1. Export route: stop_training() only signals the subprocess — wait up to
30s for it to actually exit before loading the export checkpoint, avoiding
a GPU memory race.
2. Training reset: clear _should_stop so /api/train/status returns phase=idle
instead of staying stuck on phase=stopped after a user-triggered stop.
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
Inference now runs in a persistent subprocess, solving the same
transformers version-switching problem that was fixed for training.
The subprocess stays alive between requests (model in GPU memory)
and is only restarted when switching transformers versions.
New files:
- core/inference/worker.py: subprocess entry point with command loop
- core/inference/orchestrator.py: parent-side proxy with same API
Modified:
- core/inference/__init__.py: exports orchestrator as default backend
- routes/inference.py: removed in-process ensure_transformers_version()
- Add 200-sample parallel probe using ThreadPoolExecutor + safe_num_proc
to estimate download speed and failure rate before full conversion
- Abort with clear error if >=30% of probe images fail to download
- Show estimated download time in the training overlay modal
- Parallel batch conversion for URL-based datasets (vs sequential for local)
- Add warning field to /check-format response for URL-based image datasets
- Display URL warning in dataset preview dialog (amber banner)
- Thread progress_callback from trainer through format_and_template_dataset
to convert_to_vlm_format for real-time status updates
Add Start/End index inputs under Advanced in the dataset card,
allowing users to slice a dataset by row range before training.
Wired end-to-end: frontend store, API payload, backend Pydantic
model, and trainer dataset loading (inclusive on both ends).
Add Start/End index inputs under Advanced in the dataset card,
allowing users to slice a dataset by row range before training.
Wired end-to-end: frontend store, API payload, backend Pydantic
model, and trainer dataset loading (inclusive on both ends).
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.
- Extracted shared execution utilities into `execution-helpers.ts` for reusability across features.
- Replaced deprecated `/preview` endpoint and its logic with unified job execution handling.
- Consolidated job execution flows ("Preview" and "Full Run") into shared `runJobExecution` logic.
- Enhanced execution progress tracking with support for column-level progress reporting.
- Added support for handling execution job events and improved error reporting from the backend.
- Updated backend to better manage dataset access errors and provide more informative error messages.
- Cleaned up redundant code in `use-recipe-studio-actions` and streamlined execution APIs.
- Introduced backend changes to handle dataset pagination with limit, offset, and total row support.
- Updated frontend execution view with dataset pagination controls, including "Next" and "Prev" buttons.
- Extended recipe execution logic to manage dataset pagination details like page number, page size, and total records.
- Introduced "Full Run" support in execution logic, including progress tracking, cancellation, and job status updates.
- Extended backend to manage full execution jobs, handle dataset previews, and return detailed analysis and artifacts.
- Updated frontend components to support full runs, with execution sorting, live updates, and detailed execution views.
- Enhanced `ExecutionsView` with progress indicators, status filtering, and dataset preview capabilities.
- Added IndexedDB schema migration to track additional execution metadata.
- Added `ExecutionsView` with execution history tracking, live updates, and detailed data analysis.
- Implemented IndexedDB support via Dexie to persist execution records locally.
- Enhanced backend preview logic to return execution analysis and artifacts.
- Updated studio header with view toggling between "Editor" and "Executions."
