cyberneticlibrary

Expand camera test fixtures adversarially

camera-parity-expansionworkflowsetup L34
lazyxeon/AstraWeave-AI-Native-Gaming-Engine
What it does

Generate camera test fixtures adversarially and verify under load

Best for

Generating and hardening camera test fixtures when baseline math is known but GPU output can drift across optimization passes.

Inputs
  • · fixture family specifications (matrix, cinematics, orbit, freefly types)
  • · camera API surface snapshot (Unified Camera campaign C.8)
Outputs
  • · generated fixture code (one agent per family, concrete test cases + baselines)
  • · survived fixtures (passed adversarial refute protocol)
  • · parity suite results + non-regression verdict
  • · landed fixtures ready for human review (NO commit)
Requires
  • · Rust compiler + cargo (build, test)
  • · parallel agent fleet (generate, adversarial refute × 3-lens, verify)
  • · bash shell
Preconditions
  • · Unified Camera codebase initialized
  • · Renderer + camera API accessible
  • · Cargo + Rust toolchain installed
  • · Previous fixture families as reference (parity test file structure)
Failure modes
  • · Baseline derivation fails (camera math incorrect) → fixture discarded, agent details drift
  • · Adversarial refute 3-lens returns null — retry logic (minimal-fallback) re-attempts (load-bearing for concurrency)
  • · Parity suite timeout — non-regression gate blocked, fixtures remain pending human review
  • · GPT output diverges from Rust semantics — skeptic protocol catches, fixture demoted to review
Trust signals
  • · Hardened skeptic protocol: 3 perspective-diverse lenses, 2 rich-schema attempts + 2 minimal-fallback per lens before terminal null
  • · First-run ~14% null-rate found transient on re-run hardening (not controlled A/B, but directional)
  • · Retry + minimal-fallback is load-bearing insurance, not optional — essential under concurrency
  • · Camera API snapshot cited (Unified Camera C.8 close) — agents re-verify against current source