Adding AI Models

Available

CutReady uses large language models (LLMs) to power its AI assistant — planning sketches, writing narratives, improving descriptions, and managing your demo project. This guide covers how to connect CutReady to a model provider.

Supported Providers

Provider	Endpoint Format	Auth Methods
Microsoft AI Foundry	https://your-hub.services.ai.azure.com/api/projects/your-project	API Key, Azure OAuth
Azure OpenAI	https://your-resource.openai.azure.com	API Key, Azure OAuth
OpenAI	https://api.openai.com (default)	API Key
Anthropic	https://api.anthropic.com (default)	API Key

Microsoft AI Foundry (Recommended)

AI Foundry provides a unified endpoint for accessing multiple model deployments. CutReady auto-detects Foundry endpoints (.services.ai.azure.com) and adjusts its API calls accordingly.

Prerequisites

• An Azure subscription
• An AI Foundry hub and project (create one here)
• At least one model deployed in your project (e.g., gpt-4o)

Setup Steps

1. Get your project endpoint

   In the AI Foundry portal, navigate to your project. Copy the endpoint URL — it looks like:

   https://your-hub.services.ai.azure.com/api/projects/your-project

2. Open CutReady Settings

   Click the gear icon (⚙️) pinned to the bottom of the sidebar, then select the AI Provider tab. (You can also configure AI Provider per workspace in Workspace Settings → AI Provider Override.)

3. Select Microsoft AI Foundry as the provider

   Set Provider to Microsoft AI Foundry and CutReady will detect the Foundry endpoint format automatically.

4. Paste your endpoint

   Enter the full Foundry project endpoint URL in the Endpoint field.

5. Choose authentication

   API Key (Simplest)
   • Set Authentication to API Key
   • Paste your Foundry project API key (found in the AI Foundry portal under your project’s Keys and Endpoint section)
   Azure OAuth (Recommended)
   • Set Authentication to Sign in with Azure (Entra ID)
   • Enter your Tenant ID and Client ID
   • Click Sign in with Azure — a browser window opens for the OAuth flow
   • Complete the login in your browser and return to CutReady
   • CutReady will obtain and refresh tokens automatically

6. Select a model

   Click Refresh next to the model dropdown. CutReady queries your Foundry project for available deployments and shows capability tags when the provider returns them, such as context window, vision support, and Responses API support. Select the model you want to use (e.g., gpt-4o).

Caution

Make sure your deployed model supports chat completions and function calling (tool use). Models like gpt-4o and gpt-4o-mini work well. Embedding-only or image-only models will not work with CutReady’s AI assistant.

How Foundry Detection Works

CutReady automatically detects Foundry endpoints by checking if the URL contains .services.ai.azure.com. When detected, it:

• Uses the Foundry-compatible chat completions path
• Queries the Foundry deployments API for available models and capability metadata
• Filters to chat-capable models when populating the model dropdown
• Stores context length and vision support for the selected model when available

No special configuration is needed — just paste the Foundry endpoint and CutReady handles the rest.

Azure OpenAI (Standard)

For standard Azure OpenAI resources (not Foundry):

1. Get your resource endpoint from the Azure portal (e.g., https://your-resource.openai.azure.com)
2. Set Provider to Azure OpenAI
3. Paste the endpoint URL
4. Choose API Key or Azure OAuth for authentication
5. Click Refresh to load deployed models and select one

OpenAI (Direct)

For direct OpenAI API access:

1. Set Provider to OpenAI
2. Leave the endpoint blank (defaults to https://api.openai.com)
3. Paste your OpenAI API key
4. Click Refresh to load available models

Anthropic

For Anthropic Claude models:

1. Set Provider to Anthropic
2. Leave the endpoint blank (defaults to https://api.anthropic.com)
3. Paste your Anthropic API key
4. Click Refresh to load available models

Recommended Models

Model	Best For	Available Via
gpt-4o	Best quality	Foundry, Azure OpenAI, OpenAI
gpt-4o-mini	Faster responses	Foundry, Azure OpenAI, OpenAI
claude-3-5-sonnet-latest	High quality, long context	Anthropic

Note

CutReady requires models that support function calling (tool use). This is how the AI assistant interacts with your project — reading sketches, updating notes, modifying planning rows, etc.

Model Capability Tags

After you fetch models, each result may show tags like:

Tag	Meaning
128k ctx	Reported context window for conversation and compaction planning
vision	The model can receive image inputs from notes or sketch screenshots
responses	CutReady can route the model through the Responses API

Capability metadata is provider-dependent. If a provider does not report a field, CutReady leaves it unknown rather than guessing.

Internet Search

The Internet Search setting controls whether agents receive the search_web tool. Leave it disabled when the assistant should only use your project files, notes, screenshots, and explicitly provided URLs. Enable it when you want agents to look up current public information on request.

Project content is not included in search queries unless you explicitly ask the assistant to search for that content.

Encrypted Credential Storage

All API keys, OAuth tokens, and personal access tokens are stored in an IOTA Stronghold encrypted vault (vault.hold in the app data directory). Credentials are never written to disk in plaintext.

If you’re upgrading from an earlier version, CutReady automatically migrates any plaintext credentials from settings.json to the encrypted vault on first launch. No action is needed.

Troubleshooting

”No models found”

• Verify your endpoint URL is correct
• Check that you have at least one model deployed
• Ensure your API key or OAuth token has permission to list deployments

”Authentication failed”

• For API Key: verify the key is correct and not expired
• For OAuth: check that your Tenant ID and Client ID are correct, and that the app registration has the required permissions

Model not responding

• Check the debug panel (click the bug icon in the title bar) for error details
• Ensure the selected model supports chat completions
• Verify your Azure subscription/OpenAI account has available quota