ci.yml

name: CI

on:
  push:
    branches:
      - main
      - dev
      - release/**

  pull_request:
    types: [opened, synchronize, reopened]
    branches:
      - main
      - dev
      - release/**

  workflow_dispatch:

env:
  DOTNET_CLI_TELEMETRY_OPTOUT: "1"
  DOTNET_NOLOGO: "true"
  DOTNET_VERSION: 10.0.x
  TEST_PROJECT: ".\\SharpClaw.Tests\\SharpClaw.Tests.csproj"

permissions:
  contents: read

jobs:
  correctness:
    name: Correctness / ${{ matrix.domain }}
    runs-on: windows-latest
    timeout-minutes: 45
    strategy:
      fail-fast: false
      matrix:
        include:
          - domain: API Key Provider
            filter: 'FullyQualifiedName~SharpClaw.Tests.Api'
          - domain: Architecture Core Dependencies
            filter: 'FullyQualifiedName~SharpClaw.Tests.Architecture.CoreDependencyGuardrailTests'
          - domain: Architecture Module Dependencies
            filter: 'FullyQualifiedName~SharpClaw.Tests.Architecture.ModuleDependencyGuardrailTests'
          - domain: Clients Finish Reasons
            filter: 'FullyQualifiedName~SharpClaw.Tests.Clients'
          - domain: Core Header Tags
            filter: 'FullyQualifiedName~SharpClaw.Tests.Services.HeaderTagProcessorTests'
          - domain: Core Chat Cache
            filter: 'FullyQualifiedName~SharpClaw.Tests.Services.ChatCacheTests'
          - domain: CLI Channel Commands
            filter: 'FullyQualifiedName~SharpClaw.Tests.Cli.ChannelCliCommandTests'
          - domain: CLI REPL
            filter: 'FullyQualifiedName~SharpClaw.Tests.Cli.CliReplSmokeTests'
          - domain: Gateway Abstraction Boundaries
            filter: 'FullyQualifiedName~SharpClaw.Tests.Gateway.GatewayAbstractionsIsolationTests'
          - domain: Gateway Internal API Client
            filter: 'FullyQualifiedName~SharpClaw.Tests.Gateway.InternalApiClientResolutionTests'
          - domain: Gateway Synthetic Modules
            filter: 'FullyQualifiedName~SharpClaw.Tests.Gateway.SyntheticGatewayModuleTests'
          - domain: Gateway Endpoint Catalog
            filter: 'FullyQualifiedName~SharpClaw.Tests.Gateway.GatewayEndpointGroupCatalogTests'
          - domain: Gateway Endpoint Toggles
            filter: 'FullyQualifiedName~SharpClaw.Tests.Gateway.GatewayEndpointToggleTests'
          - domain: Gateway Route Parity
            filter: 'FullyQualifiedName~SharpClaw.Tests.Gateway.GatewayRouteParityTests'
          - domain: Gateway Lifecycle
            filter: 'FullyQualifiedName~SharpClaw.Tests.Gateway.GatewayInstanceLifecycleTests'
          - domain: Gateway Hot Reload
            filter: 'FullyQualifiedName~SharpClaw.Tests.Gateway.HotReload.HotReloadHostManagerTests'
          - domain: Gateway Chat Surface
            filter: 'FullyQualifiedName~SameSpoofResponseTextMatchesCoreApiSseAndGatewaySurfaces|FullyQualifiedName~GatewaySseProxy_ForwardsRealHttpSsePath'
          - domain: Gateway Chat Errors And Cancellation
            filter: 'FullyQualifiedName~GatewayErrorEnvelopeShapeIsStable|FullyQualifiedName~MidStreamFailureErrorEnvelopeIsStableInSse|FullyQualifiedName~DeterministicTimeoutCancellationDoesNotLeaveExecutingJobsStuck'
          - domain: API Job Contracts
            filter: 'FullyQualifiedName~ApiJob'
          - domain: Gateway Job Contracts
            filter: 'FullyQualifiedName~GatewayJob'
          - domain: Frontend Instances
            filter: 'FullyQualifiedName~SharpClaw.Tests.Frontend.FrontendInstanceServiceTests'
          - domain: Frontend Gateway Process
            filter: 'FullyQualifiedName~SharpClaw.Tests.Frontend.GatewayProcessManagerTests'
          - domain: Frontend Uno Client State
            filter: 'FullyQualifiedName~SharpClaw.Tests.Frontend.UnoClientStateTests'
          - domain: Persistence Provider Configuration
            filter: 'FullyQualifiedName~SharpClaw.Tests.Persistence.ApiKeyEncryptorBytesTests|FullyQualifiedName~SharpClaw.Tests.Persistence.DatabaseProviderOptionsTests'
          - domain: JSONColdStore Provider Paths
            filter: 'FullyQualifiedName~SharpClaw.Tests.Services.AuthServiceJsonColdStoreTests|FullyQualifiedName~SharpClaw.Tests.Modules.BundledModuleStorageGatewayTests|FullyQualifiedName~SharpClaw.Tests.Modules.InProcessModuleJsonColdStoreTests|FullyQualifiedName~SharpClaw.Tests.Tasks.TaskExecutionLifecycleTests.ScheduledJobWorker'
          - domain: Persistence Instances
            filter: 'FullyQualifiedName~SharpClaw.Tests.Persistence.SharpClawInstanceLifecycleTests|FullyQualifiedName~SharpClaw.Tests.Persistence.SharpClawInstancePathsTests'
          - domain: Providers Capabilities
            filter: 'FullyQualifiedName~SharpClaw.Tests.Providers.CapabilityResolverTests'
          - domain: LlamaSharp Schema Conversion
            filter: 'FullyQualifiedName~SharpClaw.Tests.Providers.LlamaSharp.LlamaSharpRegexToGrammarTests|FullyQualifiedName~SharpClaw.Tests.Providers.LlamaSharp.LlamaSharpStringFormatGrammarsTests'
          - domain: LlamaSharp Tool Calling
            filter: 'FullyQualifiedName~SharpClaw.Tests.Providers.LlamaSharp.LlamaSharpToolGrammarShapeTests|FullyQualifiedName~SharpClaw.Tests.Providers.LlamaSharp.LlamaSharpToolPromptBuilderTests'
          - domain: Tasks Scripts Compiler Parser
            filter: 'FullyQualifiedName~SharpClaw.Tests.Tasks.TaskScriptCompilerTests|FullyQualifiedName~SharpClaw.Tests.Tasks.TaskScriptParserTests'
          - domain: Tasks Scripts Semantics Validator
            filter: 'FullyQualifiedName~SharpClaw.Tests.Tasks.TaskScriptSemanticsTests|FullyQualifiedName~SharpClaw.Tests.Tasks.TaskScriptValidatorTests'
          - domain: Tasks Step Keys
            filter: 'FullyQualifiedName~SharpClaw.Tests.Tasks.TaskStepKeyAssignmentTests'
          - domain: Tasks Execution Lifecycle
            filter: 'FullyQualifiedName~SharpClaw.Tests.Tasks.TaskExecutionLifecycleTests'
          - domain: Tasks Shared Data
            filter: 'FullyQualifiedName~SharpClaw.Tests.Tasks.TaskSharedDataStoreTests'
          - domain: Tasks Triggers
            filter: 'FullyQualifiedName~SharpClaw.Tests.Tasks.TaskTriggerAttributeParserTests'
          - domain: Chat Prompt Surface
            filter: 'FullyQualifiedName~AllDynamicChatFeaturesDisabled|FullyQualifiedName~HeaderExpansionSwitches|FullyQualifiedName~ModuleHeaderTagContext'
          - domain: Chat Tool Permissions
            filter: 'FullyQualifiedName~PermissionDeniedInlineTool|FullyQualifiedName~PermissionGrantedInlineTool|FullyQualifiedName~PermissionDeniedJobTool|FullyQualifiedName~PermissionGrantedJobTool'
          - domain: Chat Costs And Budgets
            filter: 'FullyQualifiedName~CachedCostHotPath|FullyQualifiedName~NonStreamingChatMeetsHardProviderPlusOverheadBudget|FullyQualifiedName~StreamingChatAndSseForwardingMeetHardProviderPlusOverheadBudget|FullyQualifiedName~BudgetHelperFailsAtOneMillisecondOverBudget'
          - domain: Chat Thread History
            filter: 'FullyQualifiedName~ThreadedChatSendsNewestHistory|FullyQualifiedName~ThreadedChatDefaultHistoryLimit|FullyQualifiedName~ThreadedChatCharacterLimit'
          - domain: Chat Thread Concurrency
            filter: 'FullyQualifiedName~SameThreadConcurrentSends|FullyQualifiedName~DifferentThreadConcurrentSends|FullyQualifiedName~CancelledThreadedSend'
          - domain: Chat Agent Selection
            filter: 'FullyQualifiedName~RequestedAllowedAgent|FullyQualifiedName~RequestedDisallowedAgent'
          - domain: Chat Prompt Assembly
            filter: 'FullyQualifiedName~TestHarnessPromptAssemblyExpandedTests'
          - domain: Chat Streaming
            filter: 'FullyQualifiedName~TestHarnessStreamingCorrectnessTests'
          - domain: Jobs Direct
            filter: 'FullyQualifiedName~DirectJobSubmissionCompletes|FullyQualifiedName~DirectJobDetailAndLifecycleActionsReuseCachedLogs|FullyQualifiedName~DirectJobDeniedStops|FullyQualifiedName~DirectJobAwaitingApproval|FullyQualifiedName~DirectJobApprovalDenies|FullyQualifiedName~DirectJobFailure'
          - domain: Jobs Lifecycle
            filter: 'FullyQualifiedName~LongRunningDirectJobPauseResumeStopAndCompletedCancelAreStable'
          - domain: Jobs Streaming
            filter: 'FullyQualifiedName~DirectStreamingJobToolConcatenatesChunksIntoResult'
          - domain: Tools Inline Permissions
            filter: 'FullyQualifiedName~AllowedInlineModuleTool|FullyQualifiedName~DeniedInlineModuleTool|FullyQualifiedName~ToolPermissionDoesNotStayCached|FullyQualifiedName~JobToolAliasUsesCanonicalPermissionDescriptor'
          - domain: Tools Arguments Output And Latency
            filter: 'FullyQualifiedName~MalformedToolArgumentsProduceStableError|FullyQualifiedName~LargeToolOutputIsBounded|FullyQualifiedName~ToolLatencyIsMeasuredSeparately'
          - domain: Tools Provider Stream Bulk
            filter: 'FullyQualifiedName~ProviderStream_100ToolCalls_CorrectnessVariants|FullyQualifiedName~ProviderStream_100MixedToolCalls|FullyQualifiedName~ProviderStream_100ToolCalls_MalformedArguments'
          - domain: Tools Provider Stream Cache
            filter: 'FullyQualifiedName~ProviderStream_RepeatedToolCalls_RolePermissionRemovalInvalidatesWarmPath|FullyQualifiedName~ProviderStream_RepeatedToolCalls_RolePermissionGrantInvalidatesWarmPath'
          - domain: Tools Provider Stream Lifecycle
            filter: 'FullyQualifiedName~ProviderStream_ModuleLifecycle|FullyQualifiedName~ProviderStream_TextBeforeToolsAndTextAfterTools|FullyQualifiedName~ProviderStream_CancelAfterToolCalls|FullyQualifiedName~ProviderStream_ToolNameVariants'
          - domain: Costs Job Accounting
            filter: 'FullyQualifiedName~JobCostIsNonNull|FullyQualifiedName~FailedJobCostBehavior|FullyQualifiedName~SharpClaw.Tests.Services.AgentJobTokenAccountingTests.WhenCurrentExecutionRecordsTokens'
          - domain: Costs Streaming
            filter: 'FullyQualifiedName~StreamingCostReportedOnlyAtEnd'
          - domain: Costs Multi Agent
            filter: 'FullyQualifiedName~MultipleAgentsInOneChannelDoNotCrossContaminateCosts'
          - domain: Costs Module Hooks
            filter: 'FullyQualifiedName~CustomModuleCostHookAddsUsage|FullyQualifiedName~SharpClaw.Tests.Services.AgentJobTokenAccountingTests.WhenModuleCostTrackerRecordsTokens'
          - domain: Cache Header Invalidations
            filter: 'FullyQualifiedName~ChannelMutationInvalidates|FullyQualifiedName~AgentMutationInvalidates|FullyQualifiedName~ThreadLimitUpdateInvalidates|FullyQualifiedName~DefaultResourceMutationInvalidates'
          - domain: Cache Tool Settings
            filter: 'FullyQualifiedName~EffectiveToolDefinitionsStayWarmUntilAgentToolSettingsChange'
          - domain: Agent Orchestration Access
            filter: 'FullyQualifiedName~ListAccessibleThreadsScales|FullyQualifiedName~AdminIndependentCanList|FullyQualifiedName~NonAdminDoubleGate|FullyQualifiedName~PrimaryAgentAndAllowedAgents|FullyQualifiedName~DisabledAccessibleThreadsHeader'
          - domain: Agent Orchestration History
            filter: 'FullyQualifiedName~ReadThreadHistoryHandles'
          - domain: Defaults Mutations
            filter: 'FullyQualifiedName~ChannelDefaultsClear|FullyQualifiedName~ContextDefaultsClear|FullyQualifiedName~ClearingCoreAgentDefault|FullyQualifiedName~SetClearSetSameDefaultKey|FullyQualifiedName~UnknownDefaultKeyClear|FullyQualifiedName~ModuleDefinedDefaultResourceKeyCannotShadow'
          - domain: Defaults Resource Resolution
            filter: 'FullyQualifiedName~PerResourceJob'
          - domain: Module API Host Startup
            filter: 'FullyQualifiedName~SharpClaw.Tests.Modules.ApiHostStartupTests|FullyQualifiedName~SharpClaw.Tests.Logging.SessionLogWriterTests'
          - domain: Module Bundled Outputs
            filter: 'FullyQualifiedName~SharpClaw.Tests.Modules.BundledModuleOutputTests'
          - domain: Module External Lifecycle
            filter: 'FullyQualifiedName~SharpClaw.Tests.Modules.SyntheticExternalModuleLifecycleTests'
          - domain: Module Foreign Runtime
            filter: 'FullyQualifiedName~SharpClaw.Tests.Modules.ForeignModule|FullyQualifiedName~SharpClaw.Tests.Modules.JavaScriptModuleSdkTests|FullyQualifiedName~SharpClaw.Tests.Modules.PythonModuleSdkTests'
          - domain: Module Sidecar Parity
            filter: 'FullyQualifiedName~SharpClaw.Tests.Modules.BundledDotNetSidecarDefaultTests|FullyQualifiedName~SharpClaw.Tests.Modules.BundledTestHarnessSidecarConformanceTests|FullyQualifiedName~SharpClaw.Tests.Modules.OutOfProcessModuleHostTests|FullyQualifiedName~SharpClaw.Tests.Modules.SidecarReadinessInventoryTests'
          - domain: Module Harness Integration
            filter: 'FullyQualifiedName~ModuleRegistersDynamicProvidersToolsHeaderTagsAndCostFeed|FullyQualifiedName~ProviderCaptureRedactsSecretsButKeepsAssertablePromptShape'
          - domain: Module Harness Cost And Matrix
            filter: 'FullyQualifiedName~CostFeedUsesHarnessBehaviorAndRecordsTiming|FullyQualifiedName~ProviderMatrixScalesPastOneThousandDeterministicCases'
          - domain: Module Contracts Architecture
            filter: 'FullyQualifiedName~TestHarnessArchitectureTests'
          - domain: Provider Registration
            filter: 'FullyQualifiedName~TestHarnessProviderRegistrationTests'
    steps:
      - name: Checkout
        uses: actions/checkout@v4

      - name: Checkout ModuleSDK test fixture
        if: matrix.domain == 'Module Foreign Runtime'
        uses: actions/checkout@v4
        with:
          repository: SharpClaw-NET/SharpClaw.ModuleSDK
          path: SharpClaw.ModuleSDK
          ref: main

      - name: Set ModuleSDK test fixture root
        if: matrix.domain == 'Module Foreign Runtime'
        shell: pwsh
        run: |
          "SHARPCLAW_MODULESDK_ROOT=$env:GITHUB_WORKSPACE\SharpClaw.ModuleSDK" |
            Out-File -FilePath $env:GITHUB_ENV -Encoding utf8 -Append

      - name: Setup .NET
        uses: actions/setup-dotnet@v4
        with:
          dotnet-version: ${{ env.DOTNET_VERSION }}

      - name: Restore
        shell: pwsh
        run: dotnet restore $env:TEST_PROJECT

      - name: Run correctness domain
        shell: pwsh
        run: >
          dotnet test $env:TEST_PROJECT
          --configuration Release
          --no-restore
          --logger GitHubActions
          --filter "TestCategory!=PerformanceDiagnostic&TestCategory!=PerformanceGate&(${{ matrix.filter }})"

  performance:
    name: Performance / ${{ matrix.domain }}
    runs-on: windows-latest
    timeout-minutes: 30
    strategy:
      fail-fast: false
      matrix:
        include:
          - domain: Streaming End To End
            filter: 'FullyQualifiedName~PerformanceGate_StreamingOverhead_1000ms_AllSurfaces'
          - domain: Streaming API
            filter: 'FullyQualifiedName~PerformanceGate_StreamingOverhead_100ms_ApiStream|FullyQualifiedName~PerformanceGate_StreamingOverhead_Large1000ChunkStream|FullyQualifiedName~PerformanceGate_StreamingOverhead_ApiSse1000TinyChunks'
          - domain: Streaming Gateway
            filter: 'FullyQualifiedName~PerformanceGate_StreamingOverhead_GatewaySseProxy1000TinyChunks'
          - domain: Streaming Concurrency
            filter: 'FullyQualifiedName~PerformanceGate_TenConcurrentStreams'
          - domain: Chat Non Streaming
            filter: 'FullyQualifiedName~PerformanceGate_NonStreamingOverhead'
          - domain: Chat Concurrent Throughput
            filter: 'FullyQualifiedName~PerformanceGate_FiftyConcurrentShortChats'
          - domain: Chat Sequential Warm Cache
            filter: 'FullyQualifiedName~PerformanceGate_OneHundredSequentialWarmCacheChats'
          - domain: Jobs Direct Actions
            filter: 'FullyQualifiedName~PerformanceGate_DirectJobAllowedWarmNoOp|FullyQualifiedName~PerformanceGate_DirectJobDeniedWarmNoOp'
          - domain: Jobs Summaries
            filter: 'FullyQualifiedName~PerformanceGate_DirectJobSummaries|FullyQualifiedName~PerformanceGate_DirectJobDetailHotLogs'
          - domain: Jobs Parallel
            filter: 'FullyQualifiedName~PerformanceGate_DirectJob_TwentyParallelAllowedNoOpJobs'
          - domain: Tools Permission Checks
            filter: 'FullyQualifiedName~PerformanceGate_InlineToolPermissionCheck'
          - domain: Tools Provider Stream Serial
            filter: 'FullyQualifiedName~PerformanceGate_ProviderStream_100AllowedNoOpToolCalls|FullyQualifiedName~PerformanceGate_ProviderStream_100DeniedToolCalls'
          - domain: Tools Provider Stream Parallel
            filter: 'FullyQualifiedName~PerformanceGate_ProviderStream_10ParallelChatsEachWith100AllowedToolCalls|FullyQualifiedName~PerformanceGate_ProviderStream_10ParallelChatsEachWith100DeniedToolCalls'
          - domain: Accessible Threads
            filter: 'FullyQualifiedName~PerformanceGate_AccessibleThreads'
          - domain: Prompt Assembly
            filter: 'FullyQualifiedName~PerformanceGate_PromptAssembly'
          - domain: Cost Lookup
            filter: 'FullyQualifiedName~PerformanceGate_CostLookup'
          - domain: Provider Resolution
            filter: 'FullyQualifiedName~PerformanceGate_ProviderResolution'
          - domain: Cold Chat Cache
            filter: 'FullyQualifiedName~PerformanceGate_ColdChatAfterCacheClear'
    steps:
      - name: Checkout
        uses: actions/checkout@v4

      - name: Setup .NET
        uses: actions/setup-dotnet@v4
        with:
          dotnet-version: ${{ env.DOTNET_VERSION }}

      - name: Restore
        shell: pwsh
        run: dotnet restore $env:TEST_PROJECT

      # Require this check in GitHub branch protection. It is the blocking
      # performance signal; diagnostic tiers remain local opt-in tests.
      - name: Run performance domain
        shell: pwsh
        run: >
          dotnet test $env:TEST_PROJECT
          --configuration Release
          --no-restore
          --logger GitHubActions
          --filter "TestCategory=PerformanceGate&(${{ matrix.filter }})"