Error codes Models

PULSE-MODEL-005 No connection to the model

What does this mean

Pulse sent a request to the model server and did not receive a response: the server is not running, is not accessible over the network, or has been silent for longer than Pulse is willing to wait. The agent's progress is stopped; what the agent managed to do remains in the working copy of the task.

Why does this happen

The built-in Pulse agent goes to the model server at the address from “Settings → Models”. There is no connection when:

  • local server (Ollama, LM Studio, llama.cpp) is not running or is listening to a different address and port;
  • the server is behind the VPN or on the organization’s network, and the VPN is turned off;
  • the model thinks for more than ten minutes, and Pulse aborts the request by timeout.

In “More details” - the words Pulse “request to the model: fetch failed” with the system code (ECONNREFUSED, ENOTFOUND, ETIMEDOUT) or “the request to the model was interrupted by a timeout.”

What to do

  1. Start the model server and check the address: the “Check” button in “Settings → Models” will show whether it responds.
  2. Server behind VPN - enable VPN.
  3. Click “Repeat turn” in the error card - in the session feed or in the task. Pulse will send the same request again.
  4. If you have a spare model point, assign it in “Settings → Models”: in case of a break, the main turn will continue to the spare one.

When to write to support

Write to us if the model server responds to other programs at the same address, but Pulse writes that there is no connection. Click “Send report” in the error card - a report form will open, where the code and details have already been entered, the secrets are hidden in them. Add which server and which address are indicated, and click “View and Send”: the window will show exactly what is left - check and click “Submit”. If the cloud is turned off, the button is called “View and save” - send the resulting file to us in any way.