market-pattern

Load normalized OHLCV candles from the market-data namespace, then scan candle sequences for known formations like doji, hammer, engulfing, and morning star. Each detection is classified by type, direction, and a reliability score from 0.0 to 1.0.

It detects single-candle patterns (doji, hammer, inverted hammer), two-candle patterns (bullish and bearish engulfing), three-candle patterns (morning star, evening star, three white soldiers, three black crows), and multi-candle patterns (head and shoulders, double top/bottom, cup and handle).

Yes, data must be ingested first via the market-ingest skill. Pattern detection reads normalized OHLCV candles from the market-data memory namespace, so running it without prior ingestion returns no data to scan.

Patterns are stored through the agentdb pattern-store with type market-pattern, or via memory_store in the market-patterns namespace. You can retrieve them later with memory search queries or the CLI memory search command.

The reliability score is a value from 0.0 to 1.0 assigned to each detection indicating confidence in the pattern. Results are ranked by this score in descending order so the strongest signals appear first in the report.