Skip to content

Talemate Client (Vision)

New in 0.36.0

The Talemate Client vision backend allows you to use any vision-capable Talemate LLM client for image analysis.

The Talemate Client backend for image analysis lets you reuse an existing Talemate LLM client connection for vision tasks. Instead of configuring a separate endpoint, you select one of your already-configured LLM clients that has vision capabilities enabled.

Prerequisites

  • A Talemate LLM client configured and connected with a vision-capable model
  • Vision must be enabled in the client's settings (see Enabling Vision on a Client below)

Enabling Vision on a Client

Before a client can be used for image analysis, you need to enable its vision capabilities:

  1. Open the client's settings by clicking on the client in the sidebar
  2. Look for the Vision configuration section
  3. Enable the Enable Vision toggle

Vision configuration in LLM client settings

Model Must Support Vision

Only enable vision if the model currently loaded on the client actually supports multimodal vision input. For local models this generally means a mmproj (multimodal projector) file was loaded alongside the model. Enabling vision on a text-only model will result in errors when image analysis is attempted.

Vision support is currently available on the following clients:

  • KoboldCpp
  • llama.cpp
  • text-generation-webui

Setup

  1. Ensure at least one of your LLM clients has vision enabled (see above)
  2. Open the Visualizer agent settings
  3. Set the Backend (image analysis) dropdown to Talemate Client
  4. In the backend settings that appear, select the client to use from the Client dropdown

Talemate Client vision backend settings

The client dropdown only shows clients that have vision enabled. If no clients appear, you need to enable vision on at least one client first.

Advantages

Using the Talemate Client backend has several advantages over configuring a separate OpenAI Compatible endpoint:

  • No duplicate configuration -- reuses your existing client connection and credentials
  • Model awareness -- the backend automatically knows which model is loaded
  • Centralized management -- client status, connection, and model changes are handled in one place
  • Dynamic selection -- the client list updates automatically when you enable or disable vision on clients

Usage

Once configured, the backend is used automatically whenever image analysis is requested. The selected client handles the vision request using the same connection it uses for text generation.

Troubleshooting

No clients in dropdown: Ensure you have at least one client with vision enabled. Open your client settings and check the Vision section.

Client not connected: The selected client must be connected and enabled. Check the client's status in the sidebar.

Vision not enabled error: The selected client has vision_enabled set to false. Open the client settings and enable vision in the Vision section.

Analysis errors: Verify that the model loaded on the client actually supports multimodal vision input. Some models may be listed as vision-capable but require specific configurations.