How dpduc313/k5-l3a-day10-daidaidi-poctopilotgate1 works
This repository is a hands-on Gate 1 lab for deciding whether a simulated LogiX-VN refund assistant is ready for a customer pilot. The central workflow is to sign acceptance criteria, run the refund simulation and inspect its evidence, implement and checkpoint preventive gates, then evaluate an incident and record a Gate 1 decision. Supporting branches cover operational/legal limits, monitoring, incident grading, and submission completeness. Some modules are represented from the source index and README rather than sampled implementation; edges are limited to relationships visible in those sources.
Main entry points: eval.py Suite evaluation · ops.py Ops gate logic · criteria.py Acceptance criteria · agent.py Refund agent
Components 19
Criteria & Operations
- Acceptance criteria
lab10/criteria.py - Signed thresholds
config/acceptance_criteria.json - Operating limits
config/operating_gates.json - Ops gate logic
lab10/ops.py
Refund Simulation
- Refund agent
lab10/agent.py - Mock LLM
lab10/mock_llm.py - Read/write tools
lab10/tools.py - Preventive guards
lab10/guards.py - Policy helpers
lab10/helpers.py - Mock world
lab10/world.py - World and tickets
data/world.json
Evaluation & Gate
- Suite evaluation
lab10/eval.py - Custom monitors
lab10/monitors.py - Criteria comparison
lab10/gate.py
Incident & Grading
- Round-one snapshot
lab10/checkpoint.py - Incident grading
lab10/grade.py
Submission
- Submission checks
lab10/check_submission.py - Submission record
submission/submission.json
Other components
- Lab team
Connections 29
- Lab team → Signed thresholds: sets thresholds
README.md - Lab team → Preventive guards: implements guards
README.md - Lab team → Operating limits: sets limits
README.md - Acceptance criteria → Signed thresholds: signs and verifies
lab10/criteria.py - Refund agent → Mock LLM: extracts with
lab10/agent.py - Refund agent → Read/write tools: uses tools
lab10/agent.py - Refund agent → Ops gate logic: uses gate defaults
lab10/agent.py - Read/write tools → Preventive guards: checks writes
lab10/tools.py - Read/write tools → Mock world: reads and writes
lab10/tools.py - Preventive guards → Policy helpers: uses facts
lab10/guards.py - Ops gate logic → Mock world: uses world facts
lab10/ops.py - Ops gate logic → Operating limits: loads limits
lab10/ops.py - Mock world → World and tickets: loads base data
lab10/world.py - Suite evaluation → Refund agent: runs trials
lab10/eval.py - Suite evaluation → Mock world: loads simulation data
lab10/eval.py - Suite evaluation → Acceptance criteria: checks signing state
lab10/eval.py - Suite evaluation → Ops gate logic: computes operations
lab10/eval.py - Suite evaluation → Custom monitors: runs monitors
lab10/eval.py - Criteria comparison → Acceptance criteria: loads signed rules
lab10/gate.py - Lab team → Round-one snapshot: freezes round one
README.md - Round-one snapshot → Preventive guards: snapshots guards
lab10/checkpoint.py - Round-one snapshot → Ops gate logic: snapshots gate config
lab10/checkpoint.py - Incident grading → Suite evaluation: compares stages
lab10/grade.py - Incident grading → Round-one snapshot: uses prepared stage
lab10/grade.py - Incident grading → Acceptance criteria: checks criteria
lab10/grade.py - Lab team → Submission record: records decisions
README.md - Submission checks → Submission record: validates record
lab10/check_submission.py - Submission checks → Incident grading: collects grade report
lab10/check_submission.py - Submission checks → Acceptance criteria: checks team criteria
lab10/check_submission.py