Since upgrading to 2026.1 getting intent failed on Voice PE and ESP32 voice assistants

Configuration Voice Assistant
1

I’v read several help articles that their is an automation or custom file somewhere with bad intents but I wen through every automation i have and found nothing out of the ordinary. Plus nothing has been modified in months (and several upgrades) so it doesn’t explain why its happening now in 2026.1.

Oddly, voice assitant to google and chatgpt work fine from my Home Assistant mobile. Both as text and voice. So this only effects the ESP32 based assistants.

In the sys logs I get this message, which does seem to indicate I have someone custom somewhere but I’m hoping something can help me track it down.

Log details (ERROR)
Logger: homeassistant.components.assist_pipeline.pipeline
Source: components/assist_pipeline/pipeline.py:1298
integration: Assist pipeline (documentation, issues)
First occurred: 9:02:47 AM (10 occurrences)
Last logged: 9:33:11 AM

Unexpected error during intent recognition
Traceback (most recent call last):
  File "/usr/src/homeassistant/homeassistant/components/assist_pipeline/pipeline.py", line 1298, in recognize_intent
    conversation_result = await conversation.async_converse(
                          ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
    ...<9 lines>...
    )
    ^
  File "/usr/src/homeassistant/homeassistant/components/conversation/agent_manager.py", line 126, in async_converse
    result = await method(conversation_input)
             ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
  File "/usr/src/homeassistant/homeassistant/components/conversation/entity.py", line 55, in internal_async_process
    return await self.async_process(user_input)
           ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
  File "/usr/src/homeassistant/homeassistant/components/conversation/entity.py", line 68, in async_process
    return await self._async_handle_message(user_input, chat_log)
           ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
  File "/usr/src/homeassistant/homeassistant/components/openai_conversation/conversation.py", line 84, in _async_handle_message
    await self._async_handle_chat_log(chat_log)
  File "/usr/src/homeassistant/homeassistant/components/openai_conversation/entity.py", line 485, in _async_handle_chat_log
    messages = _convert_content_to_param(chat_log.content)
  File "/usr/src/homeassistant/homeassistant/components/openai_conversation/entity.py", line 186, in _convert_content_to_param
    output=json.dumps(content.tool_result),
           ~~~~~~~~~~^^^^^^^^^^^^^^^^^^^^^
  File "/usr/local/lib/python3.13/json/__init__.py", line 235, in dumps
    return _default_encoder.encode(obj)
           ~~~~~~~~~~~~~~~~~~~~~~~^^^^^
  File "/usr/local/lib/python3.13/json/encoder.py", line 202, in encode
    chunks = self.iterencode(o, _one_shot=True)
  File "/usr/local/lib/python3.13/json/encoder.py", line 263, in iterencode
    return _iterencode(o, 0)
  File "/usr/local/lib/python3.13/json/encoder.py", line 182, in default
    raise TypeError(f'Object of type {o.__class__.__name__} '
                    f'is not JSON serializable')
TypeError: Object of type time is not JSON serializable
2
3

Its different I’m afraid. I don’t get any response back. I get a red flashing light and this error in debug:

finished: 2026-01-08T14:33:12.000Z
error:
  code: intent-failed
  message: Unexpected error during intent recognition
4

Don’t ask it for the time, or in your AI prompt ensure the time/date looks like this :

The current time is {{ now().strftime('%H:%M:%S') }}. Today's date is {{ now().strftime('%Y-%m-%d') }}

Oh and restart HA - so the “history” is gone.

5

Cause is same I think. time

Neil recommendation may be good fix so probably no need to worry

6

I did too many steps at once unfortuantely but the thing that sort of fixed it is I removed "Terminal & SSH’ addon on and retested and now Gemini works. ChatGPT still doesn’t but it now seems to be the issue you posted. I no longer get the red flashing light, it just replies it can’t get a resonse.

I found an obscure post about that add on messing up the Python version and since I don’t use it I figured I might as well remove it and retest.

I’ll reboot and try again to see if the logs show something different. I was using time as a test as that worked (I think it was local) but “how many lights are on” failed with red ring because it had to go out to AI model to do the math.

7

This is the issue.

We’re discussing it in github but the root is a bit of code that works a little too well. A tweak was made to time intents and it now implements something that was updated in 2026.1.

That will have to be changed to output a format openai api compatible integrations can use correctly.

Impacts all openai api compatible integrations

The author of the affected code is looking at his options. Which will likely include a patch tonight that would then likely be in Friday’s build.

My recommendation if you’re using llm skip 2026.1.0

1 Like
8

Thank you. Unfortunately I think I got caught up in two issues. Gemini and GPT doing the red ring of death on Voice PE (not related to the bug), and then after fixing that issue, just GPT not working (because of your aformentioned bug).

I’m good with just Gemini TBH. I think it generally works better for me.

I’m still pretty sure the fix was fully removing Terminal and SSH add on, in case that helps anyone else, but if not it was just a random issue that self resolved after about 6 hours which seems suspicious.

1 Like
9

It’s not Terminal and SSH as I use that. It’s the time format. You likely restarted HA and cleared the history it was trying to send (with the badly formatted time)

10

I appreciate the info, as I am having the same error. Where I am sitting, it is currently a little after 1pm on Saturday, the 10th. Will there be a patch, or incremental update anytime soon, or should I try to downgrade to the last version (never done that before and I am not anxious to try lol)

11

They typically release patches on Fridays. I’d expect end of the week. If you need that functionality you need to roll back and skip the release.

1 Like
12

Looks like everything is back with 2026.1.1. Chatgpt and Gemini are working for me.

13

I have the issue on my Android devices and Assist in web browser on my PC, looking everywhere.
also I can disable OpenAI and use just HA Cloud, same issue

14

It’s a different issue than this one. They’ve fixed this. They also tried to implement streaming in the android client I suspect yours is that.

1 Like