Resources

Non-thinking generation prompt uses non-canonical `<think>\n</think>\n` → duplicated output under streaming

#43 opened 7 days ago by

batsclamp

Remove extra newline that breaks prefix cache after tool calls

#42 opened 9 days ago by

choongng

Breaks claude code

#41 opened 13 days ago by

widdae

Fix: Context Retention Fixes After Compression

👍 5

#40 opened 14 days ago by

ross232323

v20: Agent looping/repeating actions in agentic workflows

#39 opened 15 days ago by

kratto

`</think>` strings appears meaningless in content text

#38 opened 15 days ago by

tooltd

preserve_thinking: false or true

👍 1

#37 opened 15 days ago by

vikuloff

122B-A10B 4bit

#35 opened 17 days ago by

O-ishi

Lmstudio Prompt Template override not available in Server Mode only works in Chat tab

#34 opened 18 days ago by

dasss-st

v20 released: C++ AST optimization & minja compatibility patch

👍 10

#33 opened 19 days ago by

froggeric

Will this work on spicyneuron/Qwen3.5-397B-A17B-MLX-2.6bit ?

#32 opened 22 days ago by

openSourcerer9000

Reasoning off mode issue

👍 3

#30 opened 28 days ago by

GergelyZsolt

Still issues with tooling from time to time....

#29 opened 28 days ago by

boldos

Response will sometimes show full thought process to Frigate, which will fail to create a GenAI Summary

#27 opened about 1 month ago by

ampersandru

toolcall inside thinking

👀 3

#25 opened about 1 month ago by

snapo

Keeps emitting <tool_call> json into the context

➕ 1

#23 opened about 1 month ago by

theliphant