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().
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.
- Added constrained dependency files for single-env installations: `constraints.txt`, `data-designer.txt`, and `data-designer-deps.txt`.
- Implemented a `patch_metadata.py` script to resolve metadata conflicts between dependency versions.
- Updated `setup.sh` to integrate single-env setup, including dependency installation and metadata patching.
- Upgraded `fastmcp` and `websockets` versions in `extras.txt` for compatibility.
- Commented out unused "Start Tutorial" button in `data-recipes-page.tsx`.
- Added support for configuring markdown note block styles, including color and opacity.
- Enabled double-click on markdown notes to open their configuration dialog.
- Adjusted layout styles in markdown previews for better interaction control.
- Updated relevant payloads, types, and UI logic to support added styling features.
- Integrated multiple example notes in learning recipes for better visualization.
- Added "Markdown Note" block to allow users to add UI-only markdown notes to the canvas for documentation purposes.
- Integrated note creation, editing, and rendering in the `recipe-studio` UI, including markdown previews.
- Updated payload generation logic to omit markdown notes from backend payloads.
- Enhanced block types, definitions, and dialog support to include the new "Markdown Note" feature.
- Introduced "Multi-Turn Chat" recipe to generate structured user-assistant conversations with domain/topic-based goals and constraints.
- Added `conversation.json` with model configuration, sampling strategies, and LLM prompts.
- Updated UI nodes, layout, and graph rendering logic to support new recipe.
- Enhanced `recipe-studio` fit view logic to improve editor layout responsiveness.
- Added three new learning recipes: "Instruction from Answer," "PDF Grounded QA," and "Structured Outputs Jinja," with respective metadata and configuration.
- Integrated support for unstructured and structured input handling, including sampling strategies, prompt definitions, and model specifications.
- Enhanced JSON structure and UI nodes to facilitate better recipe visualization and execution.
