The Raw CLI (AGISystem2Raw.sh) provides direct access to the AGISystem2 engine through structured Sys2DSL commands. It is designed for automation, testing, and scripting scenarios where determinism and precise control are essential. Unlike the Chat Interface, it requires no external LLM dependencies and executes commands exactly as specified.

Requirements

The Raw CLI requires only:

• Node.js 18+ – For running the JavaScript engine
• Bash – For the launcher script (Linux, macOS, WSL)

No external API keys or network access required. The engine runs entirely locally.

Starting the Raw CLI

Interactive Mode

# Start interactive REPL
./bin/AGISystem2Raw.sh

# The prompt appears:
AGIS2> 

Batch Mode

# Execute single command
./bin/AGISystem2Raw.sh --exec "add Dog IS_A Animal"

# Execute with JSON output
./bin/AGISystem2Raw.sh --exec "ask Is Dog an Animal?" --json

# Execute multiple commands from file
./bin/AGISystem2Raw.sh --batch commands.sys2dsl

# Pipe commands via stdin
echo "add Cat IS_A Animal" | ./bin/AGISystem2Raw.sh --batch -

Command Syntax

Commands follow the Sys2DSL syntax. The full reference is in the Syntax Guide.

Adding Facts

# Basic fact assertion
AGIS2> add Dog IS_A Animal
OK: ASSERT Dog IS_A Animal

# With explicit ASSERT keyword
AGIS2> ASSERT Cat IS_A Mammal
OK: ASSERT Cat IS_A Mammal

# Multiple relations
AGIS2> add Fire CAUSES Smoke
AGIS2> add Fire CAUSES Heat
AGIS2> add Water EXTINGUISHES Fire

Asking Questions

# Yes/no question
AGIS2> ask Is Dog an Animal?
Result: TRUE_CERTAIN
Confidence: 0.95

# What-type question
AGIS2> ask What is Dog?
Result: Animal, Mammal

# Structured query
AGIS2> ASK "Dog IS_A ?"
Result: Animal

Listing Knowledge

# List all facts
AGIS2> list facts

# List concepts
AGIS2> list concepts

# Match pattern
AGIS2> match ? IS_A Animal
Dog IS_A Animal
Cat IS_A Animal

Theory Management

The Raw CLI supports the full theory stack operations:

# Push new theory layer
AGIS2> THEORY_PUSH name="medical_context"
OK: Pushed theory layer "medical_context"

# Check current stack
AGIS2> LIST_THEORIES
{ "depth": 1, "name": "medical_context" }

# Add facts to current layer
AGIS2> add Aspirin REDUCES Inflammation

# Pop layer (discards facts added in that layer)
AGIS2> THEORY_POP
OK: Popped theory layer "medical_context"

# Save theory to file
AGIS2> SAVE_THEORY medical_knowledge
Saved to .AGISystem2/theories/medical_knowledge.sys2dsl

# Load theory from file
AGIS2> LOAD_THEORY medical_knowledge
Loaded 15 facts from medical_knowledge.sys2dsl

Output Formats

Text Output (Default)

$ ./bin/AGISystem2Raw.sh --exec "ask Is Dog an Animal?"
Result: TRUE_CERTAIN
Confidence: 0.95
Reasoning: Direct inclusion in concept "Animal"

JSON Output

$ ./bin/AGISystem2Raw.sh --exec "ask Is Dog an Animal?" --json
{
  "success": true,
  "result": {
    "truth": "TRUE_CERTAIN",
    "confidence": 0.95,
    "reasoning": [
      { "step": "inclusion_check", "concept": "Animal", "result": "inside" }
    ]
  }
}

Variable Bindings

Commands can use variable bindings for complex operations:

# Store result in variable
AGIS2> @result ASK "Dog IS_A ?"

# Use variable in subsequent command
AGIS2> @facts FACTS_MATCHING Dog ? ?

# Chain operations
AGIS2> @f ASSERT Bird IS_A Animal
AGIS2> @q ASK "Bird IS_A Animal?"
AGIS2> echo @q.truth
TRUE_CERTAIN

Batch File Format

Batch files (.sys2dsl) contain one command per line:

# comments.sys2dsl - Example batch file
# Lines starting with # are comments

# Define taxonomy
ASSERT Dog IS_A Mammal
ASSERT Cat IS_A Mammal
ASSERT Mammal IS_A Animal
ASSERT Bird IS_A Animal

# Define properties
ASSERT Dog HAS_PROPERTY barks
ASSERT Cat HAS_PROPERTY meows
ASSERT Bird HAS_PROPERTY flies

# Define relations
ASSERT Fire CAUSES Smoke
ASSERT Water EXTINGUISHES Fire

Execute with:

./bin/AGISystem2Raw.sh --batch comments.sys2dsl

Exit Codes

Code	Meaning
0	Success
1	General error
2	Parse error (invalid syntax)
3	File not found (batch mode)
4	Query returned FALSE

Exit codes allow integration with shell scripts and CI pipelines:

#!/bin/bash
./bin/AGISystem2Raw.sh --exec "ask Is system consistent?" --json
if [ $? -eq 0 ]; then
  echo "System is consistent"
else
  echo "Inconsistency detected"
  exit 1
fi

Environment Variables

Variable	Description	Default
AGIS2_PROFILE	Configuration profile	manual_test
AGIS2_DATA_DIR	Data directory path	.AGISystem2/data
AGIS2_DEBUG	Enable debug output	0

Integration Examples

CI/CD Pipeline

# .github/workflows/knowledge-test.yml
- name: Test knowledge consistency
  run: |
    ./bin/AGISystem2Raw.sh --batch tests/rules.sys2dsl
    ./bin/AGISystem2Raw.sh --exec "ask Is system consistent?" --json

Shell Script Automation

#!/bin/bash
# load-and-query.sh

# Load base knowledge
./bin/AGISystem2Raw.sh --batch base_knowledge.sys2dsl

# Query specific facts
result=$(./bin/AGISystem2Raw.sh --exec "ask $1?" --json)
echo "$result" | jq '.result.truth'

Node.js Integration

const { execSync } = require('child_process');

function query(question) {
  const result = execSync(
    `./bin/AGISystem2Raw.sh --exec "ask ${question}?" --json`,
    { encoding: 'utf8' }
  );
  return JSON.parse(result);
}

const answer = query('Is Dog an Animal');
console.log(answer.result.truth); // TRUE_CERTAIN

Comparison with Chat Interface

Aspect	Raw CLI	Chat Interface
Input	Structured Sys2DSL commands	Natural language
Determinism	100% reproducible	LLM introduces variability
Dependencies	Node.js only	Node.js + LLM API
Batch mode	Full support	Interactive only
Best for	Testing, CI/CD, automation	Exploration, demos, learning