AQL Playground

Agent Query Language — In-Browser Demo (WASM) — GitHub · Spec v0.5

Loading WASM module...

Examples — Click to Execute

STORE → Working

STORE INTO WORKING (name = "alice", score = 95)

STORE → Working

STORE INTO WORKING (name = "bob", score = 82)

STORE → With TTL

STORE INTO WORKING ... TTL 60s

STORE → Scoped

STORE ... SCOPE private NAMESPACE "auth"

STORE → Episodic

STORE INTO EPISODIC (event = "purchase")

STORE → Procedural

STORE INTO PROCEDURAL (pattern = "retry")

STORE → Tools

STORE INTO TOOLS (name = "web_search")

STORE → Tools

STORE INTO TOOLS (name = "calculator")

RECALL → Filter

RECALL FROM WORKING WHERE score > 80

RECALL → Complex

WHERE (A AND B) OR C

RECALL → ORDER + LIMIT

ORDER BY score DESC LIMIT 3

RECALL → RETURN

RETURN name, score

RECALL → Scoped

RECALL ... SCOPE private NAMESPACE

RECALL → Procedural

RECALL FROM PROCEDURAL

AGGREGATE → All Funcs

COUNT, AVG, MAX, MIN

AGGREGATE + HAVING

AGGREGATE ... HAVING avg > 80

AGGREGATE → SUM

SUM(amount) AS total_spend

LOAD → Ranked Tools

LOAD FROM TOOLS ORDER BY ranking

LOOKUP → Pattern

LOOKUP FROM PROCEDURAL

SCAN → Last N

SCAN WINDOW LAST 5

SCAN → TOP BY

SCAN WINDOW TOP 3 BY score

SCAN → Duration

SCAN WINDOW LAST 30s

SEMANTIC → ML

RECALL SEMANTIC LIKE "machine learning"

SEMANTIC → DB

RECALL FROM SEMANTIC LIKE "database"

LINK → Create

LINK FROM ... TO ... AS "performed"

LINK → Proc→Tool

LINK PROCEDURAL TO TOOLS AS "uses"

WITH LINKS ALL

RECALL ... WITH LINKS ALL

FOLLOW LINKS

FOLLOW LINKS "performed" DEPTH 1

REFLECT → Multi

REFLECT FROM WORKING, EPISODIC, PROC

REFLECT → All+Links

REFLECT FROM ALL WITH LINKS

REFLECT → THEN

REFLECT ... THEN STORE

PIPELINE → Multi-Stage

PIPELINE setup TIMEOUT 100ms {...}

PIPELINE → RTB

PIPELINE rtb_flow TIMEOUT 50ms

PIPELINE → Analyze

PIPELINE with AGGREGATE

UPDATE → Additive

UPDATE ... (score = 100, level = "expert")

UPDATE → Procedural

UPDATE PROCEDURAL confidence

FORGET → Inactive

FORGET WHERE active = false