Files

Apple 831f361f0f docs: expand lint scope batch3 (6 files)

2026-02-16 03:44:58 -08:00

8.1 KiB

Raw Permalink Blame History

DAARWIZZ - AI Agent for DAARION.city

DAARWIZZ is the official AI agent for the DAARION.city ecosystem, designed to help community members navigate microDAO processes, understand roles, and interact with the DAGI Stack.

🎯 What is DAARWIZZ?

DAARWIZZ is an intelligent agent that:

Guides users through microDAO operations
Explains roles and permissions (RBAC)
Answers questions about DAO processes
Provides onboarding for new members
Enforces security by respecting user entitlements

🧠 Personality & Behavior

Core Traits

Helpful: Provides practical, actionable guidance
Concise: Avoids unnecessary explanations
Security-aware: Respects RBAC permissions
Honest: Admits when it doesn't know something
Professional yet friendly: Approachable but not casual

Style Guidelines

Short paragraphs, no fluff
Step-by-step instructions when needed
Uses numbered lists for procedures
Always considers user's role and permissions

📋 System Prompt

DAARWIZZ's behavior is defined by its system prompt located at:

gateway-bot/daarwizz_prompt.txt

Key elements:

Identity: Official DAARION.city AI agent
Role: Help with microDAO, roles, processes
Context awareness: Uses dao_id, user_id, RBAC data
Limitations: No legal/financial/medical advice
Knowledge boundaries: Only DAARION/microDAO topics

🔄 How DAARWIZZ Works

Message Flow

User (Telegram)
    ↓ "Привіт! Що це за DAO?"
Gateway (:9300)
    ↓ Load daarwizz_prompt.txt
    ↓ Build request with agent="daarwizz"
Router (:9102)
    ↓ Fetch RBAC context
RBAC (:9200)
    ↓ role: member, entitlements: [chat, vote, comment]
Router
    ↓ Inject RBAC + system_prompt
LLM Provider (Ollama/OpenAI)
    ↓ Generate response with full context
Gateway
    ↓ Send to Telegram
User receives answer

🛠️ Technical Implementation

Gateway Integration

File: gateway-bot/http_api.py

# Load DAARWIZZ prompt on startup
DAARWIZZ_SYSTEM_PROMPT = load_daarwizz_prompt()

# Build request to Router
router_request = {
    "prompt": text,
    "mode": "chat",
    "agent": "daarwizz",  # Agent identifier
    "metadata": {
        "dao_id": "greenfood-dao",
        "user_id": "tg:12345",
        ...
    },
    "context": {
        "agent_name": "DAARWIZZ",
        "system_prompt": DAARWIZZ_SYSTEM_PROMPT,
        # RBAC injected by Router
    },
}

LLM Provider Support

File: providers/llm_provider.py

def _get_system_prompt(self, req: RouterRequest) -> Optional[str]:
    # 1. Check context.system_prompt (from Gateway)
    context = req.payload.get("context") or {}
    if "system_prompt" in context:
        return context["system_prompt"]

    # 2. Agent-specific fallback
    if req.agent == "daarwizz":
        return "Ти — DAARWIZZ, AI-агент DAARION.city..."

    return None

Docker Configuration

File: gateway-bot/Dockerfile

# Copy DAARWIZZ prompt
COPY gateway-bot/daarwizz_prompt.txt /app/gateway-bot/daarwizz_prompt.txt

ENV DAARWIZZ_NAME=DAARWIZZ
ENV DAARWIZZ_PROMPT_PATH=/app/gateway-bot/daarwizz_prompt.txt

📖 Example Interactions

Basic Greeting

User: Привіт! Хто ти?

DAARWIZZ:

Привіт! Я — DAARWIZZ, AI-агент екосистеми DAARION.city.

Я допомагаю учасникам розібратися з:
• microDAO та їх структурою
• Ролями і правами доступу (RBAC)
• Процесами голосування та пропозиціями
• Завданнями та винагородами

Чим можу допомогти?

Role Information

User: Які у мене права в цьому DAO?

DAARWIZZ (with RBAC context):

У greenfood-dao ти маєш роль: member

Твої права (entitlements):
1. chat_access - писати в чат
2. read_proposals - читати пропозиції
3. vote - голосувати за пропозиції
4. comment - коментувати

Для більших прав (наприклад, створення пропозицій) потрібна роль contributor або вища.

Onboarding

User: Як розпочати роботу в DAO?

DAARWIZZ:

Ласкаво просимо! Ось кроки для старту:

1. Ознайомся з документацією DAO (команда /docs)
2. Переглянь активні пропозиції (/proposals)
3. Познайомся з командою в чаті
4. Виконай перше завдання (/tasks)
5. Накопичуй репутацію для підвищення ролі

Твоя поточна роль: guest
Наступна роль: member (після 1 виконаного завдання)

Є питання? Запитуй!

🎛️ Customization

Updating System Prompt

Edit prompt file:
```
nano gateway-bot/daarwizz_prompt.txt
```

Rebuild Gateway:

docker-compose build gateway
docker-compose restart gateway

Verify loading:

docker-compose logs gateway | grep "DAARWIZZ system prompt loaded"

Environment Variables

# .env
DAARWIZZ_NAME=DAARWIZZ
DAARWIZZ_PROMPT_PATH=/app/gateway-bot/daarwizz_prompt.txt

🧪 Testing DAARWIZZ

Local Test (curl)

curl -X POST http://localhost:9300/telegram/webhook \
  -H "Content-Type: application/json" \
  -d '{
    "message": {
      "message_id": 1,
      "from": {"id": 12345, "username": "testuser"},
      "chat": {"id": 12345, "type": "private"},
      "text": "Привіт! Хто ти?"
    }
  }'

Telegram Bot

Set up webhook:

curl -X POST "https://api.telegram.org/bot$TELEGRAM_BOT_TOKEN/setWebhook" \
  -d "url=https://your-domain.com/telegram/webhook"

Send message to bot in Telegram
Check logs:
```
docker-compose logs -f gateway router
```

Expected Log Output

gateway | INFO: DAARWIZZ system prompt loaded (1243 chars)
gateway | INFO: Telegram message from testuser (tg:12345): Привіт! Хто ти?
gateway | INFO: Sending to Router: agent=daarwizz, dao=greenfood-dao
router  | INFO: Received request: mode=chat, agent=daarwizz
router  | INFO: RBAC context: role=member, entitlements=4
router  | INFO: Routing to llm_local_qwen3_8b
router  | INFO: Response generated (345 chars, 2.3s)
gateway | INFO: Telegram message sent to chat 12345

📊 Metrics & Monitoring

Health Check

curl http://localhost:9300/health

Response:

{
  "status": "healthy",
  "agent": "DAARWIZZ",
  "system_prompt_loaded": true,
  "timestamp": "2024-11-15T14:30:00Z"
}

Performance Targets

Metric	Target	Notes
Prompt loading	< 100ms	On startup
Request enrichment	< 50ms	Gateway processing
End-to-end response	< 5s	Including LLM generation
RBAC lookup	< 100ms	From RBAC service

🚀 Future Enhancements

Phase 4 Roadmap

Knowledge Base Integration
- Connect to microdao-daarion docs
- RAG (Retrieval-Augmented Generation)
- Contextual answers from official docs
Multi-language Support
- Ukrainian (default)
- English
- Auto-detect user language
Workflow Triggers
- User: "Onboard me" → triggers CrewAI workflow
- User: "Review proposal #123" → triggers multi-agent review
Analytics
- Track common questions
- Identify knowledge gaps
- Improve prompt iteratively
Personalization
- Remember user context across sessions
- Adapt responses based on user role
- Suggest relevant actions

Version: 0.2.0 Agent: DAARWIZZ Status: Production-ready ✅ Last Updated: 2024-11-15

8.1 KiB Raw Permalink Blame History Unescape Escape