`attachment.type` resolves to `string & {}` via @assistant-ui/store@0.1.6's
generic type chain when installed through npm (package-lock.json), breaking
the `const _exhaustiveCheck: never = type` exhaustive check pattern.
Replace with a direct throw that compiles cleanly across library versions
while preserving identical runtime behaviour.
Fixes#263
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
GGUF was in the global EXCLUDED_TAGS set which filtered it from all
consumers of useHfModelSearch, including the chat page. Move GGUF
exclusion to an opt-in excludeGguf option so only training and
onboarding pages filter out GGUF models.
GGUF models can't be fine-tuned, so hide them from the training/studio
page while keeping them available for inference on the chat page.
- Add "gguf" to EXCLUDED_TAGS in HF model search hook
- Filter local models with .gguf extension or -GGUF in ID
- Replaced `CookBookIcon` with `ChefHatIcon` in navbar for improved clarity.
- Added dark mode-specific gradient styles to recipe cards for better visual differentiation.
Remove os.chdir(save_directory) from export.py which was causing all of
unsloth-zoo's relative-path internals (check_llama_cpp, use_local_gguf,
_download_convert_hf_to_gguf) to resolve against the export directory
instead of the repo root. This caused llama.cpp to be cloned inside each
export dir and destroyed the repo root's llama-server build on cleanup.
Now passes absolute paths to save_pretrained_gguf so unsloth resolves
llama.cpp from the repo root where setup.sh already built it.
Also builds llama-quantize in setup.sh (needed by unsloth-zoo's export
pipeline) and symlinks it to llama.cpp root for check_llama_cpp().
- Replace datetime.UTC with datetime.timezone.utc in authentication.py and storage.py
- Fixes ImportError on Python versions < 3.11
- timezone.utc works on Python 3.9+
Resolves#237
Fixes two bugs:
1. Chat template tags (<|im_start|>, <|im_end|>) leaking into output
because /v1/completions treated them as literal text
2. Image hallucination because image_b64 was never passed to llama-server
Now llama-server handles chat templates natively and receives images
as OpenAI-format multimodal content parts for vision models.
Replace Python-side GGUF download with llama-server's native -hf flag for
HuggingFace repos. Add frontend variant picker so users can choose
quantization (Q4_K_M, Q8_0, BF16, etc.) with file sizes. Fix vision
detection via mmproj files instead of hardcoding is_vision=False.
