Repository navigation
Release v1.0.3 — /modelRank interleaved ranking, runaway-scroll fix, full-superset config lists - #6
Merged
Merged
Conversation
…nd force-keep pool-configured models 1. Persist pre-pruning candidates (every conversable model of credentialed providers) in shell-superset.json; config surfaces read candidates first, falling back to the injection face, then the bundled manifest (shells semantics unchanged for banner/deny consumers) 2. selectInjectableDefs gains keepModelIds: models referenced by pool-config.json stay injectable and dispatchable; runtime chains keep the pool-selection ∩ activation (favorites/visible set) intersection semantics 3. Add tests: keepModelIds force-keep and composition (shells-select), candidate-source precedence, cross-provider dedup and fallbacks (candidates-cli)
… movement 1. Feeding onMove back into the `current` prop re-fired the host DialogSelect's current-effects on every mouse-over; the centering re-scroll kept sliding fresh rows under the stationary cursor, each firing another onMouseOver — an unbounded self-sustaining scroll loop 2. Drive `current` from a separate cursor signal updated only on explicit ctrl/alt+up/down hotkey moves; highlight tracking via onMove is unchanged
…ase-score neighbors 1. Merged ordering: manual entries interleave with base-score models by effective capability (dialog + rank list), no longer a block on top 2. One-spot moves anchor the moved model's manual score between its new neighbors (same-tier midpoint, ±0.001 decimal step at tier boundaries and tie groups); unranked models materialize an entry on first move 3. Persist as an optional `scores` side-map in capability-rank.json; entries without a stored score keep the legacy ladder semantics, and CLI rank set/add/remove preserve surviving anchored scores
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Summary
Release v1.0.3 from
dev.Added
/modelRankinterleaved ranking — manual entries merge into the base-score ordering (one merged list, no more manual block on top); a move anchors the model's manual score between its new neighbors (same-tier midpoint, ±0.001 decimal step at tier boundaries / tie groups), so one-spot nudges work without ranking every model above. Anchored scores persist as an optionalscoresside-map incapability-rank.json; legacy/hand-written files keep ladder semantics; CLIrank set/add/removepreserve surviving anchored scores./modelRank&/poolConfig— every conversable model of credentialed providers is now listable/configurable (not just the pruned injection face); pool-configured models stay force-kept in the lists.Fixed
/modelRankdialog — hover changes were mirrored into the dialog'scurrentprop, re-triggering the host's centering scroll in a self-sustaining loop;currentis now driven only by explicit hotkey moves.中文摘要
/modelRank交错排名:手动条目与基础分模型合并排序,上下移动一次即锚定介于新上下邻居之间的手动分(同档中点、跨档/同分 ±0.001 小数步进),无需整列重排;锚定分存于capability-rank.json可选scores边表,旧文件保持阶梯语义,CLI 写入保留既有锚定分。/modelRank//poolConfig候选扩为全量超集:有凭据 provider 的全部可对话模型均可列出并配置,池内已配置模型强制保留。/modelRank弹窗鼠标划过列表后疯狂滚动:悬停选中不再回写current,仅快捷键移动驱动。Test plan:
bun run typecheck,bun test(438 pass),bun run build— all green on this branch.