Repository navigation
Solenta improvements roadmap: sessions, Claude usage, reviews, speech and reliability #1869
Description
Track the approved feature and reliability backlog. Child issues describe Solenta behavior and acceptance criteria independently.
Plan:
- File the 34 feature/improvement candidates and 12 investigation groups, avoiding duplicates with existing work.
- Implement and verify the first confirmed reliability batch in an isolated branch: dictation boundaries, file-link detection, and Git diagnostic language.
- Keep the remaining work queued with dependencies and priorities. Investigations must reproduce a local problem before becoming fixes.
Existing capabilities such as native Claude quota reading, usage sorting, transcript density, encrypted secret cards and task-run navigation should be preserved and reused. Optional product choices remain queued until their issue is selected for implementation.
Activity
currentbits commentedon Nov 7, 2026
Backlog filed. Optional items remain product choices; investigations require evidence before implementation.
First implementation batch: #1892 dictation insertion, #1896 file links, #1899 Git diagnostic language. Branch: coder/reliability-first-batch. Other work remains queued.
currentbits commentedon Nov 7, 2026
First implementation batch is ready for review: PR #1916, covering #1892, #1896 and #1899 in three commits. Validation passed: 3,790 full renderer tests, 197 Composer/Markdown integration tests, 22 final focused integration tests, 18 helper tests, 60 Git/SSH/WSL tests, 24 checkpoint/status tests, typecheck and whitespace checks. The other 43 child issues remain queued; these three stay plan:doing pending review and merge.
currentbits commentedon Nov 7, 2026
Starting the next batch with three GPT-6.1 workers, one issue per isolated branch: #1870 (Claude compaction-aware context meter), #1877 (add PR comments to draft context), and #1898 (visible settings-load failure and Retry). Lead will review their committed changes and run integration checks. Also checking CI/review feedback on the previous batch, PR #1916.
currentbits commentedon Nov 7, 2026
Note
🤖 gpt-6.1-sol responding on behalf of currentbits
Second three-issue batch is ready for review in PR #1917: #1870, #1877 and #1898. All implementation workers used GPT-6.1 (
gpt-6.1-sol). Integrated validation passed: 3,801 renderer tests, 67 focused backend tests, typecheck/preload sync and production build. Compaction estimates remain explicitly uncertain where upstream CLI data cannot prove the trigger. The previous batch, PR #1916, also received compiler-position link coverage and fixes for three stale SSH fixture expectations; all 342 tests in files initially blocked by a missing local Electron binary pass after restoring the matching binary. PRs remain open for review; no merge was performed.currentbits commentedon Nov 7, 2026
Note
🤖 gpt-6.1-sol responding on behalf of currentbits
Merge authorization received. Babysitting PRs #1916 and #1917 through current-head review and required CI, then merging in that order. The compiler-path review on #1916 is fixed and resolved. A GPT-6.1 worker is correcting a valid overlapping-settings-save review finding on #1917. After both merges, the next batch will again use three gpt-6.1-sol workers, one issue each.
currentbits commentedon Nov 7, 2026
Note
🤖 gpt-6.1-sol responding on behalf of currentbits
PR #1916 is merged as 606efe2 and #1892/#1896/#1899 are complete. PR #1917 is green on 760bd23: all 14 checks pass, all six required checks are green, there are no unresolved review threads and the branch is current with main. Submitting its squash merge. Next batch assignments are prepared for #1871, #1873 and #1875, using three gpt-6.1-sol workers after the merge.
currentbits commentedon Nov 7, 2026
Note
🤖 gpt-6.1-sol responding on behalf of currentbits
PR #1917 merged as 2e4b0fd after all 14 checks passed. #1877 and #1898 are complete; #1870 retains its documented upstream reporting follow-up. Starting three GPT-6.1 (
gpt-6.1-sol) workers from fresh main, one isolated branch each: #1871 independent Ultracode controls, #1873 fresh provider context within the same thread, and #1875 opt-in process cleanup on settlement. Lead will review, integrate, test and babysit the next PR.currentbits commentedon Nov 7, 2026
Note
🤖 gpt-6.1-sol responding on behalf of currentbits
Batch three is implemented in PR #1918: #1871 separate Ultracode mode, #1873 fresh provider context in the same thread, and #1875 optional managed process cleanup on settlement. Three GPT-6.1 Sol workers built the changes; cross-review fixed legacy defaults, named-account controls, explicit older-history forks/rewinds, settings inheritance and private worktree ownership after merge cleanup. All three issues remain plan:doing until merge. Renderer suite (3,817 tests), typecheck/preload sync and production build pass; full Electron and GitHub CI are running.
currentbits commentedon Nov 7, 2026
Note
🤖 gpt-6.1-sol responding on behalf of currentbits
PR #1918 is ready at e701e84: all 14 GitHub checks pass, no unresolved review threads, and the branch is current with main. Local validation includes 3,817 renderer tests, 5,851 Electron passes (four skips), 32 Codex provider tests after the usage-ledger fix, typecheck/preload sync and production build. Cross-review and PR review fixes cover active Claude retry callbacks, usage-ledger deltas after fresh context, named accounts and legacy defaults. Awaiting the required merge sign-off; #1871/#1873/#1875 remain doing until merged. Next planned batch: #1874 automation destination/session modes, #1878 PR completions and #1879 editable description checklists, again with three GPT-6.1 Sol workers.
currentbits commentedon Nov 7, 2026
Note
🤖 gpt-6.1-sol responding on behalf of currentbits
PR #1918 squash-merged as ca38be9. #1871, #1873 and #1875 are closed and plan:done. Batch four is now plan:doing: #1874 automation destination/session modes, #1878 PR/MR reference completions, #1879 editable description checklists. Three GPT-6.1 Sol workers are building these in isolated worktrees based on the merged main; the lead will integrate, validate and babysit the batch PR.
currentbits commentedon Nov 7, 2026
Note
🤖 gpt-6.1-sol responding on behalf of currentbits
Batch four is implemented in #1919: fixed automation destinations/session modes (#1874), repository-scoped editor completions (#1878), and safe description task toggles (#1879). Three GPT-6.1 Sol workers built the changes. All four review findings are fixed; local checks pass: 3,842 renderer tests, 5,875 Electron tests (4 skips), typecheck/preload sync, production build and generated site checks. CI is running on the updated head before squash merge.
currentbits commentedon Nov 7, 2026
Note
🤖 gpt-6.1-sol responding on behalf of currentbits
#1919 is ready at db66273: all 14 CI checks pass, all four review findings are resolved, and the branch is clean/current with main. The macOS fixture race was fixed by removing unused shared-state writes from read-only fake-gh calls; all 62 affected backend tests pass. Requesting the host's squash merge sign-off next. The next batch will cover #1872, #1880 and #1888 with three GPT-6.1 Sol workers after this merge.
currentbits commentedon Nov 7, 2026
Note
🤖 gpt-6.1-sol responding on behalf of currentbits
#1919 merged by squash at c0d9943. #1874, #1878 and #1879 are done. Starting the next batch from merged main: CLIProxy account usage (#1872), GitLab request-changes reviews (#1880), and multilingual local dictation (#1888), using three GPT-6.1 Sol workers in isolated worktrees. Each change will receive focused verification, integration review and CI before the next squash merge.
currentbits commentedon Nov 7, 2026
Note
🤖 gpt-6.1-sol responding on behalf of currentbits
The three worker branches are combined in #1923: CLIProxy usage (#1872), GitLab request-changes verdicts (#1880), and multilingual local dictation (#1888). Review fixes include clearing stale draft credentials on destination changes, current v8 API support with bounded legacy fallback, and pinning the original local fallback during cloud recording/upload. All 3,859 renderer tests, typecheck/preload sync and production build pass. Full Electron verification and CI are running before squash merge.
currentbits commentedon Nov 7, 2026
Note
🤖 gpt-6.1-sol responding on behalf of currentbits
#1923 is ready for squash merge at 79ecfbf. The three forks are combined and reviewed; all 14 CI checks pass, including native server bundle boot on Apple Silicon, Intel Mac, Linux x64 and Linux ARM. Local checks passed 3,859 renderer tests and 5,908 Electron tests (4 skips), typecheck/preload sync and production build. No unresolved review threads. The next selected batch is #1889, #1894 and #1900, still queued until this merge.
currentbits commentedon Nov 7, 2026
Note
🤖 gpt-6.1-sol responding on behalf of currentbits
Batch five merged in #1923: optional Claude proxy usage, GitLab request-changes verdicts, and multilingual local dictation. All 14 CI jobs passed, including the four native server bundles. Issues #1872, #1880 and #1888 are plan:done.
Batch six is in progress with three GPT-6.1 Sol workers: #1889 reuse verified speech models from selected folders; #1894 narrower desktop windows with layout checks; #1900 atomic generated provider/MCP configuration. Work is isolated from the main checkout. I will integrate, review, test and babysit the combined PR through the same squash-merge flow.
currentbits commentedon Nov 7, 2026
Note
🤖 gpt-6.1-sol responding on behalf of currentbits
Batch six is implemented in #1924 for #1889, #1894 and #1900. Independent worker reviews and automated review found no material issues on head 3ef2a47. Typecheck/preload parity, production build, 3,862 renderer tests, 247 memory-server tests and all six main/detached Electron layout scenarios passed. Local Electron: 5,927 passed, four skipped, one existing Ultracode version-probe timeout; that test file passed separately. Cross-platform CI remains the merge gate. Issue labels stay plan:doing until merge.
Next batch after merge: #1881 complete GitLab discussion/finding coverage; #1887 show the actual comparison base alongside change totals; #1902 bounded, redacted and actionable failure explanations. Each starts by verifying the existing Solenta path before changing it.
currentbits commentedon Nov 8, 2026
Note
🤖 gpt-6.1-sol responding on behalf of currentbits
#1924 head 58a47c9 fixes three Windows-only test comparisons between equivalent short and expanded canonical paths. The full Windows test job now passes, including the previously transient Cursor fixture setup. The application code is unchanged by those fixes.
The remaining CI gate is currently blocked before tests by repeated HTTP504 downloads in Electron installation and optional ONNX CUDA binary installation. Failed workflows have been rerun with the babysit-pr flow. No application or dependency changes were made to bypass this external failure, and #1889/#1894/#1900 remain plan:doing until merge. A separate suggested task investigates skipping optional CUDA downloads in CPU-only CI.
currentbits commentedon Nov 8, 2026
Note
🤖 gpt-6.1-sol responding on behalf of currentbits
The final infrastructure rerun for #1924 again fails before tests with HTTP504 during Electron installation (examples: jobs128690328876,128690328658,128690327864). The current head remains58a47c94, clean and current with main. Full Windows CI is green; local validation and the product changes are complete, but the cross-platform merge gate is externally blocked. The babysit-pr watcher exited on these failures. #1889/#1894/#1900 stay plan:doing; the next selected batch #1881/#1887/#1902 stays queued until this PR can merge.
currentbits commentedon Nov 8, 2026
Note
🤖 gpt-6.1-sol responding on behalf of currentbits
Next batch is tracked in #1926. Planboard now contains the full new feature list: #1927 model history, #1928 model usage filters, #1929 resizable file columns, #1930 recording privacy, #1931 parent worker approvals, #1932 Claude CLI model discovery, #1933 dictation vocabulary, #1934 silence/cloud chunking, #1935 interactive MCP apps, #1936 file-backed HTML artifacts, #1937 linked-computer HTTPS thread links, #1938 PDF preview, #1939 mixed RTL Markdown, #1940 original screenshot export, and conditional #1941 Cursor account usage. Reproduced reliability fixes are #1942 Git diff paths and #1943 GitLab origins.
The focused usage extension adds #1944 monthly Claude extra-usage spend, #1945 partial/per-model Claude accounting and #1946 interrupted/replayed Codex accounting; #1904/#1905/#1906 received additional recovery, identity and import cases.
Three GPT-6.1 Sol workers are implementing #1927/#1928/#1929 in isolated branches. #1928 is committed with focused checks passing and independent review underway. All unmerged issues remain open. Existing #1924 is preserved and still blocked by failing Linux checks; it is not counted as merged.