simple terms without technical jargon"
For Developers "You are a Python expert. Provide only clean, production-ready code"
For Learning "Provide detailed explanations with examples"
For Quick Answers "Be very concise and brief, use bullet points"
Professional "Use formal and professional language"
Creative "Be creative and think outside the box"

✅ When to use:

  • You want a specific tone/style
  • You want code-only or explanation-only answers
  • You want specific formatting (bullet points, tables, etc.)

❌ When NOT to use:

  • Just asking general questions - leave it blank!
  • System prompts won't help find information

4. Allow Search (Optional - Unchecked by default)

Enable web search for real-time information.

What it does:

  • The AI searches the internet for current information
  • Useful for latest news, recent events, current data
  • Adds a small delay (searching takes time)

✅ When to use:

  • "What's the latest in AI/Tech news today?"
  • "What's the current weather?"
  • "Show me recent updates about [topic]"
  • "Find latest prices for [product]"

❌ When NOT to use:

  • General knowledge questions (slower, unnecessary)
  • Historical facts
  • Technical explanations
  • Code help

Example Comparison:

❌ Question: "What is Python?" 
→ Don't enable search (unnecessary)

✅ Question: "What Python features were released in 2024?"
→ Enable search (needs current info)

How They Work Together

Scenario 1: Learning Python

Model: Llama 4 Scout (17B)
Message: "Teach me Python decorators"
System Prompt: "Explain like I'm a beginner, use examples"
Allow Search: OFF ✓ (Learning material doesn't change)

Scenario 2: Current News

Model: Qwen 3 (32B)
Message: "What happened in AI this week?"
System Prompt: [Leave blank]
Allow Search: ON ✓ (Needs real-time info)

Scenario 3: Professional Code

Model: Qwen 3 (32B)
Message: "Write a REST API in Python"
System Prompt: "Provide production-ready code with error handling"
Allow Search: OFF ✓ (Code patterns don't change)

Tips for Best Results

  1. Be Specific - "Write a Python function" is better than "Write code"
  2. Use System Prompt for Style - Not for finding information
  3. Enable Search Only When Needed - It adds latency
  4. Choose Right Model - Qwen 3 for complex tasks, Llama 4 for quick answers
  5. Combine Features - System Prompt + Model choice = Best results

Future Enhancements

  • Multi-turn conversations (remember context)
  • More model options
  • File upload support
  • Response history
  • Custom model fine-tuning options