163 apply v1 100 instance

.azure/modules/appi-applicationinsights.bicep

@@ -20,7 +20,7 @@ param name string
 
 @description('The resource ID of the Log Analytics workspace to link to Application Insights. Must be a valid resourceId string.')
 @minLength(1)
-param workResourceId string
+param logResourceId string
 
 @description('Controls whether Application Insights accepts public telemetry ingestion.')
 @allowed([
@@ -44,7 +44,7 @@ resource appiResource 'Microsoft.Insights/components@2020-02-02' = {
   properties: {
     Application_Type: 'web'
     Flow_Type: 'Bluefield'
-    WorkspaceResourceId: workResourceId
+    WorkspaceResourceId: logResourceId
     publicNetworkAccessForIngestion: publicNetworkAccessForIngestion
     publicNetworkAccessForQuery: publicNetworkAccessForQuery
   }

.azure/modules/aca-ollama.bicep → .azure/modules/ca-ollama.bicep

@@ -22,7 +22,7 @@ param stAccountName string
 @minLength(3)
 @maxLength(63)
 @description('Name of the Azure Files share mounted into Ollama for persistent model data.')
-param storageShareName string = 'ollama-models'
+param storageShareName string = 'share-ollamamodels'
 
 @description('Ollama model to pull and serve on startup.')
 @allowed([

.azure/modules/law-loganalyticsworkspace.bicep → .azure/modules/log-loganalyticsworkspace.bicep

@@ -22,7 +22,7 @@ param tags object = {}
 @minValue(30)
 param retentionInDays int = 30
 
-resource workResource 'Microsoft.OperationalInsights/workspaces@2023-09-01' = {
+resource logResource 'Microsoft.OperationalInsights/workspaces@2023-09-01' = {
   name: name
   location: location
   tags: empty(tags) ? null : tags
@@ -34,4 +34,4 @@ resource workResource 'Microsoft.OperationalInsights/workspaces@2023-09-01' = {
   }
 }
 
-output id string  = workResource.id
+output id string  = logResource.id

.azure/modules/sent-loganalyticsworkspace.bicep → .azure/modules/log-sentinelworkspace.bicep

@@ -22,9 +22,9 @@ param tags object = {}
 @minValue(30)
 param retentionInDays int = 30
 
-// Sentinel = the plain workspace (work-) plus onboarding
-module workspace 'law-loganalyticsworkspace.bicep' = {
-  name: 'sent-${name}'
+// Sentinel = the plain Log Analytics workspace (log-loganalyticsworkspace) plus onboarding
+module workspace 'log-loganalyticsworkspace.bicep' = {
+  name: 'log-sentinel-${name}'
   params: {
     name: name
     location: location
@@ -34,7 +34,7 @@ module workspace 'law-loganalyticsworkspace.bicep' = {
   }
 }
 
-resource workResource 'Microsoft.OperationalInsights/workspaces@2023-09-01' existing = {
+resource logResource 'Microsoft.OperationalInsights/workspaces@2023-09-01' existing = {
   name: name
   dependsOn: [
     workspace
@@ -43,7 +43,7 @@ resource workResource 'Microsoft.OperationalInsights/workspaces@2023-09-01' exis
 
 resource onboarding 'Microsoft.SecurityInsights/onboardingStates@2021-03-01-preview' = {
   name: 'default'
-  scope: workResource
+  scope: logResource
   properties: {}
 }
 

.azure/templates/workload-digitalproduct.bicep

@@ -59,10 +59,10 @@ var sharedSubscriptionId = shared.?subscriptionId ?? subscription().subscription
 var sharedResourceGroupName = shared.?resourceGroupName ?? resourceGroup().name
 var vnetSubscriptionId = vnet.?subscriptionId ?? subscription().subscriptionId
 
-module lawModule '../modules/law-loganalyticsworkspace.bicep' = if (isStandalone) {
-  name: 'digitalproduct-law'
+module logModule '../modules/log-loganalyticsworkspace.bicep' = if (isStandalone) {
+  name: 'digitalproduct-log'
   params: {
-    name: hosting.?workspaceName ?? '${hosting.appiName}-work'
+    name: hosting.workspaceName!
     location: location
     tags: tags
   }
@@ -74,12 +74,12 @@ module appiModule '../modules/appi-applicationinsights.bicep' = if (isStandalone
     name: hosting.appiName
     location: location
     tags: tags
-    workResourceId: lawModule!.outputs.id
+    logResourceId: logModule!.outputs.id
   }
 }
 
-module planModule '../modules/plan-appserviceplan.bicep' = if (isStandalone) {
-  name: 'digitalproduct-plan'
+module planModule '../modules/asp-appserviceplan.bicep' = if (isStandalone) {
+  name: 'digitalproduct-asp'
   params: {
     name: hosting.planName
     location: location

.azure/types/digitalproduct-types.bicep

@@ -20,7 +20,7 @@ type hostingConfig = {
 
   @minLength(4)
   @maxLength(63)
-  @description('Standalone only. Log Analytics workspace name backing Application Insights.')
+  @description('Required in standalone mode. Log Analytics workspace name backing Application Insights.')
   workspaceName: string?
 }
 

.azure/variables/workload-digitalproduct-standalone-dev.bicepparam

@@ -2,12 +2,12 @@ using '../templates/workload-digitalproduct.bicep'
 
 // Phase 1 (mode 1, standalone): plan, Log Analytics, App Insights, web, API and SQL all live in this RG.
 // RG: {product}-{env}-{region}-{instance}-rg. Secrets: set SQL_ADMIN_USER and SQL_ADMIN_PASSWORD in the environment before running az deployment group.
-// Deploy: az deployment group create -g agent-dev-wus2-001-rg -f .azure/templates/workload-digitalproduct.bicep -p .azure/variables/workload-digitalproduct-standalone-dev.bicepparam
+// Deploy: az deployment group create -g agent-dev-wus2-100-rg -f .azure/templates/workload-digitalproduct.bicep -p .azure/variables/workload-digitalproduct-standalone-dev.bicepparam
 
 var productIac = 'agent'
 var environmentIac = 'dev'
 var regionIac = 'wus2'
-var instanceIac = '001'
+var instanceIac = '100'
 var prefix = '${productIac}-${environmentIac}-${regionIac}-${instanceIac}'
 
 param mode = 'standalone'
@@ -17,9 +17,9 @@ param tags = { Environment: environmentIac, CostCenter: '0000' }
 
 param hosting = {
   appiName: '${prefix}-appi'
-  planName: '${prefix}-F1-plan'
+  planName: '${prefix}-F1-asp'
   planSku: 'F1'
-  workspaceName: '${prefix}-work'
+  workspaceName: '${prefix}-log'
 }
 
 // Components: delete a line to skip that component.

.github/workflows/agent-app.yml

@@ -42,7 +42,7 @@ env:
   # ===== EDIT PER INSTALL: naming slots (names: {product}-{env}-{region}-{instance}-{type}) =====
   PRODUCT: "agent"
   REGION: "wus2"
-  INSTANCE: "001"
+  INSTANCE: "100"
   # Same value as MODE in the IaC workflow. standalone: App Insights is in the product RG; any other mode: in the shared management RG.
   MODE: "standalone"
   # Shared management RG and App Insights from platform-shared-mgmt; product is normally 'mgmt'.

.github/workflows/agent-iac.yml

@@ -39,7 +39,7 @@ env:
   # ===== EDIT PER INSTALL: naming slots (RG: {product}-{env}-{region}-{instance}-rg) =====
   PRODUCT: "agent"
   REGION: "wus2"
-  INSTANCE: "001"
+  INSTANCE: "100"
   LOCATION: "westus2"
   # standalone | shared | network | zerotrust
   MODE: "standalone"

Initialize-Local.ps1

@@ -9,7 +9,7 @@ Run from any folder in PowerShell (not 'pwsh -File', so a splatted hashtable wor
 3. Database: applies the EF migrations to the local SQL Server (LocalDB by default).
 Local only: it does not touch GitHub, Azure Pipelines or Azure resources. CI/CD secrets and federation are set up separately.
 Safe to re-run: existing registrations are reused, and the Web client secret is only regenerated with -RotateSecrets.
-Leave -ApplicationInsightsConnectionString empty until the infrastructure is deployed; the apps start without it, with no Azure telemetry. Deployment names are only stored when you pass them, because the appsettings files already carry the template defaults (chat-fast, chat-balanced, embedding-fast).
+Leave -ApplicationInsightsConnectionString empty until the infrastructure is deployed; the apps start without it, with no Azure telemetry. Deployment names are only stored when you pass them, because the appsettings files already carry the template defaults (chat-fast, tool-standard, tool-reasoning, embedding-standard).
 
 .EXAMPLE
 # Omit -AzureOpenAIApiKey to be prompted without echo:
@@ -31,6 +31,8 @@ param(
     [Parameter(Mandatory)][string]$AzureOpenAIEndpoint,
     [string]$AzureOpenAIApiKey,
     [string]$ChatDeploymentName,
+    [string]$ToolDeploymentName,
+    [string]$ToolReasoningDeploymentName,
     [string]$EmbeddingDeploymentName,
     [string]$ApplicationInsightsConnectionString = "",
 
@@ -108,6 +110,8 @@ $openAi = @{
     "AzureOpenAI:ApiKey"   = $AzureOpenAIApiKey
 }
 if ($ChatDeploymentName) { $openAi["AzureOpenAI:ChatDeploymentName"] = $ChatDeploymentName }
+if ($ToolDeploymentName) { $openAi["AzureOpenAI:ToolDeploymentName"] = $ToolDeploymentName }
+if ($ToolReasoningDeploymentName) { $openAi["AzureOpenAI:ToolReasoningDeploymentName"] = $ToolReasoningDeploymentName }
 if ($EmbeddingDeploymentName) { $openAi["AzureOpenAI:EmbeddingDeploymentName"] = $EmbeddingDeploymentName }
 $appInsights = @{ "ApplicationInsights:ConnectionString" = $ApplicationInsightsConnectionString }
 

README.md

@@ -80,7 +80,7 @@ dotnet user-secrets set "ApplicationInsights:ConnectionString" "<connection-stri
 ```
 In Azure, the deployment sets the connection string for you.
 
-**Azure OpenAI deployment names:** the committed appsettings use `chat-fast` (Development and local), `chat-balanced` (Production) and `embedding-fast`. These are placeholders for the names of your own Azure OpenAI or AI Foundry model deployments. Replace them in `src/Presentation.Api/appsettings.*.json` to match your deployments (deployment names are not secrets, so they need no GitHub secrets), or pass `-ChatDeploymentName` and `-EmbeddingDeploymentName` to `Initialize-Local.ps1` to override them in your local user-secrets only.
+**Azure OpenAI deployment names:** all committed environments use `chat-fast`, `tool-standard`, `tool-reasoning`, and `embedding-standard`, matching the shared `platform-azure` Foundry deployments. To use different purpose-compatible deployments, update `src/Presentation.Api/appsettings.*.json`, or pass `-ChatDeploymentName`, `-ToolDeploymentName`, `-ToolReasoningDeploymentName`, and `-EmbeddingDeploymentName` to `Initialize-Local.ps1` for local user-secret overrides only. Deployment names are not secrets.
 
 ## Quick-Start: Re-Setup for Existing Entra (no Entra setup scripts)
 ```powershell
@@ -283,7 +283,7 @@ dotnet user-secrets set "AzureOpenAI:ApiKey" "<azure-openai-api-key>" --project
 dotnet user-secrets set "AzureOpenAI:Endpoint" "https://<resource>.openai.azure.com/openai/v1" --project .\src\Tests.Integration\Tests.Integration.csproj
 dotnet user-secrets set "AzureOpenAI:ApiKey" "<azure-openai-api-key>" --project .\src\Tests.Integration\Tests.Integration.csproj
 ```
-The deployment names come from `AzureOpenAI:ChatDeploymentName` and `AzureOpenAI:EmbeddingDeploymentName` in `src/Presentation.Api/appsettings.*.json` (`chat-fast`, `chat-balanced`, `embedding-fast`). Replace them with your own deployment names, or override them with `dotnet user-secrets set` using the same keys.
+Deployment names come from `AzureOpenAI:ChatDeploymentName`, `AzureOpenAI:ToolDeploymentName`, `AzureOpenAI:ToolReasoningDeploymentName`, and `AzureOpenAI:EmbeddingDeploymentName` in `src/Presentation.Api/appsettings.*.json`. Override these same keys with `dotnet user-secrets set` when using different purpose-compatible deployments.
 
 Alternately you can set in Environment variables
 ```
@@ -441,7 +441,7 @@ doc's **UI Changes** and **API Changes** sections for details.
 # Github Actions for Azure IaC and CI/CD
 ## GitHub Actions (.github folder)
 
-The `.github/workflows` folder contains the GitHub Actions pipelines for CI/CD. Workflows follow the [github-actions](https://github.com/goodtocode/github-actions) conventions and are named `{product}-{capability}.yml`. The naming slots (`PRODUCT`, `REGION`, `INSTANCE`, `MODE`) at the top of each workflow produce the Azure names `{product}-{env}-{region}-{instance}-{type}`, for example `agent-dev-wus2-001-rg`. Below is a summary of the workflow files, their purposes, and triggers:
+The `.github/workflows` folder contains the GitHub Actions pipelines for CI/CD. Workflows follow the [github-actions](https://github.com/goodtocode/github-actions) conventions and are named `{product}-{capability}.yml`. The naming slots (`PRODUCT`, `REGION`, `INSTANCE`, `MODE`) at the top of each workflow produce the Azure names `{product}-{env}-{region}-{instance}-{type}`, for example `agent-dev-wus2-100-rg`. Below is a summary of the workflow files, their purposes, and triggers:
 
 ### Triggers
 All workflow YAML files in this repo are designed to:
@@ -567,4 +567,11 @@ This setup ensures your GitHub Actions workflows can securely deploy to Azure us
 | 1.0.0   | 2026-Feb-02 | Initial Release                                                  |
 | 1.1.0   | 2026-Jul-27 | AI-ize md/yml, AgentProvider                                     |
 
-This project is licensed with the [MIT license](https://mit-license.org/).
+This project is licensed with the [MIT license](https://mit-license.org/).
+
+### Shared Foundry deployment configuration
+
+Use `AzureOpenAI:ChatDeploymentName=chat-fast` for text-only Phi chat, `AzureOpenAI:ToolDeploymentName=tool-standard` for Azure-direct Mistral-Large-3 function calls, and `AzureOpenAI:ToolReasoningDeploymentName=tool-reasoning` for MAI playbook inference. Azure-direct Cohere embed-v-4-0 intent embeddings use `AzureOpenAI:EmbeddingDeploymentName=embedding-standard` and `EmbeddingDimension=1536`.
+These names match the `platform-azure` shared Foundry definitions; provision them and verify regional model access/quota before running live AI scenarios. MAI-Thinking-1 remains preview. Existing embedding vectors must be regenerated after switching models.
+All three purpose deployments are required for the selected provider. Old user-secret/environment overrides may need migration; a chat-only override does not select a tool deployment.
+See [semantic intent classification](docs/product/features/feature-semantic-intent-classification.md#purpose-specific-foundry-inference) for routing, governance, and migration details.

docs/product/features/feature-semantic-intent-classification.md

@@ -97,7 +97,7 @@ The API configuration exposes the following section in every environment file:
 
 The section is bound through `IOptions<IntentClassificationOptions>`. `EnableSemantic` must remain `false` until evaluation and rollout approval are complete.
 
-Azure embedding settings are bound through `AzureOpenAIOptions`, including `EmbeddingDeploymentName` with the default deployment name `embedding-fast`.
+Azure embedding settings are bound through `AzureOpenAIOptions`, including `EmbeddingDeploymentName` with the default deployment name `embedding-standard`.
 
 ## Shared Implementation
 
@@ -190,7 +190,7 @@ Both repositories require:
 - Reqnroll scenarios covering the disabled switch and an enabled semantic match using deterministic embeddings and in-memory storage.
 - API configuration parsing tests or equivalent validation confirming the switch is present and disabled by default.
 
-Live Azure-backed seeding is a separate environment validation. It requires a configured `embedding-fast` deployment, SQL database, credentials, and an applied migration.
+Live Azure-backed seeding is a separate environment validation. It requires a configured `embedding-standard` deployment, SQL database, credentials, and an applied migration.
 
 ## Rollout
 
@@ -221,3 +221,34 @@ Live Azure-backed seeding is a separate environment validation. It requires a co
 - [ ] Production enablement is approved and reversible.
 - [ ] Governance documentation contains durable principles only.
 - [ ] Tactical implementation details remain in this product feature document.
+
+
+## Purpose-specific Foundry inference
+
+These applications consume the shared `platform-azure` Foundry deployments:
+
+| Purpose | Configuration | Deployment | Model |
+| --- | --- | --- | --- |
+| Text-only chat | `AzureOpenAI:ChatDeploymentName` | `chat-fast` | Phi-4-mini-instruct |
+| Interactive functions | `AzureOpenAI:ToolDeploymentName` | `tool-standard` | Mistral-Large-3 (Azure-direct) |
+| Playbook reasoning | `AzureOpenAI:ToolReasoningDeploymentName` | `tool-reasoning` | MAI-Thinking-1 |
+| Intent embeddings | `AzureOpenAI:EmbeddingDeploymentName` | `embedding-standard` | Cohere embed-v-4-0 (Azure-direct, 1536 dimensions) |
+
+`embedding-standard` is shared by intent classification and retrieval consumers, using Cohere embed-v-4-0 at 1536 dimensions.
+Callable function names must fit the OpenAI-compatible API's 64-character limit. Names within the limit
+remain unchanged; longer tool-qualified names use a readable 47-character prefix and a deterministic
+16-hex-character SHA-256 suffix to avoid truncation collisions and provider HTTP 422 errors.
+MAI-Thinking-1 remains a preview model despite the deployment name having no preview suffix.
+Verify model availability, access, and quota before provisioning the shared deployments; the names alone do not establish capabilities.
+
+The keyed `InferencePurpose.Chat` agent advertises no application functions. Direct mode and reply-wording correction use only this agent. Correction receives the original reply, retains the governed policy context, and cannot repeat application operations.
+The default agent is `InferencePurpose.Tool`: described public tool methods become real `AIFunction` instances with typed schemas and qualified names.
+After classification misses, general requests use one automatic-function turn. Recognized chat-history requests require a function call; missing calls, provider failures, and cancellation propagate without a second open-agent attempt.
+Agentic Essay and Taxonomy stages explicitly consume `InferencePurpose.ToolReasoning`; these stages do not expose chat administration functions.
+Existing durable chat governance and hosted catalog/embedding initialization remain enabled.
+
+AzureAIFoundry uses the corresponding `ChatCompletionModelId`, `ToolCompletionModelId`, and `ToolReasoningModelId` deployment settings. OpenAI, GitHubCopilotSDK, and Ollama also require explicit purpose models; unused providers remain unvalidated. Ollama defaults retain text-only phi4 and use OpenAI gpt-oss:20b for function/reasoning purposes, requiring a local model pull before use.
+
+Embedding calls use the official OpenAI SDK and the normalized `/openai/v1/embeddings` endpoint, passing the deployment name and configured dimension. Root endpoints and existing `/openai/v1/` endpoints are both supported.
+Before enabling semantic classification, regenerate persisted intent embeddings when changing the embedding model, even if the dimension is unchanged. Reconcile SQL vector storage when changing dimensions; do not compare embeddings from different model spaces.
+No timeout increases, automatic model fallback, or Azure deployment are part of this change.

src/Core.Application/Abstractions/ChatRoutingMode.cs

@@ -5,9 +5,9 @@ namespace Goodtocode.AgentFramework.Core.Application.Abstractions;
 /// </summary>
 public enum ChatRoutingMode
 {
-    /// <summary>Uses deterministic routing, forced-tool inference, then an open agent turn.</summary>
+    /// <summary>Uses deterministic routing, then purpose-specific function inference without open-agent retries.</summary>
     Routed = 0,
 
-    /// <summary>Bypasses deterministic and forced-tool routing for diagnostics or explicit direct-agent requests.</summary>
+    /// <summary>Uses text-only inference without classification or application functions.</summary>
     Direct = 1
 }