Sys2DSL

Sys2DSL (System 2 Domain-Specific Language) is the structured command language used by AGISystem2. It provides a deterministic, testable, and auditable way to express knowledge and queries. This page explains the role of Sys2DSL and shows how natural language maps to structured commands.

Why a DSL?

AGISystem2 uses a constrained DSL rather than free-form natural language for several important reasons:

Determinism – The same command always produces the same result. No interpretation variability.
Testability – Commands can be scripted, versioned in git, and executed in CI pipelines.
Auditability – Every operation is logged with its exact command, enabling complete audit trails.
Precision – Unambiguous Subject-Relation-Object structure maps directly to geometric operations.

Natural Language to Sys2DSL

The Chat Interface uses an LLM to translate natural language into Sys2DSL commands. Here are examples of how common statements translate:

Teaching Facts

Natural Language	Sys2DSL Command
"Dogs are animals"	`ASSERT Dog IS_A Animal`
"Fire causes smoke"	`ASSERT Fire CAUSES Smoke`
"Paris is located in France"	`ASSERT Paris LOCATED_IN France`
"A car has wheels"	`ASSERT Car HAS_PART Wheel`
"Driving requires a license"	`ASSERT Driving REQUIRES License`
"Speeding is prohibited by law"	`ASSERT Speeding PROHIBITED_BY Law`

Asking Questions

Natural Language	Sys2DSL Command
"Is a dog an animal?"	`ASK "Dog IS_A Animal?"`
"What causes smoke?"	`ASK "? CAUSES Smoke"`
"Where is Paris located?"	`ASK "Paris LOCATED_IN ?"`
"What if water were a gas?"	`CF "Water IS_A Gas?" \| Water TEMPERATURE_AT Celsius200`

Command Structure

Every Sys2DSL command follows a consistent pattern:

@variable COMMAND arguments

The @variable captures the result. The COMMAND specifies the operation. Arguments provide the data.

Core Commands

Command	Purpose	Example
`ASSERT`	Add a fact	`@f ASSERT Dog IS_A Animal`
`ASK`	Query truth value	`@q ASK "Dog IS_A Animal?"`
`RETRACT`	Remove a fact	`@r RETRACT Dog IS_A Fish`
`FACTS_MATCHING`	Find matching facts	`@m FACTS_MATCHING ? IS_A Animal`

Theory Management

Command	Purpose	Example
`THEORY_PUSH`	Create new context layer	`@_ THEORY_PUSH name="experiment"`
`THEORY_POP`	Discard top layer	`@_ THEORY_POP`
`LOAD_THEORY`	Load theory file	`@_ LOAD_THEORY medical`
`SAVE_THEORY`	Save current state	`@_ SAVE_THEORY session_1`

Benefits for Testing and Validation

Automated Testing

Sys2DSL commands can be stored in test files and executed automatically:

# test_taxonomy.sys2dsl
@f1 ASSERT Dog IS_A Mammal
@f2 ASSERT Mammal IS_A Animal
@q ASK "Dog IS_A Animal?"
# Expected: TRUE_CERTAIN (transitive)

CI/CD Integration

# In GitHub Actions or similar
./bin/AGISystem2Raw.sh --batch tests/knowledge_base.sys2dsl
./bin/AGISystem2Raw.sh --exec "ask Is system consistent?" --json

Audit Logs

Every command execution is logged with timestamp and result:

2024-01-15 10:23:45 ASSERT Dog IS_A Animal → OK
2024-01-15 10:23:46 ASK "Dog IS_A Animal?" → TRUE_CERTAIN (0.95)
2024-01-15 10:23:47 THEORY_PUSH name="experiment" → depth=1

The Two Interfaces

AGISystem2 provides two ways to work with Sys2DSL:

Interface	Input	Best For
Raw CLI	Direct Sys2DSL commands	Testing, automation, scripting
Chat Interface	Natural language (translated by LLM)	Exploration, demos, learning

Both interfaces execute the same underlying Sys2DSL commands. The Chat Interface adds an LLM translation layer for convenience, while the Raw CLI provides direct control for precision work.

AGISystem2 – Sys2DSL