Skip to content

Build production-oriented AionBench benchmark - #1

Merged
tensorlink-dev merged 1 commit into
mainfrom
agent/build-production-benchmark
Jul 30, 2026
Merged

tensorlink-dev merged 1 commit into
mainfrom
agent/build-production-benchmark

Conversation

@tensorlink-dev

Copy link
Copy Markdown
Contributor

What changed

  • builds the executable AionBench 0.1 vertical slice across capacity, inventory, and workforce domains
  • adds deterministic episode generation, SQLite worlds, operational actions, oracles, regret and safety scoring
  • adds bounded agent execution, complete run bundles, replay and scripted-oracle agents, batch evaluation, paired bootstrap statistics, and reports
  • integrates the pinned Aion 0.2.0 CLI contract with artifact capture and determinism verification
  • adds 12 development episodes, 29 benchmark tests, container packaging, CI, security guidance, and release-readiness documentation

Why

The repository previously contained only the build specification. This creates the production-oriented alpha needed to execute and validate the benchmark while explicitly documenting the remaining external release blockers.

Validation

  • 29 passed in AionBench
  • 152 passed in the pinned Aion runtime
  • 12/12 episodes validated
  • 48-run oracle matrix: zero regret, Success@1/ Reliable@4/ Safe@4 = 1.0
  • 100 repeated Aion forecasts produced equivalent numerical artifacts
  • wheel and source distribution built successfully
  • Docker image built and validated with networking disabled

@tensorlink-dev
tensorlink-dev merged commit 1cac502 into main Jul 30, 2026
4 checks passed
@tensorlink-dev
tensorlink-dev deleted the agent/build-production-benchmark branch July 30, 2026 06:16
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

1 participant