Skip to content

USER_GUIDE.md

Latest commit

 

History

History
333 lines (243 loc) · 8.13 KB

USER_GUIDE.md

File metadata and controls

333 lines (243 loc) · 8.13 KB

📖 StillMe User Guide

Welcome to StillMe! This guide will help you get started and make the most of StillMe's capabilities.

🌟 What is StillMe?

StillMe is a Learning AI system that:

  • Continuously learns from multiple sources (RSS feeds, arXiv, Wikipedia, etc.)
  • Provides context-aware responses using RAG (Retrieval-Augmented Generation)
  • Validates responses to reduce hallucinations and ensure accuracy
  • Admits when it doesn't know (Intellectual Humility)

Core Philosophy: "In the AI era, true value lies not in what AI can do, but in what AI chooses NOT to do."

🚀 Quick Start

Option 1: Use StillMe Dashboard (Easiest)

  1. Access the Dashboard:

  2. Start Chatting:

    • Type your question in the chat interface
    • StillMe will search its knowledge base and provide context-aware answers
    • Check citations to see sources StillMe used

  3. Explore Features:

    • View RAG retrieval results
    • See validation chain results
    • Monitor learning metrics

Option 2: Use API Directly

Simple Chat Request:

curl -X POST https://stillme-backend-production.up.railway.app/api/chat/rag \
  -H "Content-Type: application/json" \
  -d '{
    "message": "What is RAG?",
    "use_rag": true,
    "context_limit": 3
  }'

Response:

{
  "response": "RAG (Retrieval-Augmented Generation) is...",
  "confidence_score": 0.85,
  "citations": ["[1]", "[2]"],
  "sources": [...]
}

💡 How to Use StillMe Effectively

1. Ask Clear Questions

Good:

  • "What is the difference between RAG and fine-tuning?"
  • "Explain how StillMe's validation chain works"
  • "What sources does StillMe learn from?"

Less Effective:

  • "tell me stuff" (too vague)
  • "ai" (too broad)

2. Use Citations

StillMe provides citations [1], [2], [3] in responses. These refer to sources in StillMe's knowledge base:

  • Click on citations to see source details
  • Verify information by checking sources
  • Citations indicate StillMe found relevant context

3. Understand Confidence Scores

StillMe provides confidence scores (0.0-1.0):

  • 0.8-1.0: High confidence, reliable answer
  • 0.5-0.8: Medium confidence, verify with sources
  • 0.0-0.5: Low confidence, StillMe may not have enough information

What to do:

  • High confidence: Trust the answer, but still verify if critical
  • Medium confidence: Check citations, verify sources
  • Low confidence: Ask for clarification or rephrase your question

4. Multi-Language Support

StillMe supports multiple languages:

  • Vietnamese (Tiếng Việt)
  • English
  • Chinese (中文)
  • French (Français)
  • Spanish (Español)
  • German (Deutsch)
  • Japanese (日本語)
  • Korean (한국어)
  • Arabic (العربية)
  • Russian (Русский)
  • And more...

How to use:

  • Ask questions in your preferred language
  • StillMe will respond in the same language
  • No need to specify language - StillMe detects automatically

5. Check Validation Results

StillMe uses a Validation Chain to ensure quality:

  • Citation Required: Ensures StillMe cites sources when available
  • Evidence Overlap: Checks if answer matches retrieved context
  • Language Match: Ensures StillMe responds in your language
  • Confidence Score: Indicates how certain StillMe is
  • Ethics Validation: Ensures responses are ethical

What this means:

  • If validation passes: High-quality, reliable answer
  • If validation fails: StillMe may use fallback or ask for clarification

🎯 Use Cases

1. Research & Learning

Example:

You: "What are the latest developments in RAG systems?"

StillMe: [Provides answer with citations from recent papers]

Best for:

  • Understanding complex topics
  • Getting up-to-date information
  • Learning from multiple sources

2. Technical Questions

Example:

You: "How does StillMe's validation chain work?"

StillMe: [Explains validation chain with technical details]

Best for:

  • Technical documentation
  • Architecture questions
  • Implementation details

3. Multi-Language Communication

Example:

You: "Giải thích RAG là gì bằng tiếng Việt"

StillMe: [Responds in Vietnamese with detailed explanation]

Best for:

  • Non-English speakers
  • Learning in your native language
  • Cross-cultural communication

4. Fact-Checking

Example:

You: "Is it true that RAG reduces hallucinations?"

StillMe: [Provides answer with citations and confidence score]

Best for:

  • Verifying information
  • Understanding claims
  • Getting reliable sources

⚠️ Limitations & What StillMe Can't Do

StillMe CAN:

  • ✅ Search its internal knowledge base (RAG)
  • ✅ Provide context-aware answers
  • ✅ Cite sources from its knowledge base
  • ✅ Admit when it doesn't know
  • ✅ Validate responses for quality

StillMe CANNOT:

  • ❌ Search the internet in real-time
  • ❌ Access live websites
  • ❌ Provide information not in its knowledge base
  • ❌ Guarantee 100% accuracy (no AI can)
  • ❌ Replace human judgment for critical decisions

What This Means:

If StillMe says "I don't know":

  • StillMe doesn't have information in its knowledge base
  • This is a feature, not a bug - StillMe admits limitations
  • Try rephrasing your question or asking about a different topic

If StillMe provides low confidence:

  • StillMe found some information but isn't certain
  • Check citations to verify
  • Consider asking for clarification

🔍 Understanding StillMe's Responses

Response Structure

Typical Response:

[Answer text with citations [1], [2]]

Sources:
[1] Source name - Brief description
[2] Source name - Brief description

Confidence: 0.85
Validation: Passed

Citation Format

  • [1], [2], [3] = References to sources in StillMe's knowledge base
  • Click citations to see full source details
  • Citations indicate StillMe found relevant context

Confidence Scores

  • 0.9-1.0: Very high confidence
  • 0.7-0.9: High confidence
  • 0.5-0.7: Medium confidence
  • 0.3-0.5: Low confidence
  • 0.0-0.3: Very low confidence

🛠️ Advanced Usage

Custom LLM Provider

StillMe supports multiple LLM providers:

  • DeepSeek (default)
  • OpenAI
  • Claude (Anthropic)
  • Gemini (Google)
  • Ollama (local)
  • Custom providers

How to use:

{
  "message": "Your question",
  "llm_provider": "openai",
  "llm_api_key": "your-api-key",
  "llm_model_name": "gpt-4"
}

Self-Hosted Deployment

For self-hosted deployments:

  1. Set up your own backend
  2. Configure your LLM provider
  3. Use your own API keys
  4. See docs/SELF_HOSTED_SETUP.md for details

❓ Troubleshooting

StillMe doesn't understand my question

Solutions:

  • Rephrase your question more clearly
  • Break complex questions into smaller parts
  • Check if StillMe supports your language
  • Try asking in English first, then your language

StillMe gives low confidence

What to do:

  • Check citations - are sources relevant?
  • Rephrase your question
  • Ask about a different aspect of the topic
  • StillMe may not have enough information in its knowledge base

StillMe responds in wrong language

Solutions:

  • StillMe should auto-detect language, but if it doesn't:
  • Explicitly request: "Please respond in [your language]"
  • Report the issue - this is a bug we want to fix

StillMe says "I don't know"

This is normal:

  • StillMe admits when it doesn't know (Intellectual Humility)
  • Try rephrasing or asking about a different topic
  • StillMe's knowledge base may not have information on this topic

📚 Learn More

🤝 Getting Help

  • GitHub Discussions: Ask questions, share ideas
  • Issues: Report bugs or request features
  • Documentation: Check docs for detailed information

StillMe - Learning AI system with RAG foundation 🤖✨