I am unsure if this is a bug, or something with my configuration or if I am just too stupid to understand how things should work in current (2026.2) Home Assistant versions, but I just cannot get streaming to work for the whole pipeline...
Wyoming Cloud Streamer itself does process the input in chunks - I can see that thanks to added logging - there are multiple calls to my OpenAI-compatible TTS server.
I also know that my LLM is streaming its replies, because I also added debug logging to the integration I am using on this side (local llms).
However, what I just cannot get working is taking those chunks from the llm and streaming it to the tts.
For some reason I cannot wrap my head around, somehow HA seems to collect all chat_log_delta messages doing nothing, and then finally, after the last message is received, it sends them all in a single message to the streaming proxy...
That can't be right, can it?
I am unsure if this is a bug, or something with my configuration or if I am just too stupid to understand how things should work in current (2026.2) Home Assistant versions, but I just cannot get streaming to work for the whole pipeline...
Wyoming Cloud Streamer itself does process the input in chunks - I can see that thanks to added logging - there are multiple calls to my OpenAI-compatible TTS server.
I also know that my LLM is streaming its replies, because I also added debug logging to the integration I am using on this side (local llms).
However, what I just cannot get working is taking those chunks from the llm and streaming it to the tts.
For some reason I cannot wrap my head around, somehow HA seems to collect all chat_log_delta messages doing nothing, and then finally, after the last message is received, it sends them all in a single message to the streaming proxy...
That can't be right, can it?