Remote main includes PR #694, which updates the MiniMax preset to MiniMax-M3 and removes deprecated built-in MiniMax context-window entries. Merge that provider baseline before applying the transcript context-window estimate fix so the local line keeps upstream provider defaults.
Constraint: Remote main advanced independently with PR #694 while local main carried unreleased commits.
Rejected: Reset local main to origin/main | would discard local-only main commits.
Confidence: high
Scope-risk: moderate
Directive: Preserve MiniMax-M3 provider preset changes when touching provider context-window logic.
Tested: bun test src/server/__tests__/conversations.test.ts src/server/__tests__/conversation-service.test.ts src/server/__tests__/providers.test.ts src/server/__tests__/provider-presets.test.ts src/server/__tests__/provider-runtime-env.test.ts src/utils/__tests__/context.test.ts src/utils/__tests__/contextBudget.test.ts
Not-tested: bun run check:server blocked by expired quarantine entries: server:cron-scheduler, server:providers-real, server:tasks, server:e2e:business-flow, server:e2e:full-flow
Add MiniMax-M3 as the new default for the minimax provider preset and
remove deprecated M2.5/M2.1/M2 entries. MiniMax-M2.7 and the highspeed
variant are kept for users who pinned the previous generation.
Updated places:
- providerPresets.json: default models -> M3, modelContextWindows trimmed
- modelContextWindows.ts: built-in window list mirrors the new set
- provider-presets / providers-real / thinking tests updated to assert
the new default
- .env.example and third-party docs (zh + en) recommend M3
- scripts/repro-issue-247-real.ts default model bumped to M3
Co-Authored-By: Octopus <liyuan851277048@icloud.com>
Large screenshots were being treated like oversized text payloads,
which pushed readable images through an aggressive resize/compression
path before the model saw them. The image path now preserves images
that already fit API byte and dimension limits, prefers lossless or
high-quality compression when only bytes are too large, and only
downscales by vision pixel budget when the image would exceed the
Read token budget.
Constraint: Fixes pasted/read image failures reported in #615 and #677 while preserving screenshot readability requested in #663
Rejected: Always resize to 2048px | would unnecessarily degrade screenshots that already fit provider limits
Rejected: Continue estimating image budget from base64 text length | image blocks are billed/read as vision pixels, not text payload
Confidence: high
Scope-risk: moderate
Directive: Do not lower fallback JPEG quality or long-edge limits without revalidating real screenshot OCR/vision tasks
Tested: bun test src/utils/__tests__/imageResizer.test.ts src/server/__tests__/conversation-attachments.test.ts
Tested: bun run check:server
Tested: bun run check:coverage (changed-line coverage 95.25%; adapters/desktop coverage lanes failed from existing non-target issues)
Tested: GPT-5.5 Read smoke on 1394x4404, 4096x2304, and 7900x900 images
Not-tested: Real resize/re-encode branch in this source worktree because optional sharp/native image processor is not installed locally
Related: #615
Related: #677
Related: #663
Unsupported image rejections from text-only compatible providers should not
poison the session, and low-trust multimodal usage spikes should not make
context indicators report a full window.
Constraint: Third-party Anthropic-compatible providers may report encoded media bytes as usage tokens.
Rejected: Trust all provider usage uniformly | third-party media responses can pin context to 100% incorrectly.
Confidence: high
Scope-risk: moderate
Directive: Do not remove the media-aware fallback without checking text-only provider recovery and desktop context indicators.
Tested: bun run check:server
Tested: focused media/context regression suite
Desktop reasoning effort now follows the selected session runtime instead of the General settings surface. Session metadata records provider, model, and effort so reconnects and prewarm restarts preserve the per-session runtime, while the default effort is max for new sessions.
Constraint: Existing transcript and localStorage runtime data must remain readable without a required migration.
Rejected: Keep effort as a General setting | it made one session change leak into every other session.
Confidence: high
Scope-risk: moderate
Directive: Do not reintroduce global desktop effort controls without proving multi-session isolation.
Tested: bun run check:server
Tested: bun run check:desktop
Tested: bun run check:native
Tested: bun run check:policy
Tested: bun run check:coverage
Tested: cd desktop && bun run test -- src/components/controls/ModelSelector.test.tsx src/stores/chatStore.test.ts
Tested: bun test src/server/__tests__/settings.test.ts --test-name-pattern "effort"
Tested: bun test src/server/__tests__/conversations.test.ts --test-name-pattern "runtime"
Not-tested: Live provider smoke for domestic providers without local credentials.
Bring the verified worktree implementation into local main while preserving the existing local main commits and unrelated dirty working tree changes.
Constraint: Local main already had unrelated uncommitted provider/runtime edits, so the merge only touches the desktop branch feature paths.
Confidence: high
Scope-risk: moderate
Tested: bun run verify before merge; merge completed without conflicts
Not-tested: Post-merge full quality gate on dirty local main
Desktop users need the same branch-from-here workflow that the CLI already exposed, so the branch creation logic now lives in a shared transcript utility and the desktop app routes completed message actions through the server API. The UI hydrates transcript ids after live completions so newly generated turns can be branched immediately without a refresh.
Constraint: Source sessions must remain unmodified while branch sessions inherit the active transcript chain and persistence metadata.
Rejected: Keep a desktop-only branch implementation | it would drift from CLI /branch semantics and duplicate transcript filtering rules
Confidence: high
Scope-risk: moderate
Directive: Do not remove the post-completion transcript hydration without a real-model desktop E2E for just-finished messages
Tested: bun run verify; Chrome Web UI E2E with real gpt-5.5 provider on ports 45678/45679
Not-tested: Provider-specific behavior beyond the configured Sub2API-ChatGPT route
Provider-managed environments should preserve Claude attribution only when the selected model is actually Claude-prefixed. Non-Claude provider models now disable the attribution header so third-party providers do not inherit Claude-specific prompt metadata.
Constraint: Claude model detection is intentionally model-name based to match the requested provider policy.
Rejected: Remove attribution headers globally | official Claude-prefixed models should keep existing behavior.
Confidence: high
Scope-risk: moderate
Directive: Do not broaden this policy without live provider request-shape verification.
Tested: Targeted server, provider, settings, cron, and managed-env tests in prior verification pass.
Not-tested: Full verify gate after this commit.
The contributor PR adds useful Windows terminal and portable mode support, but it also reintroduced an older General settings zoom block and left the new native settings paths without enough regression coverage. This commit keeps the feature direction intact while removing the duplicate UI, making invalid bash paths fail at save time, and covering the portable cache and app-mode paths with focused tests.
Constraint: This commit lands directly on the contributor PR branch to avoid a long review-comment loop.
Rejected: Ask the contributor to rework the PR from scratch | the remaining issues are narrow and maintainable by us.
Confidence: high
Scope-risk: moderate
Directive: Keep future portable-mode changes covered at the native boundary and the desktop store boundary.
Tested: cd desktop && bun run test src/pages/TerminalSettings.test.tsx src/__tests__/generalSettings.test.tsx src/stores/settingsStore.test.ts
Tested: bun test src/utils/__tests__/cachePaths.test.ts
Tested: cd desktop/src-tauri && cargo test
Tested: cd desktop && bun run lint
Tested: cd desktop/src-tauri && cargo check
Tested: bun run check:server
Tested: bun run check:desktop
Not-tested: Manual Windows packaged-app portable-mode smoke; to be covered before a future release.
Desktop sessions must treat the global Thinking toggle as the source of truth across active CLI sessions, title generation, and provider-compatible request bodies. DeepSeek-style streams can start with reasoning blocks before text, so the WebSocket bridge now keeps the UI in thinking state until text content actually starts.
Constraint: DeepSeek and MiniMax use Anthropic-compatible surfaces but expose thinking behavior through provider-specific request and stream shapes
Rejected: Keep provider-specific disabled-thinking env defaults | stale defaults kept overriding the user's global Thinking setting
Rejected: Treat every non-tool stream block as text | reasoning-only startup blocks created a blank streaming gap before the visible thinking bubble
Confidence: high
Scope-risk: moderate
Directive: Do not change thinking stream state transitions without replaying reasoning-before-text provider events
Tested: bun test src/server/__tests__/ws-memory-events.test.ts
Tested: bun run check:server
Tested: bun run verify
Not-tested: Live DeepSeek provider smoke with real quota
Virtualized chat rendering keeps inactive long-running sessions cheap without disconnecting their CLI process, while DeepSeek now relies on the shared Thinking setting instead of a provider-specific disabled-thinking override. Existing legacy DeepSeek managed env is normalized so old local settings do not keep suppressing reasoning output after upgrade.
Constraint: Multiple desktop tabs must keep live sessions running and remain quick to switch without reconnecting.
Rejected: Pause or disconnect hidden sessions | would delay tab switching and interrupt streaming/tool state visibility.
Rejected: Keep DeepSeek disabled-thinking preset | conflicts with the General Settings Thinking control.
Confidence: high
Scope-risk: moderate
Directive: Do not reintroduce provider-specific disabled-thinking defaults for DeepSeek without testing both General Settings toggle states.
Tested: bun test src/server/__tests__/conversations.test.ts -t "global Thinking setting control DeepSeek"
Tested: bun test src/server/__tests__/persistence-upgrade.test.ts src/server/__tests__/provider-presets.test.ts src/server/__tests__/providers.test.ts src/server/__tests__/title-service.test.ts src/utils/__tests__/thinking.test.ts src/utils/__tests__/providerManagedEnvCompat.test.ts
Tested: cd desktop && bun run test -- MessageList.test.tsx ContextUsageIndicator.test.tsx
Tested: bun run check:server
Providers like DeepSeek with 1M context windows can report
input_tokens that, when combined with output_tokens, exceed the
declared context window — causing the UI to show >100% usage.
Add an optional contextWindow parameter to calculateCurrentContextTokenTotal
that clamps the result to the window size. Pass it from analyzeContext
and sessionService where the window size is already available.
Backward compatible: callers that omit the parameter keep existing behavior.
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
Expose local Claude Code CLI transcript usage in Settings so users can inspect recent token consumption and daily activity without leaving the desktop app.
The page uses server-side transcript aggregation for session, message, tool, model-token, and subagent token data. Daily token buckets use assistant message timestamps, and daily session counts use active parent sessions for the same date bucket so resumed sessions and cross-midnight work do not produce token-only days. Cache accounting is bumped to v5 to force recomputation under the corrected daily semantics.
Constraint: Usage data must come from local Claude Code CLI transcripts rather than mock/demo data.
Constraint: Desktop navigation keeps Token usage directly above Diagnostics.
Rejected: Bucket all token usage by session start date | hides resumed-session and cross-midnight consumption from the actual day it was spent.
Confidence: high
Scope-risk: moderate
Directive: Keep daily token and daily session counts on the same date-bucketing semantics.
Tested: bun run check:desktop
Tested: bun run check:server
Tested: Browser verification for Token usage in English and Chinese locale date labels
Not-tested: Full bun run verify quality gate
Bring the detached worktree fix onto local main after verifying the title-generation regression path and local quality checks.
Constraint: The worktree commit was based on an older main commit, while local main already contained later chat scroll fixes.
Confidence: high
Scope-risk: moderate
Directive: Keep title sanitation shared across CLI, server, and desktop read paths.
Tested: bun run check:server
Tested: bun run check:desktop
Tested: bun run verify (7 passed, 1 failed on existing aggregate agent-utils coverage baseline)
Not-tested: live model title-generation smoke
Slash-command and skill prompts can enter the title-generation path as internal XML breadcrumbs. The async title request previously treated that transport metadata as user prose, so a generated ai-title could persist raw command tags and override the normal session title.
This routes title sources and generated title output through a shared sanitizer, preserving user-visible command names and arguments while dropping unrelated internal XML metadata. Desktop fallback title rendering now applies the same cleanup so existing transcripts with bad ai-title entries recover on read.
Constraint: Session titles can be produced by both desktop server logic and the SDK generate_session_title control request.
Rejected: Only clean desktop session display | leaves CLI and remote title generation able to persist bad titles again
Confidence: high
Scope-risk: moderate
Directive: Keep generated-title input and persisted-title readback on the shared sanitizer; do not add a title path that reads command XML directly.
Tested: bun test src/utils/__tests__/sessionTitle.test.ts src/utils/__tests__/sessionTitleText.test.ts
Tested: bun run check:server
Tested: bun run check:desktop
Tested: bun run verify (7 passed, 1 failed on existing aggregate agent-utils coverage baseline)
Not-tested: live model title-generation smoke
Chat sessions need predictable navigation when users switch tabs or read history during streaming. This preserves each session's scroll position, defaults fresh sessions to the latest message, and adds a compact jump-to-latest affordance when auto-follow is paused.
The PR gate also exposed that provider-scoped scheduled tasks must force the sdk-cli entrypoint when launched through the sidecar, so the same change records that runtime contract and its regression assertion.
Constraint: Desktop chat should not force-scroll while the user is reading older messages
Constraint: Scheduled task provider env must not inherit stale parent model runtime values
Rejected: Persist scroll positions in localStorage | session scroll is transient UI state and should not survive app restarts
Confidence: high
Scope-risk: moderate
Directive: Do not remove the sdk-cli entrypoint marker from provider-scoped cron tasks without rerunning the sidecar launcher regression
Tested: cd desktop && bun run test -- MessageList.test.tsx
Tested: cd desktop && bun run lint
Tested: bun test src/server/__tests__/cron-scheduler-launcher.test.ts src/utils/__tests__/themeWords.test.ts
Tested: bun run check:desktop
Tested: bun run check:server
Tested: bun run check:native
Not-tested: bun run verify still fails on pre-existing agent-utils global coverage baseline below threshold
Not-tested: Chrome extension E2E blocked by Codex Chrome Extension communication timeout after plugin diagnostics passed
The repository now has a measurable PR quality path instead of a loose set of
manual checks. Coverage, quarantine governance, provider smoke, desktop smoke,
and workflow wiring all produce durable reports that contributors and maintainers
can inspect without reconstructing terminal output.
This also fixes the desktop smoke current-runtime path so browser-driven smoke
runs use the desktop default active provider instead of forcing the official
current model, and records that runtime decision as an artifact.
Constraint: Default PR gates must remain non-live and contributor-safe while live model checks stay explicit.
Constraint: Release packaging is still GitHub Actions based, so release preflight must run before the build matrix.
Rejected: Make live provider or desktop smoke mandatory on every PR | secrets, quotas, and model availability are maintainer-controlled.
Rejected: Let PRs lower coverage baselines in the same change | base-branch ratchet comparison must remain authoritative.
Confidence: high
Scope-risk: moderate
Directive: Do not relax coverage or quarantine policy without a maintainer approval label and a fresh quality report.
Tested: ALLOW_CLI_CORE_CHANGE=1 ALLOW_COVERAGE_BASELINE_CHANGE=1 bun run quality:gate --mode pr
Tested: bun run quality:gate --mode baseline --allow-live --only provider-smoke:* --provider-model nvidia-custom:main:nvidia-custom-main --artifacts-dir /tmp/quality-gate-live-smoke
Tested: bun run quality:gate --mode baseline --allow-live --only desktop-smoke:* --provider-model current:current:current-runtime --artifacts-dir /tmp/quality-gate-desktop-smoke-fixed
Tested: git diff --check
Not-tested: Full live release mode with multiple providers in hosted CI; provider credentials and quota remain maintainer-controlled.
Context inspection was switching from the local context estimate to the latest
provider usage total once a response completed. That provider total omitted the
assistant output tokens, even though those tokens become part of the next turn's
context, so the desktop meter could fall from 11% to 10% after work finished.
The context total now includes output tokens and keeps the local estimate as a
lower bound.
Constraint: Provider usage and local context estimates use different token accounting paths
Rejected: Trust provider input tokens alone | it omits the latest assistant output from the next-turn context
Confidence: high
Scope-risk: narrow
Directive: Context usage totals must represent the next-turn conversation context, not billing-only input tokens
Tested: bun test src/utils/__tests__/context.test.ts
Tested: bun run check:server
Desktop users can now disable thinking for new sessions, and Anthropic-compatible providers that opt in receive explicit disabled-thinking requests across main turns, side queries, and AI title generation.
Constraint: DeepSeek/Kimi/GLM Anthropic-compatible endpoints need a non-thinking path without scattering provider-specific logic through the CLI.
Rejected: Per-model if/else branches | centralized provider preset env keeps the native CLI surface smaller and easier to audit.
Confidence: high
Scope-risk: moderate
Tested: bun test src/utils/__tests__/thinking.test.ts src/server/__tests__/title-service.test.ts src/server/__tests__/provider-presets.test.ts src/server/__tests__/conversations.test.ts
Tested: cd desktop && bun run test -- generalSettings.test.tsx
Tested: cd desktop && bun run lint
Tested: cd desktop && bun run build
Tested: agent-browser DeepSeek desktop E2E with transparent proxy captured main and title requests with thinking.type=disabled
Default desktop sessions with an active third-party provider were only passing the selected model into the CLI startup path. That allowed the child CLI to fall back to stale cc-haha settings env instead of the provider index, so the UI-selected provider could diverge from the key and endpoint used by the subprocess.
Route default sessions through the active provider id and make third-party Sonnet capability overrides explicit with a non-empty sentinel. This keeps provider runtime env sourced from providers.json and avoids treating third-party Sonnet models as first-party thinking-capable models.
Constraint: Third-party provider behavior is the product baseline for desktop users.
Rejected: Rely on synced cc-haha/settings.json env | it can be stale and caused provider/key routing mismatches.
Confidence: high
Scope-risk: narrow
Directive: Third-party desktop sessions must carry providerId into ConversationService; do not depend on global settings env for active providers.
Tested: bun test src/server/__tests__/conversations.test.ts src/server/__tests__/conversation-service.test.ts src/server/__tests__/provider-presets.test.ts src/utils/__tests__/thinking.test.ts
Tested: git diff --check
Third-party Anthropic-compatible providers can expose Claude model names without supporting every first-party runtime capability. Desktop sessions now preserve provider-specific capability overrides, keep selected provider runtime state across reconnects and restarts, and keep the provider dialog behavior aligned with the updated presets.
Constraint: Custom ANTHROPIC_BASE_URL endpoints must not be treated as first-party Anthropic for adaptive thinking support.
Rejected: Rename provider default Sonnet models | that would hide the compatibility issue instead of fixing runtime capability detection.
Confidence: high
Scope-risk: moderate
Directive: Do not enable Claude first-party capability defaults for custom Anthropic-compatible base URLs without provider verification.
Tested: bun test src/utils/__tests__/thinking.test.ts src/server/__tests__/conversation-service.test.ts src/server/__tests__/provider-presets.test.ts src/server/__tests__/conversations.test.ts
Tested: cd desktop && bun run lint
Tested: cd desktop && bun run test src/components/shared/Modal.test.tsx --run
Tested: git diff --check
