Skip to content

LR-Agent v1.1.0 — ChronoForge

Latest
Latest

Choose a tag to compare

View all tags
@github-actions github-actions released this 27 Sep 03:29
v1.1.0
5e13ec5

LR-Agent v1.1.0 — ChronoForge

Headline

Let a patch live through simulated future repository generations before you merge it.

What's new

ChronoForge

LR-Agent can now age the current workspace or a completed Counterfactual Forge candidate through multiple sequential future repository trajectories.

Future-maintainer agents can evolve the shadow repository under pressures such as:

  • dependency upgrades;
  • API deprecation;
  • adjacent features;
  • schema migration;
  • module-boundary changes;
  • runtime/platform changes;
  • performance pressure;
  • configuration-contract changes.

Generation N inherits the actual code produced by generation N-1.

Patch Life Report

ChronoForge reports:

  • Temporal Survival Curve
  • Future Maintenance Cost
  • Maintenance Option Value
  • Invariant Survival
  • Dependency Robustness
  • Patch Surface Stability
  • repository-generation Half-Life
  • Temporal Death Modes

Reality calibration

Later real repository events can be recorded with:

lr-agent chrono-observe dependency_upgrade "framework major-version migration"

Those observations affect future scenario scheduling weights.

Causal Genome research stack

v1.1.0 also includes the Causal Genome mechanisms introduced in the 1.0 line:

  • quarantine-first Strategy Genes;
  • Treatment vs Control ablation;
  • active falsification;
  • Anti-Genes;
  • genealogy / contamination propagation;
  • Proof-Carrying Genes;
  • Invariant DNA;
  • Epistemic Tripwire.

Quick start

git clone https://github.com/LLR6/LR-agent.git
cd LR-agent
python -m venv .venv
pip install -e ".[dev]"

Configure an OpenAI-compatible endpoint/model in .env, then:

lr-agent doctor
lr-agent serve

Try ChronoForge:

lr-agent chrono "preserve the current core behavior" --generations 4 --trajectories 3

Demo

The animation is explicitly a deterministic explanatory showcase, not a benchmark result.

See:

Research status

ChronoForge is a research prototype.

Synthetic futures are stress scenarios, not guaranteed forecasts. Maintenance Option Value is an engineering heuristic, not a formal maintainability proof.

The next major research target is Chrono Tournament: exposing multiple present-time-correct patches to the exact same frozen Future Matrix.

Feedback wanted

The most useful feedback is a reproducible counterexample:

  • a patch ChronoForge scores highly but is clearly fragile;
  • a low-scoring patch that later proves easy to maintain;
  • a Causal Genome strategy that should not transfer;
  • an Invariant or verifier design that exposes a blind spot.
Assets 3
Loading