Quan starts with the read
The user gets price and session context first, without a rigid template or forced table.
Developer Guide · Quan 3.0 and 3.4
Easy stock questions should still feel smart.
Quan Light Market Insights add a flexible intent layer for questions like "How is AAPL performing today?" Quan answers in natural text first, then API developers can opt into compact source-grounded metadata for cards, chips, and follow-up prompts.
light-insight.json
{
"model": "quan-3.4",
"messages": [{
"role": "user",
"content": "How is AAPL performing today?"
}],
"stream": false,
"includeInsightMetadata": true
}
What changed
Text by default. Metadata when you ask for it.
Only the best extras
Quan may add move character, a watch item, or source-backed context when it helps. Unsupported catalysts are not invented.
Optional metadata
Set includeInsightMetadata to receive quanLightMarketInsight for simple stock questions.
Response shape
Use the object for chips, not as the answer.
{
"quanLightMarketInsight": {
"intent": "performance_snapshot",
"ticker": "AAPL",
"asOf": "2026-07-14T15:30:00.000Z",
"summary": "AAPL is up today, trading near $212.40 (+1.20%).",
"signals": [{
"type": "session_move",
"label": "Today's move",
"value": "+1.20%",
"sourceName": "Yahoo Finance quote",
"sourceUrl": "https://finance.yahoo.com/quote/AAPL/"
}],
"suggestedFollowups": [
"Why is it moving?",
"Bullish or bearish?",
"Compare AAPL to MSFT"
]
}
}
For streaming responses, listen for event: quan_light_market_insight. If Quan cannot ground the insight to a real quote source, the metadata is omitted.