Recently switched to Gemini 2.5 Flash (free tier). Getting this error:
Error: Stream cancelled; RPC from prefill servable to decode servable failed; Failed to close the streaming context; status = CANCELLED: Stream cancelled; RPC from prefill servable to decode servable failed [type.googleapis.com/util.ErrorSpacePayload=‘RPC::CANCELLED’]
=== Source Location Trace: ===
net/rpc/rpc-errorspace-util.cc:19
learning/serving/servables/wiz/remote_wiz_servable.cc:192
learning/serving/servables/wiz/prefill_remote_wiz_servable.cc:229
learning/serving/servables/wiz/wiz_servable.cc:2600
; Failed to run inference for model: go/debugstr
name: “prod-common-global__/aistudio/gemini-v3p1s-rev19-calmriver-sc__main__/aistudio/gemini-v3p1s-rev19-calmriver-sc__2025112500__prefill__variantvlp__6069839c-f472-46e0-90c0-6f87c941837a”
version {
value: 1
}
signature_name: “serving_stream”
The issue started happening only after switching to Gemini 2.5 flash (from 2.0 flash). Has anyone else seen this? Any known workarounds?