Conversation
- Apply compact-tool-card-wrapper--dense-command to read/grep/glob/ls and delete cards\n- Drop redundant rightStatusIconWithDivider on search/list/read headers for cleaner density
- Add 37 missing keys to zh-TW/flow-chat.json (widgetContextMenu, exploreRegion.commandCount*, terminalRegion, generativeUI, generativeWidget) - Add 45 missing keys to zh-TW/panels/files.json (search index backend, index actions/phases/summary/indicator/taskState, notifications) - Remove 4 unused keys (stopSubagent*) from zh-CN and zh-TW flow-chat.json that have no en-US counterpart and no code references - Translate advancedShow/advancedHide in Installer zh/zh-TW locales - Fix zh-TW.ftl header comment: s/簡體中文/繁體中文/ Generated with BitFun Co-Authored-By: BitFun
… fixes (GCWing#602) - Wire desktop tool API ToolUseContext with remote WorkspaceBinding and workspace services - Propagate workspace_services into execution engine tool availability context - Ignore whitespace-only stream chunks for effective output / first visible timing - Short-circuit snapshot session files and stats for remote workspace paths - Add unit tests for remote skill discovery and whitespace stream handling
…dense-header fix(web-ui): tighten compact tool card headers for file ops
Move compact-tool-card-wrapper--dense-command from individual tool cards into CompactToolCard so all compact cards share the same header density without per-card duplication.
…dense-header fix(web-ui): flow chat compact tool card dense headers
- Track structurally present reasoning/thinking fields through stream processing and round execution. - Replay empty reasoning content for OpenAI and Anthropic adapters when providers require it. - Add regression tests for empty reasoning/thinking stream cases and provider message conversion.
…ormats - expose DeepSeek reasoning effort as high/max in the model config UI - add DeepSeek reasoning_effort handling for OpenAI-compatible and Anthropic-compatible requests
feat(ai): support DeepSeek reasoning effort and preserve empty reasoning replay
- prevent remote exec from stopping at `Eof` before `ExitStatus` - keep consuming SSH channel messages until the channel actually closes
…ime refactor (GCWing#610) Backend changes: - Remove orphan ModelRound struct and model_round.rs module; the actual round lifecycle is fully tracked through execution engine state - Simplify DialogTurn / DialogTurnState — only TurnStats and new_turn_id helper survive; on-disk shape lives in SessionState as before - Refactor execution engine, round_executor, and stream_processor to align with the simplified runtime model - Add user-steering support in coordinator and scheduler: mid-turn messages from the user can be injected without aborting the current round - Improve agentic tool pipeline (pipeline/tool_pipeline.rs, types.rs), deep-review policy, and review specialist agents - Anthropic stream handler and tool-call accumulator hardening - MCP server manager reconnect and interaction improvements - Snapshot service dedup / reference-safety fixes - Various service cleanups: session manager, persistence, side_question, project context, search, workspace Frontend changes: - Add PendingQueuePanel: shows queued user messages above the chat input; supports inline edit, "send now" mid-turn steering, and delete - Add UserSteeringBubble: renders a user-steering flow item as a normal user bubble inside the active model round - Add PendingQueueModule: manages per-session pending message queue state - Add modelRoundItemGrouping: groups consecutive flow items by model round for virtualised rendering - Update FlowChatManager, EventHandlerModule, TextChunkModule, and MessageModule to handle steering events and pending queue lifecycle - Update modernFlowChatStore and FlowChatContext for new item types - Update AgentAPI with steering / queue endpoints - Update flow-chat types with FlowUserSteeringItem - Update AgentCompanionDesktopPet click-through and pixel-pet UI - Update ChatInput / ChatInputPixelPet for queue-aware send behaviour - Theme minor colour fixes across all presets - i18n: add pending-queue and steering keys for en-US, zh-CN, zh-TW; remove stale error keys - Remove deprecated SessionConfig option and related locale keys
…-multiarch fix: suppress windows helper command popups
fix: correct remote SSH command completion handling
GCWing#611) - Align desktop tool API workspace binding with remote connection lookup and session identity - Propagate remote_connection_id and remote_ssh_host into subagent SessionConfig - Allow MCP local stdio servers in remote workspace sessions - Update miniapp host dispatch, service config types, coding-selfie git errors, and stream replay tests
- Extract AppearanceConfig from Basics with i18n (en/zh-CN/zh-TW) - NavPanel/sessions/workspaces and shell menu state tweaks - Add fixedPopoverViewport helper; refresh ChatInput styles
feat(web-ui): appearance settings tab, nav layout, chat input styles
… activity (GCWing#614) - Treat whitespace-only assistant output as non-visible in round executor retries. - Allow explicit skill tool invocation for default-hidden builtin team skills (remote/workspace). - Finalize finishing turns when backend reports Idle; wire handleSessionStateChanged with context. - Companion pet uses rest mood when no active transient tracked turn; add Vitest coverage.
…ifecycle fix: flashgrep lifecycle cleanup
- Remove agent mermaid_interactive tool, desktop APIs, and announcements - Remove standalone editor scene, flow chat tool card, and mermaid-editor tooling UI - Trim FlexiblePanel and ide-control types; keep Markdown MermaidBlock rendering - Fix MermaidBlock SVG cache key regression; drop mermaid export E2E spec
…ctive-editor chore(core, desktop, web-ui): remove Mermaid interactive editor
…op_host - Nav: browser as footer primary control, insights in overflow menu (see prior diff) - desktop_host: import log::info to fix CI (info! macro in scope)
…ghts-menu fix(web-ui): nav footer browser shortcut and insights in menu
…list subagent
Adds an opt-in Pro mode to the deep-research agent triggered by user messages
beginning with `深度研究：` or `deep:`. Pro mode runs a 6-phase quality
protocol: query understanding → 4 parallel specialists → citation registry →
adversarial debate (2 rounds) → fact check → manager arbitration → report.
The existing longitudinal/cross-sectional flow stays as the default mode.
Introduces ResearchSpecialist, a new readonly built-in subagent with
WebSearch/WebFetch/Read tools. The previous research flow dispatched to the
Explore subagent, which only carries code-search tools and silently produced
empty findings for web research. ResearchSpecialist returns markdown via the
Task result and the parent agent owns all file writes — required because the
Task tool only treats readonly subagents as concurrency-safe.
Adds bilingual research policy: the parent detects USER_LANG and drives all
prose in USER_LANG, while specialists issue queries in both USER_LANG and
English so source ecosystems on both sides are covered. Quotes stay verbatim
in their source language.
Adds {SESSION_ID} placeholder substitution in prompt_builder so Pro mode can
anchor a per-session WORK_DIR under .bitfun/sessions/{SESSION_ID}/research.
Adds AskUserQuestion to the deep-research agent's default tools for Phase 0
clarification and Phase 5 GAP confirmation.
Co-Authored-By: Claude Opus 4.7 <noreply@anthropic.com>
feat(update): download daily app updates in background
Session model updates still sync into Flow Chat state; downgrade the listener log and remove duplicated warning UX that raced on restore.
…-flashgrep bundle linux flashgrep in windows builds
fix(flow-chat): prevent card toggle while selecting text
feat(cli): expose ACP server and external agent controls & prevent ACP write turns from hanging
Move remote command and response DTOs, model catalog DTOs, poll response assembly, and model catalog delta policy into services-integrations while keeping core as the execution adapter. Add owner-crate contract tests and boundary checks to preserve existing remote poll, image, cancel, and wire-shape behavior.
Move MiniApp allowlist and function-agent AI response parsing policy into bitfun-product-domains while keeping host execution, AI calls, JSON extraction, and runtime adapters in core. Add contract tests, boundary-check anchors, and plan docs that mark runtime owner migration as still deferred.
Move remote tracker storage and lifecycle into services-integrations while keeping core as the host adapter for product runtime state. Add owner-crate contract coverage plus boundary and planning docs for the current remote-connect closure.
Keep PR detail scroll position stable when switching changed files, mark draft and ready PRs in the queue, collapse long overview text by default, constrain the changed-file list height, and make generated review drafts focus on actionable findings instead of publishing selected summary comments by default.
Add repeatable Web UI bundle-size and main-entry static graph reporters for the staged startup optimization work. The graph reporter supports dependency reachability and direct-import assertions, so future changes can verify that optional runtimes stay out of the startup path without changing Vite warning thresholds.
Route startup tool initialization through narrow Git and LSP entry points so main.tsx no longer imports the full tools barrel. Replace lucide namespace lookups with explicit icon maps while keeping fallback behavior for unknown icons. Defer Mermaid and Prism syntax highlighting to cached async loaders, keeping plain Markdown and code fallback rendering available before optional runtimes load.
locale: fix Nursery->Assistant, Office Session->Cowork Session
GCWing#752) Honor IME-owned Escape/Enter in chat inputs and global shortcuts so East Asian IME candidate windows behave correctly. Extend modern flow tool cards with transitional styling, refine explore grouping around active or recently finished tools, and surface DeepSeek reasoning effort in the thinking-mode control. Improve terminal expanded output min height for stable layout during streaming.
SWangHash
force-pushed
the
rebase
branch
2 times, most recently
from
c503e1f to
a4c2986
Compare
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
9 participants
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.
No description provided.