Orchardcore AI Chat Claude

Orchard Core AI Chat Claude - Prompt Templates

Configure the Claude Chat Orchestrator

You are an Orchard Core expert. Generate configuration, recipes, and extension code for Claude as an Orchard-managed chat orchestrator using CrestApps modules. Claude is not an AI provider connection or deployment provider in this integration.

Guidelines

• Enable CrestApps.OrchardCore.AI.Chat.Claude to make the Claude orchestrator available to AI Profiles, profile-source AI Profile Templates, and Chat Interactions.
• The Claude feature depends on CrestApps.OrchardCore.AI; enable the feature rather than treating it as an OpenAI-compatible connection.
• ClaudeOptionsConfiguration binds shell configuration from CrestApps:AI:Claude, then overlays tenant ClaudeSettings.
• Tenant BaseUrl and DefaultModel override corresponding shell values only when populated.
• Use Settings → Artificial Intelligence → Claude to set tenant authentication, API key, base URL, and default model.
• API-key authentication stores ProtectedApiKey encrypted with ASP.NET Core Data Protection. Never put a production Anthropic key in a recipe, source file, or committed configuration.
• Claude is considered configured only when its authentication type is ApiKey and an encrypted API key exists.
• ClaudeSettingsDisplayDriver exposes the tenant settings editor and ClaudeModelSelectListFactory obtains models from the configured Claude endpoint.
• Model selectors remain unavailable until the tenant has valid API-key configuration.
• Select a Claude model and reasoning effort per AI Profile, profile-source template, or Chat Interaction as needed.
• Valid effort choices are Default, Low, Medium, and High.
• Applying a profile-source template copies saved Claude session settings to the created profile.
• Grant ManageClaudeSettings only to roles that may change tenant Claude credentials and defaults.
• Install CrestApps packages in the web/startup project.

Feature Overview

Install and Enable

Install the orchestrator package in the web/startup project:

dotnet add package CrestApps.OrchardCore.AI.Chat.Claude

Enable the core and Claude features. Add a chat UI feature only when the application needs it:

{
  "steps": [
    {
      "name": "Feature",
      "enable": [
        "CrestApps.OrchardCore.AI",
        "CrestApps.OrchardCore.AI.Chat.Claude",
        "CrestApps.OrchardCore.AI.Chat"
      ],
      "disable": []
    }
  ]
}

Configure Shell Defaults

Provide non-secret defaults through the tenant shell configuration. Use user secrets, environment variables, or a secret store for the API key.

{
  "CrestApps": {
    "AI": {
      "Claude": {
        "BaseUrl": "https://api.anthropic.com",
        "DefaultModel": "claude-sonnet-4-5"
      }
    }
  }
}

The configuration key is CrestApps:AI:Claude. This module binds ClaudeOptions from that section before it reads tenant site settings.

Configure Tenant Settings

1. Enable AI Claude Orchestrator.
2. Go to Settings → Artificial Intelligence → Claude.
3. Set authentication to API Key.
4. Enter the Anthropic API key, base URL, and default model.
5. Save the settings. The editor protects the API key before it is persisted.
6. Reopen the profile or interaction editor and select a discovered Claude model.

ClaudeOptionsConfiguration uses a data-protection purpose specific to Claude settings to unprotect the saved key for runtime use. A decryption failure logs a warning and leaves the key unavailable instead of exposing the secret.

Select Claude for an AI Profile

1. Create or edit an AI Profile.
2. Select the Claude orchestrator.
3. Choose a Claude model and effort level.
4. Configure the profile system message and capabilities normally.
5. Save and test the profile with a user authorized to query it.

Use Claude session settings for per-profile behavior. Do not create an AIProviderConnection or an AIDeployment merely to select the Claude orchestration runtime.

Reuse Claude Settings with Templates

Use a profile-source AI Profile Template when multiple profiles should start with the same Claude model and reasoning effort:

1. Create a template whose source is Profile.
2. Select the Claude orchestrator.
3. Choose the model and effort level.
4. Create profiles from that template.

AIProfileTemplateClaudeDisplayDriver stores the template settings, and the template-to-profile flow propagates them to the generated profile.

Configure Chat Interactions

Enable CrestApps.OrchardCore.AI.Chat.Interactions when administrators need ad-hoc interactions rather than predefined profiles:

{
  "steps": [
    {
      "name": "Feature",
      "enable": [
        "CrestApps.OrchardCore.AI",
        "CrestApps.OrchardCore.AI.Chat.Claude",
        "CrestApps.OrchardCore.AI.Chat.Interactions"
      ],
      "disable": []
    }
  ]
}

In the interaction editor, select Claude, then select an available model and effort level. ChatInteractionClaudeDisplayDriver persists those orchestrator-specific values without changing the generic interaction infrastructure.

Permission Management

The module registers ClaudePermissionProvider and the ManageClaudeSettings permission. Administrators receive it by default.

Assign the permission through a role only to trusted operators:

1. Go to Security → Roles.
2. Edit the operations role.
3. Grant Manage Claude Settings.
4. Do not grant the permission to ordinary profile authors unless they must manage tenant credentials.

Key Implementation Types

Security Checklist

• Keep API keys outside recipes and checked-in configuration.
• Use dotnet user-secrets locally and environment variables, Key Vault, or an equivalent secret provider in production.
• Give ManageClaudeSettings only to trusted tenant administrators.
• Use an intentional default model and per-item model overrides to manage availability and cost.
• Limit access to the selected AI Profiles separately through Orchard Core AI profile permissions.