mudler / mudler/LocalAI

qwythos-9b-claude-mythos-5-1m breaks on the second message

Open
#10,659 4 comments 0 reactions 0 assignees View on GitHub

Nobody has claimed this yet.

bug confirmed high prio
Dominant language
Go
Stars
49.2k
Forks
4.5k
Avg merge
1d 3m
Merged PRs (30d)
239

Description

LocalAI version:
LocalAI v3.12.1 (fcecc12e57be39bad2ebf50cf729408b64409553)
latest-aio-cpu

Environment, CPU architecture, OS, and Version:
ubuntu 24.04 . x86)64 r9 9950x . 89 GB DDR5
Linux linshyv 7.0.0-27-generic #27-Ubuntu SMP PREEMPT_DYNAMIC Thu Jun 18 19:13:49 UTC 2026 x86_64 GNU/Linux

Describe the bug
i can ask one question using the chat tab to a model before it spits out a rpc error on the second question

To Reproduce
1start the aio cpu image. let it download everything until it's done.
2 install the qwythos-9b-claude-mythos-5-1m model
3/4 just send it hi twice and on the second message it'll spit out

Internal error: rpc error: code = InvalidArgument desc = Unable to generate parser for this template. Automatic parser generation failed:

While executing CallExpression at line 151, column 28 in source:
... {%- else %}↵ {{- raise_exception('Unexpected message role.') }}↵ {%- ...
^

Error: Jinja Exception: Unexpected message role.

Expected behavior

i expected the model to just keep answering questions as it did on the first one while keeping in mind previous questions(context)

Logs

Additional context
gpt 4 and gpt4o do not exhibit this behavior

this is on default settings nothing altered

Contributor guide

Open the contributing guide

First steps

  1. Read the whole issue, then the project's contributing guide.
  2. Comment on the issue to say you are picking it up — it saves two people doing the same work.
  3. Fork the repository and make your change on a branch.
  4. Open a pull request that references the issue number.

Research direction

Reproduce the failure in the chat tab using the latest-aio-cpu image and the qwythos-9b-claude-mythos-5-1m model, sending two messages. Start from the reported automatic parser generation error and its Jinja template trace; done means the second message receives a response while retaining prior context.

Written by the indexing model from the issue text.

Assessment

Domain
ai, backend
Issue type
Bug
Difficulty
4/5
Estimated time
3-5 days
Activity status
Quiet
Clarity
Mostly clear
Newbie friendliness
45/100

Get new issues in your inbox

A short digest of beginner-friendly GitHub issues.