Retire Observability packages and document OpenTelemetry migration

.claude/agents/architecture-reviewer.md

@@ -231,7 +231,6 @@ Escalate to team discussion when:
 
 ### Dependency Flow
 ```
-runtime → observability → observability-hosting → observability-extensions-*
 runtime → tooling → tooling-extensions-*
 runtime → notifications
 ```
@@ -240,16 +239,15 @@ runtime → notifications
 | Package | Responsibility |
 |---------|---------------|
 | runtime | Foundation utilities, no external SDK deps |
-| observability | OpenTelemetry tracing infrastructure |
 | tooling | MCP server configuration and discovery |
 | notifications | Agent lifecycle event handling |
 | *-extensions-* | Framework-specific adapters |
 
 ### Design Patterns
 | Pattern | Usage |
 |---------|-------|
-| Singleton | ObservabilityManager (single tracer provider) |
-| Disposable | Scope classes (automatic span lifecycle) |
-| Builder | BaggageBuilder, ObservabilityBuilder (fluent config) |
+| Configuration Provider | Runtime and Tooling dynamic configuration |
 | Strategy | McpToolServerConfigurationService (dev vs prod) |
 | Extension Methods | Notifications extending AgentApplication |
+
+Observability packages are retired. Application telemetry belongs to `@microsoft/opentelemetry`; consult `docs/observability-migration.md`, not deleted SDK APIs.

.claude/agents/code-review-manager.md

@@ -271,12 +271,12 @@ When the task-implementer completes work:
 I've completed the implementation for [feature]. Please review:
 
 Files created:
-- packages/agents-a365-observability/src/customProcessor.ts
-- tests/observability/customProcessor.test.ts
+- packages/agents-a365-tooling/src/discoveryRetry.ts
+- tests/tooling/discoveryRetry.test.ts
 
 Files modified:
-- packages/agents-a365-observability/src/index.ts
-- packages/agents-a365-observability/src/ObservabilityBuilder.ts
+- packages/agents-a365-tooling/src/index.ts
+- packages/agents-a365-tooling/src/McpToolServerConfigurationService.ts
 
 Test results: 24 passed, 0 failed
 Lint: No errors

.claude/agents/code-reviewer.md

@@ -311,28 +311,12 @@ try {
 
 ## SDK-Specific Patterns
 
-### Disposable Pattern
+### Configuration Providers
 ```typescript
-// ✅ Correct usage with 'using' keyword
-using scope = InvokeAgentScope.start(details, tenantDetails);
-// Span automatically ends when scope is disposed
-```
+import { defaultRuntimeConfigurationProvider } from '@microsoft/agents-a365-runtime';
 
-### Builder Pattern
-```typescript
-// ✅ Fluent API usage
-const baggage = new BaggageBuilder()
-  .tenantId(tenantId)
-  .agentId(agentId)
-  .correlationId(correlationId)
-  .build();
+const config = defaultRuntimeConfigurationProvider.getConfiguration();
+const cluster = config.clusterCategory;
 ```
 
-### Singleton Access
-```typescript
-// ✅ Proper singleton usage
-const manager = ObservabilityManager.getInstance();
-if (manager === null) {
-  throw new Error('ObservabilityManager not initialized');
-}
-```
+Evaluate dynamic getter values at their point of use; overrides may vary per request. Observability is retired from this SDK; application integrations use the distro described in `docs/observability-migration.md`.

.claude/agents/pr-comment-resolver.md

@@ -70,11 +70,11 @@ Types: `fix`, `refactor`, `style`, `docs`, `test`
 
 Example:
 ```
-fix(observability): add null check for tenant ID
+fix(runtime): add null check for tenant ID
 
 Addresses review comment [CRM-003]
 - Added explicit null check before accessing tenantId
-- Prevents potential runtime error when baggage is missing
+- Prevents potential runtime error when request identity is missing
 ```
 
 ### Phase 3: Verification

.claude/agents/prd-task-generator.md

@@ -79,7 +79,6 @@ When generating tasks, consider:
 
 ### Package Placement
 - **Runtime**: Shared utilities, no external SDK dependencies
-- **Observability**: OpenTelemetry integration, tracing scopes
 - **Tooling**: MCP server configuration and discovery
 - **Notifications**: Agent lifecycle event handling
 - **Extensions**: Framework-specific adapters (OpenAI, Claude, LangChain)
@@ -88,7 +87,6 @@ When generating tasks, consider:
 - Use workspace protocol: `"@microsoft/agents-a365-runtime": "workspace:*"`
 - Understand the dependency flow:
   ```
-  runtime → observability → observability-hosting → observability-extensions-*
   runtime → tooling → tooling-extensions-*
   runtime → notifications
   ```
@@ -99,9 +97,8 @@ When generating tasks, consider:
 - Build output: `dist/cjs/`, `dist/esm/`
 
 ### Observability Integration
-- New features should emit appropriate spans
-- Use existing scope classes: `InvokeAgentScope`, `InferenceScope`, `ExecuteToolScope`
-- Propagate baggage context for correlation
+- Application telemetry is owned by the external `@microsoft/opentelemetry` distro.
+- Follow `docs/observability-migration.md`; do not recreate retired scope implementations or add the distro to unaffected SDK packages.
 
 ## Deliverable Format
 
@@ -167,25 +164,25 @@ Before finalizing, verify:
 
 ## Example Output
 
-**Input**: PRD for adding custom span processors to observability
+**Input**: PRD for adding configurable MCP discovery retries to Tooling
 
 **Output**:
 ```markdown
-# Implementation Plan: Custom Span Processors
+# Implementation Plan: MCP Discovery Retries
 
 ## Executive Summary
-This plan breaks down the custom span processor feature into 6 tasks across 3 phases. The implementation adds a plugin architecture to ObservabilityBuilder while maintaining backward compatibility.
+This plan breaks down configurable discovery retries into 6 tasks across 3 phases, preserving existing defaults and request-specific configuration.
 
 ## Architecture Impact
-- **Primary Package**: `@microsoft/agents-a365-observability`
+- **Primary Package**: `@microsoft/agents-a365-tooling`
 - **Dependencies**: None new
 - **Breaking Changes**: None (additive API)
 
 ## Phase 1: Foundation (2 tasks)
 
-### Task 1.1: Define SpanProcessor Interface
-**Package**: `@microsoft/agents-a365-observability`
-**Description**: Create the interface contract for custom span processors...
+### Task 1.1: Define Retry Configuration
+**Package**: `@microsoft/agents-a365-tooling`
+**Description**: Define bounded retries and backoff with existing configuration-provider patterns...
 ...
 
 ## Phase 2: Core Implementation (2 tasks)

.claude/agents/prd-writer.md

@@ -32,15 +32,14 @@ Systematically extract:
 ### 2. Contextual Awareness
 
 Maintain alignment with the SDK's architecture:
-- **Monorepo structure**: 9 interdependent packages in `packages/`
+- **Monorepo structure**: Six maintained packages in `packages/`
 - **Core + Extensions pattern**: Framework-agnostic core with framework-specific extensions
 - **Package categories**:
   - Runtime: Foundation utilities
-  - Observability: OpenTelemetry tracing
   - Tooling: MCP server configuration
   - Notifications: Agent lifecycle events
 - **TypeScript standards**: Strict typing, interfaces over types where appropriate
-- **Design patterns**: Singleton, Disposable, Builder, Strategy, Extension Methods
+- **Design patterns**: Configuration Provider, Strategy, Extension Methods
 - **Async/await conventions**: Promise-based APIs for I/O operations
 - **Dual module format**: Both CJS and ESM builds required
 
@@ -49,7 +48,7 @@ Maintain alignment with the SDK's architecture:
 Before drafting, ask targeted questions:
 
 1. "What problem does this solve for developers?"
-2. "Which packages are affected?" (runtime, observability, tooling, notifications, extensions)
+2. "Which packages are affected?" (runtime, tooling, notifications, tooling extensions)
 3. "Does this extend core functionality or require a new extension package?"
 4. "What are the measurable success metrics?"
 5. "Are there security or compliance considerations?"
@@ -97,6 +96,7 @@ Generate documents with these sections:
 - Usage examples
 
 ### 8. Observability
+- Application-owned telemetry uses the external Microsoft OpenTelemetry distro; see `docs/observability-migration.md`. No telemetry package is maintained here.
 - Tracing integration requirements
 - Span attributes and events
 - Metrics to capture
@@ -173,15 +173,15 @@ PRDs must be:
 
 ## Example Interaction
 
-**User**: "I want to add support for custom span processors in the observability package"
+**User**: "I want to add configurable MCP discovery retries in the tooling package"
 
 **PRD Writer Response**:
 "Before I draft the PRD, I have a few clarifying questions:
 
-1. Should custom processors run before or after the built-in Agent365Exporter?
-2. Will this require changes to the ObservabilityBuilder API?
-3. Should processors have access to the baggage context (tenant ID, agent ID)?
-4. Are there specific use cases driving this? (e.g., filtering, sampling, custom export)
+1. Which transient discovery failures should be retried?
+2. Will this require changes to ToolingConfiguration?
+3. Should retry settings vary per tenant/request?
+4. What latency and attempt limits are required?
 5. Should this be backward-compatible with existing configurations?
 
 Once I understand these aspects, I'll create a comprehensive PRD covering the API design, integration points, and testing strategy."

.claude/agents/task-implementer.md

@@ -39,9 +39,7 @@ Before writing code:
 Determine:
 - **Package placement**: Core package or extension?
 - **Pattern alignment**: Which design patterns apply?
-  - Singleton (ObservabilityManager)
-  - Disposable (Scope classes with `using` keyword)
-  - Builder (BaggageBuilder, ObservabilityBuilder)
+  - Configuration Provider (Runtime and Tooling dynamic overrides)
   - Strategy (environment-based configuration)
   - Extension Methods (TypeScript declaration merging)
 - **Cross-package impacts**: Dependencies, shared types
@@ -201,10 +199,10 @@ Before requesting code review, verify:
 
 - **Node.js versions**: >=18.0.0
 - **TypeScript**: Strict mode enabled
-- **Structure**: pnpm workspace monorepo with 9 packages
+- **Structure**: pnpm workspace monorepo with six maintained packages
 - **Package naming**: `@microsoft/agents-a365-<name>`
 - **Build output**: Dual format (CJS and ESM)
-- **Observability**: OpenTelemetry for traces, spans, metrics
+- **Application telemetry**: External `@microsoft/opentelemetry` distro; see `docs/observability-migration.md`. Do not reintroduce retired SDK packages.
 - **Tool integration**: MCP (Model Context Protocol)
 - **Testing**: Jest with ts-jest preset
 

.claude/agents/test-coverage-reviewer.md

@@ -61,9 +61,6 @@ Provide specific recommendations:
 ### Test Directory Structure
 ```
 tests/
-├── observability/
-│   ├── BaggageBuilder.test.ts
-│   └── InvokeAgentScope.test.ts
 ├── runtime/
 │   └── Utility.test.ts
 ├── tooling/

.github/CODEOWNERS

@@ -1,8 +1,7 @@
 # This will eventually be replaced by security group
 * @microsoft/agent365-approvers
 /.github/ @microsoft/agent365-approvers
-/packages/agents-a365-observability*/ @microsoft/agent365-observability-approvers
-/tests/observability/ @microsoft/agent365-observability-approvers
+/docs/observability-migration.md @microsoft/agent365-observability-approvers
 /packages/agents-a365-notifications/ @microsoft/agent365-tooling-approvers
 /packages/agents-a365-tooling*/ @microsoft/agent365-tooling-approvers
 /tests/tooling*/ @microsoft/agent365-tooling-approvers

.github/workflows/ci.yml

@@ -79,94 +79,38 @@ jobs:
       with:
         version: 10.20.0
 
+    - name: Match pnpm runtime to selected Node.js
+      run: pnpm config set --location=project useNodeVersion "$(node -p 'process.versions.node')"
+
     - name: Install dependencies
-      run: pnpm i
+      run: pnpm install --frozen-lockfile
 
     - name: Lint code
-      run: npm run lint
+      run: pnpm lint
 
     - name: Set version with Nerdbank.GitVersioning
       run: node setVersion.js
 
     - name: Build package
-      run: npm run build
+      run: pnpm build
+
+    - name: Type check remaining packages
+      run: pnpm typecheck
 
     - name: Run tests
-      run: npm test
+      env:
+        A365_EXPECTED_NODE_MAJOR: ${{ matrix.node-version }}
+      run: pnpm test
 
     - name: Pack packages
-      run: npm run pack
+      run: pnpm run pack artifacts/release-${{ matrix.node-version }}
 
-    # Copy packages to drop folder
-    - name: Copy packages to drop folder
-      run: |
-        mkdir -p ~/drop/nodejs-${{ matrix.node-version }} && cp -v packages/*.tgz $_
+    - name: Verify release artifacts independently
+      run: node packPackages.js --verify artifacts/release-${{ matrix.node-version }}
 
     - name: Upload NPM packages as artifacts
       uses: actions/upload-artifact@ea165f8d65b6e75b540449e92b4886f43607fa02 # v4.6.2
       with:
         name: nodejs-${{ matrix.node-version }}
-        path: ~/drop/nodejs-${{ matrix.node-version }}/*.tgz
-
-    # - name: Publish to NPM (commented for now)
-    #   if: github.event_name == 'push' && github.ref == 'refs/heads/main'
-    #   run: npm publish
-    #   env:
-    #     NODE_AUTH_TOKEN: ${{ secrets.NPM_TOKEN }}
-
-  integration-tests:
-    name: Integration tests
-    needs: [nodejs-sdk]
-    # Pull request workflows can be edited by the PR author, so the Azure OpenAI key lives in the
-    # integration-tests environment, which only allows deployments from main.
-    if: github.event_name == 'push'
-    runs-on: ubuntu-latest
-    environment: integration-tests
-    permissions:
-      contents: read
-
-    strategy:
-      matrix:
-        node-version: ['18', '20']
-
-    steps:
-    - name: Checkout repository
-      uses: actions/checkout@11d5960a326750d5838078e36cf38b85af677262 # v4.4.0
-      with:
-        fetch-depth: 0
-
-    - uses: pnpm/action-setup@b906affcce14559ad1aafd4ab0e942779e9f58b1 # v4.3.0
-      name: Install pnpm
-      with:
-        run_install: false
-
-    - name: Set up Node.js ${{ matrix.node-version }}
-      uses: actions/setup-node@249970729cb0ef3589644e2896645e5dc5ba9c38 # v6.5.0
-      with:
-        node-version: ${{ matrix.node-version }}
-        cache: 'pnpm'
-        cache-dependency-path: '**/pnpm-lock.yaml'
-
-    - name: Setup pnpm
-      uses: pnpm/action-setup@a3252b78c470c02df07e9d59298aecedc3ccdd6d # v3.0.0
-      with:
-        version: 10.20.0
-
-    - name: Install dependencies
-      run: pnpm i
-
-    - name: Set version with Nerdbank.GitVersioning
-      run: node setVersion.js
-
-    - name: Build package
-      run: npm run build
-
-    - name: Run integration tests
-      env:
-        AZURE_OPENAI_API_KEY: ${{ secrets.AZURE_OPENAI_API_KEY }}
-        AZURE_OPENAI_ENDPOINT: ${{ vars.AZURE_OPENAI_ENDPOINT }}
-        AZURE_OPENAI_DEPLOYMENT: ${{ vars.AZURE_OPENAI_DEPLOYMENT }}
-        AZURE_OPENAI_API_VERSION: ${{ vars.AZURE_OPENAI_API_VERSION }}
-        ENABLE_OBSERVABILITY: ${{ vars.ENABLE_OBSERVABILITY }}
-      run: |
-        npm run test:integration
+        path: artifacts/release-${{ matrix.node-version }}/
+        if-no-files-found: error

.vscode/settings.json

@@ -4,6 +4,6 @@
         "code-reviewer",
         "test-coverage-reviewer"
     ],
-    "jest.jestCommandLine": "node --experimental-vm-modules node_modules/jest/bin/jest.js --config=tests/jest.config.cjs",
+    "jest.jestCommandLine": "pnpm test",
     "jest.rootPath": "."
 }