Skip to main content
Timepoint Pro provides a powerful CLI for running temporal simulations, managing models, and querying simulation results.

​
Available Modes

The CLI supports six primary modes, each accessed via the mode parameter:

autopilot

Self-testing mode that evaluates temporal chains of varying lengths

train

Train entities with historical context and populate knowledge states

temporal_train

Train entities across temporal chains with causal evolution

evaluate

Run evaluation metrics on trained entities

interactive

Natural language query REPL for exploring simulation results

models

Manage and select LLM models

​
Autopilot Mode

Autopilot mode runs automated tests of temporal chains with different lengths to evaluate system performance: 
python cli.py mode=autopilot
What it does:
  • Tests temporal chains of configurable lengths (default: 3, 5, 7 timepoints)
  • Runs temporal training on each chain
  • Computes aggregate evaluation metrics
  • Generates cost and performance reports
  • Checks for causal chain violations
Configuration: 
autopilot:
  temporal_lengths: [3, 5, 7]  # Timepoint counts to test
Output:
  • Temporal coherence scores
  • Knowledge consistency scores
  • Causal chain violation counts
  • Per-template cost and token usage
  • Summary reports in JSON and Markdown

​
Train Mode

Trains entities using historical contexts from templates: 
python cli.py mode=train training.context=founding_fathers_1789
Features:
  • Rich historical context from predefined templates
  • Graph-based relationship modeling
  • Knowledge state population via LLM
  • Exposure event tracking
  • Resolution level management
Available contexts:
  • founding_fathers_1789 - Constitutional inauguration
  • Additional contexts in entity_templates.py

​
Temporal Train Mode

Trains entities across a temporal chain with causal propagation: 
python cli.py mode=temporal_train training.context=founding_fathers_1789 training.num_timepoints=5
Key features:
  • Builds causal temporal chains
  • Propagates knowledge states forward
  • Tracks knowledge growth per timepoint
  • Records exposure events for each learning moment
  • Validates temporal causality
Configuration: 
training:
  context: founding_fathers_1789
  num_timepoints: 5

​
Evaluate Mode

Runs evaluation metrics on all entities in the database: 
python cli.py mode=evaluate
Computed metrics:
  • Temporal Coherence - Consistency across timepoints
  • Knowledge Consistency - Information conservation compliance
  • Biological Plausibility - Constraint enforcement
Output:
  • Per-entity metric scores
  • Resolution level distribution
  • Aggregate statistics
  • Cost tracking
See Evaluate for detailed metric descriptions.

​
Interactive Mode

Natural language query REPL for exploring simulation data: 
python cli.py mode=interactive
Capabilities:
  • Parse natural language queries
  • Retrieve entity knowledge states
  • Synthesize contextual responses
  • Track query costs
  • Show simulation statistics
Example queries: 
"What did George Washington think about becoming president?"
"How did Thomas Jefferson feel during the inauguration?"
"What actions did Alexander Hamilton take after the ceremony?"
See Interactive Mode for full details.

​
Models Mode

Manage LLM model selection and testing: 
python cli.py mode=models
Operations:
  • List available Llama models from OpenRouter
  • View detailed model information (context length, pricing)
  • Switch between models
  • Test model connectivity
  • Refresh model catalog

​
Branch Mode

Explore counterfactual “what-if” scenarios: 
python cli.py mode=branch
Features:
  • Interactive branching explorer
  • Counterfactual query analysis
  • Alternate timeline generation

​
Configuration

All modes use Hydra configuration from conf/config.yaml: 
mode: autopilot  # or train, evaluate, interactive, models, branch

database:
  url: sqlite:///timepoint.db

llm:
  base_url: https://openrouter.ai/api/v1
  model: meta-llama/llama-3.1-70b-instruct

training:
  context: founding_fathers_1789
  num_timepoints: 5
  graph_size: 10
  target_resolution: SCENE

autopilot:
  temporal_lengths: [3, 5, 7]

​
Command-Line Overrides

Override any config value from the command line: 
# Change mode
python cli.py mode=train

# Override training parameters
python cli.py mode=temporal_train training.num_timepoints=10

# Change LLM model
python cli.py mode=interactive llm.model=deepseek/deepseek-chat

# Set database path
python cli.py database.url=sqlite:///custom.db

​
Cost Tracking

All modes track LLM API costs and token usage: 
Cost so far: $0.0234
Tokens used: 1,245
Reports are generated in reports/ directory with:
  • Total cost breakdown
  • Token usage statistics
  • Per-operation metrics
  • Timestamp and configuration

​
Next Steps

Run Command

Execute simulations with ./run.sh

Training

Learn about training modes

Interactive Queries

Query your simulation data

Evaluation

Understand evaluation metrics

Build docs developers (and LLMs) love

Get started for freeTalk to us