save checkpoint

This commit is contained in:
master
2026-02-14 09:11:48 +02:00
parent 9ca2de05df
commit e9aeadc040
1512 changed files with 30863 additions and 4728 deletions

View File

@@ -0,0 +1,31 @@
# Natural Language to Policy Rule Compiler (Policy Studio Copilot)
## Module
AdvisoryAI
## Status
IMPLEMENTED
## Description
AI-powered natural language to lattice rule translation engine including PolicyIntentType parsing, LatticeRuleGenerator, property-based test synthesizer for generated rules, and PolicyBundleCompiler. Transforms plain-English policy descriptions into formal stella-dsl@1 rules with live preview and conflict visualization.
## Implementation Details
- **Modules**: `src/AdvisoryAi/StellaOps.AdvisoryAI/PolicyStudio/`
- **Key Classes**:
- `AiPolicyIntentParser` (`src/AdvisoryAi/StellaOps.AdvisoryAI/PolicyStudio/AiPolicyIntentParser.cs`) - parses natural language into structured policy intents using LLM
- `LatticeRuleGenerator` (`src/AdvisoryAi/StellaOps.AdvisoryAI/PolicyStudio/LatticeRuleGenerator.cs`) - generates K4 lattice rules from parsed policy intents
- `PropertyBasedTestSynthesizer` (`src/AdvisoryAi/StellaOps.AdvisoryAI/PolicyStudio/PropertyBasedTestSynthesizer.cs`) - synthesizes property-based test cases for generated rules
- `PolicyBundleCompiler` (`src/AdvisoryAi/StellaOps.AdvisoryAI/PolicyStudio/PolicyBundleCompiler.cs`) - compiles generated rules into a deployable policy bundle
- `PolicyIntent` (`src/AdvisoryAi/StellaOps.AdvisoryAI/PolicyStudio/PolicyIntent.cs`) - policy intent model with type, constraints, and conditions
- `InMemoryPolicyIntentStore` (`src/AdvisoryAi/StellaOps.AdvisoryAI/PolicyStudio/InMemoryPolicyIntentStore.cs`) - stores policy intents for iterative refinement
- `NullPolicyIntentParser` (`src/AdvisoryAi/StellaOps.AdvisoryAI/PolicyStudio/NullPolicyIntentParser.cs`) - no-op parser for testing
- **Interfaces**: `IPolicyIntentParser`, `IPolicyRuleGenerator`, `ITestCaseSynthesizer`
- **Source**: SPRINT_20251226_017_AI_policy_copilot.md
## E2E Test Plan
- [ ] Submit a natural language policy description (e.g., "block critical CVEs without a fix") and verify `AiPolicyIntentParser` produces a structured `PolicyIntent`
- [ ] Verify `LatticeRuleGenerator` translates the intent into valid stella-dsl@1 lattice rules
- [ ] Verify `PropertyBasedTestSynthesizer` generates test cases that exercise the generated rule's accept/reject boundaries
- [ ] Verify `PolicyBundleCompiler` compiles rules into a deployable bundle with correct schema version
- [ ] Verify `InMemoryPolicyIntentStore` supports iterative refinement: modify an intent and regenerate rules
- [ ] Verify conflict detection: generate two conflicting rules and verify the compiler reports the conflict