legion-llm 0.4.1 → 0.4.5

checksums.yaml

@@ -1,7 +1,7 @@
 ---
 SHA256:
-  metadata.gz: 2337a94da9139d5ba6bb4ddbc4ee3d11a7caab4b5502e85c1fe08c5822d9bd7c
-  data.tar.gz: bab4d942d0845165167fbc15bed8bd07b7e52b3475f6b328b49bb6a2fde0d9a5
+  metadata.gz: 8402009cad8567d64881d1ab4bab1c05b863a0cfa606b4d43da55f412ab04e3c
+  data.tar.gz: '0588be98c5a45fd930a5e668ec6ce1789bb9e927ef0c8748208b5dd211b89326'
 SHA512:
-  metadata.gz: c685d7db577f7be3ea567c7fd79eb09faaa4f9f6907ddd86c460e3c640a5573e0c15658fe7eb9b89500383a00b0d65e2a99aac7e1a454d1f2caa3877d835f261
-  data.tar.gz: 9439839a44b57f05e45b310ba6b1bcbb49a115f959292a6c31a2943d22f11584d2cd092963a9f79a625e25794cf6bb21506e0709da64a0e4d8f11413878497a1
+  metadata.gz: ad633666c8e5d3b25d140f9be94741997b28704a84e35aba33b87eac334a99ae2c6c4829a9caf32aae855eef510d82cb401451f9717a3f64ecef62c649ba8fba
+  data.tar.gz: 7947ea6547260b4e16c65fe92bd4154da6b2cd892dcf7097de240f97fc0e851f0fccc416224305ab6220bfd7317bb540babaa89f5c554303fe095d3f122513c7

data/CHANGELOG.md

@@ -1,5 +1,34 @@
 # Legion LLM Changelog
 
+## [0.4.5] - 2026-03-23
+
+### Fixed
+- `llm_chat` and `llm_session` helpers now accept and forward `caller:` to pipeline (unblocks consumer migration)
+
+## [0.4.4] - 2026-03-23
+
+### Added
+- `Pipeline::Steps::Rbac`: real RBAC enforcement using `Legion::Rbac.authorize!`, graceful degradation when unavailable
+- `Pipeline::Steps::Classification`: real PII/PHI scan (SSN, email, phone regex + PHI keyword list); upgrade-only classification levels
+- `Pipeline::Steps::Billing`: real budget enforcement via `CostEstimator`; spending cap rejection
+
+### Changed
+- Extracted all three governance stubs from Executor into dedicated step modules
+
+## [0.4.2] - 2026-03-23
+
+### Added
+- `Pipeline::Steps::RagContext` (step 8): context strategy selector (full/rag_hybrid/rag/none) based on utilization, queries Apollo via `retrieve_relevant`
+- `Pipeline::Steps::RagGuard`: post-response faithfulness check against retrieved RAG context via `Hooks::RagGuard`
+- `Pipeline::EnrichmentInjector`: converts RAG and GAIA enrichments into system prompt text before provider call
+- `Pipeline::GaiaCaller`: privileged helper for GAIA/GAS LLM calls with system profile (skips governance steps)
+- `Pipeline::AuditPublisher`: publishes audit events to `llm.audit` exchange for GAS subscriber consumption
+- RAG/GAS full cycle integration test (4 examples: enrichment, injection, degradation, feedback loop prevention)
+- `OverrideConfidence` module with 4-tier degrading storage (L0 memory, L1 cache, L2 SQLite, L3 Apollo)
+- Catalog-driven auto-override in `ToolDispatcher`: settings override first, then Catalog + confidence gate
+- Shadow mode execution: when confidence is 0.5-0.8, execute both MCP and LEX, compare results, update confidence
+- `hydrate_from_l2` and `hydrate_from_apollo` for override confidence persistence across restarts
+
 ## [0.4.1] - 2026-03-23
 
 ### Added

data/CLAUDE.md

@@ -325,7 +325,7 @@ In-memory signal consumer with pluggable handlers. Adjusts effective priorities
 | `lib/legion/llm/structured_output.rb` | JSON schema enforcement with native response_format and prompt fallback |
 | `lib/legion/llm/errors.rb` | Typed error hierarchy: LLMError base + AuthError, RateLimitError, ContextOverflow, ProviderError, ProviderDown, UnsupportedCapability, PipelineError |
 | `lib/legion/llm/conversation_store.rb` | ConversationStore: in-memory LRU (256 slots) + optional Sequel DB persistence + spool fallback |
-| `lib/legion/llm/version.rb` | Version constant (0.4.1) |
+| `lib/legion/llm/version.rb` | Version constant (0.4.2) |
 | `lib/legion/llm/quality_checker.rb` | QualityChecker module with QualityResult struct |
 | `lib/legion/llm/escalation_history.rb` | EscalationHistory mixin: `escalation_history`, `escalated?`, `final_resolution`, `escalation_chain` |
 | `lib/legion/llm/router/escalation_chain.rb` | EscalationChain value object |
@@ -339,6 +339,9 @@ In-memory signal consumer with pluggable handlers. Adjusts effective priorities
 | `lib/legion/llm/pipeline/timeline.rb` | Pipeline::Timeline: ordered event recording |
 | `lib/legion/llm/pipeline/executor.rb` | Pipeline::Executor: 18-step skeleton with profile-aware execution |
 | `lib/legion/llm/pipeline/steps/metering.rb` | Pipeline::Steps::Metering: metering event builder |
+| `lib/legion/llm/pipeline/steps/rag_context.rb` | Pipeline::Steps::RagContext: context strategy selection and Apollo retrieval (step 8) |
+| `lib/legion/llm/pipeline/steps/rag_guard.rb` | Pipeline::Steps::RagGuard: faithfulness check against retrieved RAG context |
+| `lib/legion/llm/pipeline/enrichment_injector.rb` | Pipeline::EnrichmentInjector: converts RAG/GAIA enrichments into system prompt |
 | `lib/legion/llm/cost_estimator.rb` | CostEstimator: model cost estimation with fuzzy pricing |
 | `lib/legion/llm/fleet.rb` | Fleet module: requires dispatcher, handler, reply_dispatcher |
 | `lib/legion/llm/fleet/dispatcher.rb` | Fleet::Dispatcher: fleet RPC dispatch |
@@ -385,6 +388,10 @@ In-memory signal consumer with pluggable handlers. Adjusts effective priorities
 | `spec/legion/llm/pipeline/steps/metering_spec.rb` | Tests: Metering event building |
 | `spec/legion/llm/fleet/dispatcher_spec.rb` | Tests: Fleet dispatch, availability, timeout |
 | `spec/legion/llm/fleet/handler_spec.rb` | Tests: Fleet handler, auth, response building |
+| `spec/legion/llm/pipeline/steps/rag_context_spec.rb` | Tests: RAG context strategy selection, Apollo retrieval, graceful degradation |
+| `spec/legion/llm/pipeline/steps/rag_guard_spec.rb` | Tests: RAG faithfulness checking |
+| `spec/legion/llm/pipeline/enrichment_injector_spec.rb` | Tests: enrichment injection into system prompt |
+| `spec/legion/llm/pipeline/rag_gas_integration_spec.rb` | Tests: RAG/GAS full cycle integration |
 | `spec/spec_helper.rb` | Stubbed Legion::Logging and Legion::Settings for testing |
 
 ## Extension Integration
@@ -444,8 +451,8 @@ The legacy `vault_path` per-provider setting was removed in v0.3.1.
 Tests run without the full LegionIO stack. `spec/spec_helper.rb` stubs `Legion::Logging` and `Legion::Settings` with in-memory implementations. Each test resets settings to defaults via `before(:each)`.
 
 ```bash
-bundle exec rspec    # 712 examples, 0 failures
-bundle exec rubocop  # 113 files, 0 offenses
+bundle exec rspec    # 794 examples, 0 failures
+bundle exec rubocop  # 142 files, 0 offenses
 ```
 
 ## Design Documents