Resume streaming output

[veeyenkay]

For a model hosted using vLLM, I am using completions endpoint in streaming mode (stream = true). Due to network issues, the connection to the vLLM server may be lost. When the network issue gets corrected, the client will create a new connection. Is there a way by which I can resume the streaming from the last generated token?
If needed, I can provide the previously generated tokens as well.