issues Search Results · language:Dune language:Python language:JavaScript language:JavaScript language:HTML linked:pr
Filter by
4.9M results
In the Appearance branch, add_field(...) is called first to mount a Select widget, but then the code calls
select_widget.watch_value(...) and add_field( Theme: , select_widget), where select_widget was ...
bug
good first issue
Description
When the WebSocket reconnects after a HA restart or network blip, StateProxy.on_reconnect() calls _load_cache() to
resync state. If get_states times out (default 15s), the exception is caught, ...
area:core
area:websocket
Maintenance type
CI/CD pipeline
Current state
The PR Gate workflow (.github/workflows/pr-gate.yml) has a single job, size-check, that labels every PR with size/S /
size/M / size/L / size/XL based on ...
needs-triage
type: chore
Line if argv != sys.argv: sys.argv = argv references a bare name argv that is never imported or defined anywhere in the
file. Every invocation from the command line that passes through this branch will ...
bug
good first issue
Summary
When a newly-detected conflict matches a prior human-ratified verdict, apply that verdict automatically instead of
re-escalating to the human. Builds on the resolved-contradiction fingerprint ...
Summary
The contradiction detector has no memory of past verdicts: it compares passage pairs per page, so an already-adjudicated
claim re-escalates as a brand-new question every time a new page carries ...
Summary
Answering a pending question must write the ratified resolution back to the source memory file(s) — not merely emit a
sibling raw/answers/ intake doc. Today the correction never reaches the source-of-truth, ...
Summary
Seeded topology canvases render every device as a plain white box with no text. Manually-built topologies (devices
dropped via the editor) render correctly. Root cause is confirmed by a headless-browser ...
bug
Issue class
implementation-slice
Roadmap goal advanced
Player-data source-of-truth / endpoint-backed simulator input reliability
Repo layer touched
runtime-core / player-data ingestion / test-harness ...
背景
检索 选择层 (build_findings)把 L1/L2/L3/L4 候选跨层混排、单一 _score nlargest 取 top-N。两个真问题:
1. 跨层量纲不可信:L1 分可负、L2 乘性、L3/L4 confidence,混排排序是量纲偶然。
2. 某层霸占/淹没:无配额,高分 L2 断言可挤光 L1 事件。
(经评审:相关性精筛已由 grounding LLM 承担,不需要跨层统一打分;真问题用 ...
enhancement

Learn how you can use GitHub Issues to plan and track your work.
Save views for sprints, backlogs, teams, or releases. Rank, sort, and filter issues to suit the occasion. The possibilities are endless.Learn more about GitHub IssuesProTip! Restrict your search to the title by using the in:title qualifier.