- POST /api/recipe-scan/import — bulk ingest from Kiwi scanner pipeline, idempotent by item id
- GET /api/recipe-scan/next — oldest-first pending item for review
- POST /api/recipe-scan/items/{id}/approve|edit|reject — label actions
- GET /api/recipe-scan/stats — counts by status and modality
- GET /api/recipe-scan/export — JSONL training pairs (messages chat format, Option B: correction prompt + extracted draft → corrected ground truth)
- GET /api/recipe-scan/image — path-traversal-safe image serving from /Library/Assets/kiwi/
- SQLite at data/recipe_scan.db with WAL mode; separate from corpus.db lifecycle
- set_db_path() testability seam; 18 tests, all passing
- RecipeScanView.vue: two-column review UI (image left, JSON diff right), keyboard shortcuts A/E/R, toast feedback, stats header, export download
- Route /data/recipe-scan and sidebar nav entry added
Pull-side companion to kiwi#141. Ingests structured JSONL pipeline logs
from /Library/Assets/logs/pipeline/ into the log corpus for Turnstone
logreading model training.
- app/data/log_corpus.py: add ingested_pipeline_files tracking table,
_pipeline_ingest_dir() config helper, _ingest_one_file() parser, and
POST /api/corpus/pipeline-ingest endpoint
- source_host = "pipeline_scrape"; source_id from logger field; extra
dict stored as matched_patterns; batch_type = "pipeline_log"
- Idempotent by filename: skips files already in ingested_pipeline_files
- config/label_tool.yaml.example: add corpus section with pipeline_ingest_dir
and push sources comment block
- tests/test_log_corpus.py: 8 new tests covering ingest, idempotency,
non-JSONL filtering, malformed line resilience, incremental runs
Replace stale llama/mistral/phi model refs with models active on the
cluster: deepseek-r1 (1.5b, 7b-4bit, 0528-qwen3-8b-gguf), granite-4.1-8b,
qwen2.5 (3b, 7b), capybarahermes-2.5-mistral-7b, darwin-9b-opus. Update
benchmark_plans.py doc examples to match.
Add _resolve_task_model() helper that looks up a product.task assignment
from the coordinator and resolves its service_type from the model registry.
Add task_ids param to run_imitate() (comma-separated "product/task" strings)
so the imitate harness can dispatch to models chosen by the assignment layer
rather than requiring explicit model IDs.
- app/models.py: add set_cf_text_models_dir() testability seam
- tests/test_models.py: redirect _CF_TEXT_MODELS_DIR in reset_models_globals
fixture so list_installed() count tests are not polluted by real NFS models
- app/cforch.py: fix get_results() return type annotation list → dict
- tests/test_cforch.py: give _BENCH_RUNNING=True test a mock proc with
poll()=None so the stale-flag check correctly returns 409; patch
_select.select in streaming tests (select requires fileno(), iter() doesn't)
- tests/test_finetune.py: mark GPU integration test @pytest.mark.gpu
- pytest.ini: register gpu and slow markers
Adds corpus.db (corpus_sources, corpus_batches, corpus_entries), a FastAPI router
at /api/corpus with receive/label/skip/stats/export endpoints, and seeds consent
tokens for xanderland + orchard nodes from label_tool.yaml. PII flag excludes
entries from JSONL export. Closes avocet#61.
- dashboard: eval card now shows last run + score for all bench types
(classifier, LLM, style, plans) via new _get_recent_bench_runs()
- dashboard: skip cforch LLM-bench list summaries when scanning for
classifier best_macro_f1 (fixes _find_latest_classifier_bench)
- cforch: stale _BENCH_RUNNING flag now auto-resets if process exited;
idle timeout (120s via select) kills hung benchmark if node crashes
- api: add /api/finetune/{run,cancel} backward-compat shims while
ClassifierTab fine-tune section is migrated to TrainJobsView
- ClassifierTab: migrate all /api/benchmark/* paths to /api/cforch/*;
fix null-safety on results.models access; load fine-tuned models from
/api/train/results instead of /api/finetune/status
- CompareTab: extend model picker to include vllm + cf-text alongside
ollama, grouped by service; pre-select all LLM_SERVICES on load
- LlmEvalTab: null-safety on quality_by_task_type lookups
- models: AVOCET_MODELS_DIR env var overrides default models/ path
- Add AbortController to SSE pull stream in OllamaModelPanel; abort on unmount
- Fix SSE loop: break on success/error events, call fetchModels() after the loop
- Add AbortController to fetchModels() and fetchProfile() one-shot fetches
- Add onUnmounted cleanup to both panel components
- Extract GpuEntry, ServiceInfo, NodeSummary to web/src/types/nodes.ts
- Remove duplicate interface definitions from NodeCard, GpuRow, NodeManagementView
- Fix aria-live regions: persistent container with v-if on inner span (avoids
screen reader announcement miss on initial mount)
- Tighten STATE_LABELS/STATE_ICONS to Record<ServiceState, string> for exhaustiveness
- Add explicit (await r.json()) as NodeSummary[] cast in fetchNodes()
- Replace bare `except Exception: pass` with `except Exception as exc` and a
logger.warning call that surfaces node_id and the exception for diagnostics.
- Move `import os as _os` from mid-file (between test functions) to the
top-level import block to satisfy PEP 8 and linter expectations.
- Remove redundant httpx.ConnectError from nodes except clause (it's a
subclass of HTTPError so the tuple catch was redundant)
- Narrow services except clause from bare Exception to httpx.HTTPError,
add logger.warning with coordinator_url for debuggability
- Move `from unittest.mock import MagicMock, patch` from mid-file to
the top-of-file import block with the other stdlib/third-party imports
- Drop unused StreamingResponse import from app/nodes.py (will be
re-added in Task 2 when the SSE endpoint is implemented)
- Replace em dash with colon in _get_ollama_url HTTPException detail
- Remove unused os and unittest.mock imports from test_nodes.py
(mock imports will return in Task 2 tests)
- Create app/nodes.py with _CONFIG_DIR testability seam, _load_config,
_profiles_dir, _profile_path, _load_profile, _get_ollama_url helpers,
and stub list_nodes endpoint returning [] when no coordinator_url is set
- Mount nodes router at /api/nodes-mgmt in app/api.py
- Add profiles_dir comment to config/label_tool.yaml.example cforch section
- Create tests/test_nodes.py with autouse fixture and two passing tests
- Add embed_model: nomic-embed-text to config/label_tool.yaml (local, gitignored)
- Add # embed_model: commented example to config/label_tool.yaml.example
- Add pyyaml>=6.0 to requirements.txt (explicit dep for _resolve_urls yaml.safe_load)
- Add params assertion to test_embed_knn_nomic_registry_entry
Adds benchmark_plans.py script, plans_bench API router, PlansBenchTab Vue
component, and registers /api/plans-bench in api.py. Also extends models
registry (cf-text catalog integration), cforch client, LlmEvalTab, and
ModelsView with cf-orch fleet support. Wires Planning mode into BenchmarkView.
- GET /api/train/jobs now returns {"jobs":[...]} instead of bare array
- GET /api/train/results now returns {"results":[...]} instead of bare array
- POST /api/train/jobs body key renamed config -> config_json to match Pydantic model
- SSE log handler now handles 'progress' event type (backend never emits 'log')
- Dashboard _get_active_jobs() adds model_key to SELECT and return dict
- corrections.py docstring updated: both /api/corrections and /api/sft prefixes noted
- test_train.py assertions updated to unwrap new envelope shapes
