luigidemasi opened a new pull request, #27649: URL: https://github.com/apache/camel/pull/27649
# Description Semantic evaluations currently expose metadata and interactive results, but no retained history or pluggable audit delivery. This adds opt-in auditing to the semantic DSL, with a context default and per-expert overrides, and exposes retained evidence through the Semantic TUI Audit view and `tui_get_audit` MCP tool. Evaluation records capture outcomes and their historical meanings. Routes explicitly record application decisions and link them to evaluation event IDs; a Boolean result is never translated into an allow/block action. Records retain the static provider identity and allowlisted model/revision strings, plus start and completion timestamps. Input, headers, arbitrary provider metadata and raw exceptions are excluded. - Separate sink and reader SPIs, with bounded memory history and JSON logging backends. Delivery is asynchronous and best effort, with bounded queues, eviction/drop/error reporting and a documented shutdown contract. - Independent observer hooks carry parent context across console dispatch. Audit can run without OpenTelemetry. Observations remain detached between callbacks; this does not add an OpenTelemetry exporter. - YAML, XML and Java semantic declarations support the same immutable configuration. This first implementation uses the semantic DSL as one configuration authority, with deployment properties available through placeholders. The configuration and reload contract is proposed here for upstream review; direct property-based configuration remains a valid alternative. Configuration freezes at service startup. Deleting its resource preserves the active audit configuration while removing obsolete declarations; changes to the settings require restarting the application. - Read-only console/connector queries and the TUI support filters, stable paging, selected records and linked evidence. Audit queries use workers independent of inference, with bulk evidence reads and coalesced TUI refreshes. `tui_get_audit` changes the visible tab, view, filters and selection; its read-only classification refers to integration data and capture configuration. Validation: 326 semantic tests, 46 connector tests and 32 YAML semantic integration tests passed; the full TUI suite passed (1,859 tests, one existing skip). The full 704-module reactor passed `clean install -DskipTests -Psourcecheck`, including formatting and import checks. Generated catalogs, documentation mirror and YAML schemas are included. Antora validation passed for the committed documentation. References: [CAMEL-25499](https://issues.apache.org/jira/browse/CAMEL-25499); builds on merged PR #27602. The change is split into three commits: runtime/DSL and generated artifacts; CLI connector; TUI/MCP and user help. # Target - [x] Targets `main` for Camel 4. # Tracking - [x] Tracked by [CAMEL-25499](https://issues.apache.org/jira/browse/CAMEL-25499). # Apache Camel coding standards and style - [x] Each commit has a meaningful subject and body. - [x] Ran the full root `clean install -DskipTests` build with source checks and committed all related generated changes. # AI-assisted contributions - [x] Commits include AI co-authorship trailers; this description identifies the AI tool and operator. _Generated by OpenAI Codex on behalf of @luigidemasi via /oss-fix-issue._ -- This is an automated message from the Apache Git Service. To respond to the message, please log on to GitHub and use the URL above to go to the specific comment. To unsubscribe, e-mail: [email protected] For queries about this service, please contact Infrastructure at: [email protected]
