- Introduced new JSON files for normalized and raw inputs, source metadata, and public inputs.
- Added findings document detailing the methodology and results of the score reconstruction.
- Included member differences and worked examples for clarity on data discrepancies.
- Implemented a Python script for reproducing scores based on public inputs.
- Created SQL for raw reaggregation of data.
- Added task metadata for tracking the research completion and validation metrics.
- Updated bundled skills documentation to clarify the structure and usage across all platforms, ensuring consistency in skill root paths.
- Introduced a new `inject-spec-context.py` hook for path-scoped spec context injection, improving the relevance of injected specs during file interactions.
- Enhanced existing hooks to support workflow resolution, allowing for dynamic selection of workflows based on task context.
- Added a command to manage workflow selections for active tasks, enabling better task management and workflow adherence.
- Updated configuration options for spec injection, including character limits and refresh windows, to optimize performance and usability.
- remove the sector_type parameter from the selection port, adapter, and
use case; the adapter now passes SectorType.INDUSTRY directly, deleting
the dead concept vocabulary mapping
- stop parsing concepts/concept_total/concept_limit from the sector-radar
membership response in the web API layer and narrow the response type to
industries only
- update tests and the task PRD accordingly; backend pyright errors drop
from the 16 baseline to 14
- Introduced ActiveMoneyflowSource to fetch optional active-order flow, enhancing the sector radar's data capabilities.
- Updated StockFactRecord and DailyAggregateRecord to include pct_change and active_buy_net_amount_yuan for improved financial insights.
- Modified the build process to incorporate active moneyflow data without invalidating main rankings on failure.
- Enhanced the HTTP API to return detailed sector history and metrics, including pct_change and active buy metrics for members.
- Updated tests to validate the new functionality and ensure data integrity across various scenarios.
- pin the selection sector vocabulary to industry: the /sectors endpoint
no longer accepts sector_type, and the port, adapter, and use case all
resolve counts and member codes with sector_type="industry"
- relabel the results filter to 细分行业 and drop concept-type plumbing
from the frontend API, query keys, and types
- show only 细分行业 in the signal detail panel; remove the concept-board
chips and keep the snapshot-date tooltip on the industry line
- update backend and frontend tests to the industry vocabulary and record
the revised scope in the task PRD
- move the filter bar out of the list column into a full-width top section;
list and detail stay side by side below on md+, stacking search > list >
detail on small screens
- add sector select fed by the run's concept-board aggregates; options show
per-sector stock counts ordered by count desc, stale selections reset when
the aggregate list changes
- persist sector in the route search, results/runs query keys, and refresh
sector aggregates when a selection run finishes
- sector_radar: add batch sector-count aggregation and sector member lookup
over the strict last-good membership snapshot (postgres + in-memory fakes)
- selection: add SelectionSectorReader port, list_sector_counts use case,
and sector_stock_codes filtering via run identity resolution; queries stay
inside the selection context per ADR 0001
- http: add GET /api/v1/selection/sectors and forward sector param on
/results and /runs/{run_id}
- fix stale positional args in pattern-scoring run tests; cover new behavior
with read-service, application, and HTTP contract tests
Add point-in-time stock membership lookup to the sector radar module
(GET /sector-radar/stocks/{ts_code}/membership) reading the existing
dc_index/dc_member snapshots, and surface industries plus concept chips
in the selection signal detail panel with graceful no-data hiding.
Match the ZXB1 sub-chart formula: each brick spans between yesterday's
and today's brick value via a stacked range bar — rising bricks are
hollow red (solid on strong_red), falling bricks solid green. The chart
API now also returns brick_strong_red per point.
Add an optional brick_chart series to the selection chart API, computed
from the shared gold-brick formula when strategy=gold_brick, and render
a fourth grid with red/green brick bars between the volume and J grids.
yuxuanhui