GhostBro supports custom AI providers, allowing you to connect AI models from platforms such as NVIDIA Build and use them directly inside your workflow.

One of the biggest advantages of NVIDIA Build is that many models can be used for free, making it possible to get started with GhostBro.app without paying for a commercial AI provider.

In this guide, we’ll connect the GLM 5.1 model from NVIDIA Build and configure it inside GhostBro.

By the end of this tutorial, you’ll be able to use free NVIDIA-hosted AI models for meeting transcript analysis, summaries, smart suggestions, and other AI-powered workflows supported by the selected model.

Step 1: Create a Free NVIDIA Build Account

Create an account at:

https://build.nvidia.com

Registration is free and gives you access to NVIDIA’s growing catalog of AI models.

NVIDIA Build homepage

Step 2: Browse Available Models

Open the models catalog:

https://build.nvidia.com/models

You’ll find a large collection of models from different providers, including reasoning models, coding models, agentic AI models, and multimodal models.

For this guide, we’ll use GLM 5.1.

Free Models Available on NVIDIA Build

Many users don’t realize that NVIDIA Build offers several models that can be used for free.

At the time of writing, models such as GLM 5.1 and Kimi are available through NVIDIA Build’s free access program.

This makes NVIDIA Build a great option for GhostBro users who want to experiment with AI models before committing to a paid provider.

What Does “Free” Mean?

Free models allow you to:

  • Use GhostBro without purchasing credits from another AI provider.

  • Compare different AI models.

  • Test workflows and prompts.

  • Explore advanced AI capabilities at no cost.

Important: Free Models Have Limits

Free access comes with usage limits.

Depending on the model, NVIDIA may enforce:

  • Requests per minute

  • Requests per day

  • Token limits

  • Concurrent request limits

  • Temporary restrictions during peak demand

If you exceed these limits, you may receive quota or rate-limit errors until your allowance resets.

Before selecting a model, review its documentation and limits on NVIDIA Build.

Step 3: Open the Model Page

Navigate to:

https://build.nvidia.com/z-ai/glm-5.1

Open the Build section and select the Code tab.

This page contains everything needed to connect the model to GhostBro.

Step 4: Copy the Base URL

Inside the Build → Code section, locate the API endpoint.

At the time of writing:

https://integrate.api.nvidia.com/v1

Copy this value.

In case that the URL looks like this

https://integrate.api.nvidia.com/v1/chat/completions

ignore everything after the v1

You’ll use it when creating the custom provider in GhostBro.

Step 5: Generate an API Key

You can generate an API key directly from the model page or from:

https://build.nvidia.com/settings/api-keys

Create a new API key and copy it.

Keep this key private, as it grants access to your NVIDIA Build account.

Step 6: Open GhostBro Settings

Launch GhostBro and navigate to:

Settings → API Keys

Scroll down to the Custom Providers section and click:

Add Custom Provider

Step 7: Create the NVIDIA Provider

Fill the form using the values from NVIDIA Build.

Provider Name

nvidia

Base URL

https://integrate.api.nvidia.com/v1

API Key

Paste the API key generated earlier.

Click:

Add Provider

Step 8: Understanding Model Capabilities

Before adding any model to GhostBro, it’s important to understand what the model can and cannot do.

Not all AI models support screenshots, images, or screen analysis.

Example: GLM 5.1

GLM 5.1 is primarily designed for:

  • Agentic AI workflows

  • Reasoning

  • Text generation

  • Meeting transcript analysis

  • Summarization

  • Question answering

  • Conversation analysis

GLM 5.1 works very well when GhostBro analyzes text generated from meeting transcripts.

However, GLM 5.1 does not support image understanding.

This means:

  • Meeting transcription analysis → Supported ✅

  • Smart Suggestions → Supported ✅

  • Summaries → Supported ✅

  • Screen Analysis → Not Supported ❌

If you select GLM 5.1 for Screen Analysis, NVIDIA will reject the request and GhostBro will display an error such as:

glm5.1 is not a multi modal model

This is expected because the model does not support image inputs.

Step 9: Choosing Models for Screen Analysis

If you want to use GhostBro’s Screen Analysis features, pay close attention to the model capabilities shown on NVIDIA Build.

Look for models that include tags such as:

  • Image-to-Text

  • Vision

  • Multimodal

  • Image Understanding

These models can analyze screenshots and visual content.

Models that only support text generation should be used for transcript analysis and text-based workflows.

Always review the model specifications before adding it to GhostBro.

Step 10: Add the Model to AI Analysis

Open:

AI Analysis

Add a new model and select the provider you created.

Example:

Model name

glm5.1

Model identifier

z-ai/glm-5.1

Provider

Nvidia

Save the configuration.

Step 11: Test the Connection

Run an AI Analysis task using the newly configured model.

If everything is configured correctly, GhostBro will send requests directly to NVIDIA’s API using your custom provider settings.

Let’s check an example with kimi 2.6 working with transcription.

and here an example with screen analysis

Troubleshooting

“Request cannot be processed”

This usually means the selected model does not support the type of content being sent.

For example:

  • GLM 5.1 + transcript analysis → Supported ✅

  • GLM 5.1 + screen analysis → Not Supported ❌

Review the model’s capabilities on NVIDIA Build and choose a model that matches your use case.

Authentication Error

Verify that:

  • The API key is valid.

  • The API key was copied correctly.

  • The Base URL matches the value shown in NVIDIA Build.

Model Not Found

Verify that the model name entered in GhostBro matches the identifier shown on NVIDIA Build.

Final Thoughts

NVIDIA Build provides an excellent collection of free AI models that can be integrated into GhostBro in just a few minutes.

The setup process is simple:

  1. Create a free NVIDIA Build account.

  2. Choose a model.

  3. Copy the Base URL from Build → Code.

  4. Generate an API key.

  5. Add a Custom Provider in GhostBro.

  6. Configure the model in AI Analysis.

  7. Verify that the model supports your intended use case.

For transcript analysis, meeting summaries, and agentic AI workflows, models such as GLM 5.1 and Kimi are excellent free options.

If you plan to use Screen Analysis, make sure to choose a model that explicitly supports image understanding capabilities such as Image-to-Text, Vision, or Multimodal processing.