How It Works

This page explains how Daiko Factory turns your trader-AI setup into actual meme calls. Under the hood, every trader AI is still powered by conditions and live data evaluation.

The Evaluation Loop

When you save a trader AI, here's what happens behind the scenes:

1. Your DSL is validated - The system checks that your condition uses valid fields and operators
2. It's compiled to SQL - Your condition becomes a parameterized database query (no raw SQL injection possible)
3. The query runs - Matching tokens are fetched from the live database
4. Results are filtered - Cooldown and limits are applied
5. Calls are delivered - You receive the meme call via Telegram (if connected)

This loop runs periodically (typically every few seconds), so your trader AI is always evaluating fresh market data.

Key Concepts

Limits and Constraints

Result Limit: 500 Tokens

Each evaluation returns at most 500 tokens. If your condition matches more than 500 tokens:

• Only the first 500 (based on your order_by or default ordering) are returned
• You won't see the rest until they rise to the top or your condition changes

Best Practice: Use more specific conditions or add an order_by clause to prioritize what matters most.

Cooldown Period

After you receive a notification about a specific token for a specific condition, that token enters a cooldown period for that condition. During cooldown:

• The token may still match your condition
• But you won't receive duplicate notifications
• Cooldown is tracked per (condition, token) pair

Active Conditions Only

Only active conditions for active agents are evaluated. If you:

• Deactivate a condition: It stops being evaluated
• Delete an agent: All its conditions are removed

Notifications

Telegram (Primary Channel)

When tokens match your trader AI:

1. The system checks if you have Telegram connected
2. If yes, it sends a message with:
   • Trader AI name
   • Matching token details (symbol, address, metrics)
   • A link to view more

What Gets Sent

Each notification includes:

Data Freshness

Your conditions run against near-real-time data:

• Token metadata (symbol, creation time): Updated as new tokens are created
• Metrics (price, market cap, holder count): Updated continuously from on-chain data
• Trade data: Streamed from the blockchain with minimal delay

Security Model

Your conditions are safe because:

1. No raw SQL: You write DSL (JSON), not SQL. The system validates and compiles it.
2. Field whitelist: Only specific, allowed fields can be queried
3. Parameterized queries: All values are passed as parameters, not string-concatenated
4. Validation at parse time: Invalid conditions are rejected before execution

This design prevents SQL injection and ensures you can only query intended data.

Summary

For details on what you can query, see the DSL Reference.