bc-telemetry-buddy-mcp 3.1.2 → 3.1.4

package/CHANGELOG.md

@@ -7,6 +7,23 @@ and this project adheres to [Semantic Versioning](https://semver.org/spec/v2.0.0
 
 ## [Unreleased]
 
+## [3.1.4] - 2026-03-02
+
+### Improved
+- **Teams Adaptive Card Tables**: Agent notifications now convert markdown tables in the alert message to native Adaptive Card `Table` elements (schema 1.5) with styled headers and grid lines, instead of rendering them as plain text. Non-table text is preserved as `TextBlock` with markdown support.
+- **CI-Aware Collapsible Groups**: Tool call execution in the ReAct loop is now wrapped in collapsible groups — `##[group]`/`##[endgroup]` in Azure DevOps, `::group::`/`::endgroup::` in GitHub Actions. Internal kusto/auth/cache logs are collapsed; tool names, reasoning, and result summaries stay visible at top level.
+- **Retry Visibility**: LLM retry attempts (429/529/503 backoff) are now logged to `stdout` with a 🔄 icon so they appear in the normal pipeline output stream, not hidden in `stderr`.
+
+## [3.1.3] - 2026-03-02
+
+### Fixed
+- **JSON Repair for Agent Output**: `parseAgentOutput()` now repairs common LLM JSON mistakes (trailing commas, unescaped newlines, truncated output) instead of failing. Three-stage repair: strict parse → fix trailing commas/unescaped chars → close truncated brackets. Logs warning when repair succeeds.
+
+### Improved
+- **Verbose Agent Runtime Logging**: ReAct loop now logs iteration numbers, LLM token usage, reasoning text, tool call names with argument summaries, result previews (row counts for queries), and final assessment — making agent behavior visible in CI/CD pipeline output
+- **KQL Query Logging**: Query success messages now include row count, column count, and duration (e.g., `1 table(s), 42 row(s), 8 column(s) in 1230ms`)
+- **Run-All Agent Output**: Per-agent summary now includes tool call count, token stats, and executed actions
+
 ## [3.1.2] - 2026-03-02
 
 ### Added

package/README.md

@@ -344,6 +344,15 @@ cp -r node_modules/bc-telemetry-buddy-mcp/templates/agents/performance-monitorin
 bctb-mcp agent run performance-monitoring --once
 ```
 
+### CI/CD Pipeline Output
+
+Agent runs produce structured, CI-friendly output:
+
+- **Top-level visibility**: Iteration numbers, tool names, LLM reasoning, and result summaries (row/column counts, durations) are always visible
+- **Collapsible detail groups**: Internal execution logs (Kusto authentication, cache operations, query URLs) are wrapped in collapsible sections — `##[group]`/`##[endgroup]` in Azure DevOps, `::group::`/`::endgroup::` in GitHub Actions
+- **Retry visibility**: LLM API retries (429/529/503) are shown inline with backoff timing
+- **Teams notifications**: Markdown tables in alert messages are automatically converted to native Adaptive Card `Table` elements for proper rendering in Teams
+
 See the [User Guide](https://github.com/waldo1001/waldo.BCTelemetryBuddy/blob/main/docs/UserGuide.md#agentic-monitoring) for full docs including action types, state management, and troubleshooting.
 
 ## Features