Highlights
Claude Fable 5.1
- Added Claude Fable 5.1 as the leading Claude Agent model, with the exact
claude-fable-5-1slug, a one-million-token context window, always-on thinking, Low through Max effort, no fast-mode lane, and no legacy ultrathink prompt mode. - Added Fable 5.1 to Pi's repaired Anthropic catalog so authenticated Pi installations whose upstream model list predates the release can still select it alongside Fable 5 and Opus 4.8.
- Changed the plain
fablealias to select Fable 5.1 while keeping explicit Fable 5 aliases backward compatible. - Added compatible one-million-token Cursor variants and Claude's model-scoped Fable weekly allowance to usage surfaces.
Faster, quieter startup
- Reduced large-database startup work by treating the live shell-stream snapshot as authoritative and querying only when that subscription generation still lacks a snapshot, including reconnect recovery.
- Pruned unrecoverable open-turn journal rows instead of replaying them on every boot, and stopped replay after the first failure.
- Scaled SQLite cache and mmap budgets to physical memory and preloaded the Google Fonts stylesheet so a slow network cannot hold back the app module.
- On the measured 1.7 GB database, shell snapshot requests fell from three to one and warm server boot-to-listen time fell from 1.06 seconds to 0.53 seconds.
Provider and process reliability
- Gave active Devin tool calls a separate one-hour inactivity budget while retaining the ordinary 30-minute idle budget for turns without an active tool.
- Fixed stale Devin events refreshing the active watchdog and normalized Devin's unexpected boolean
get_output.blockresponse without weakening other ACP messages. - Centralized native Windows, POSIX, and WSL process planning and supervised process-tree teardown across providers, Git, updater, voice, terminal, and server execution.
- Added typed provider startup phases and failure reasons, revalidated Windows process identity before escalation, and required proof that owned roots and descendants exited.
- Added a CI architecture boundary check covering Windows execution, WSL routing, teardown, filesystem durability, and provider integration.
UI and workflow polish
- Limited the Git text-generation picker to dedicated one-shot backends—Codex, Cursor, OpenCode, and Factory Droid.
- Tightened composer spacing and aligned picker capsules, project reset highlighting, and folder/reset icon treatment.
- Fixed project and project-task creation reporting success before a real task ID existed, superseded navigation replacing a newer route, and new-task creation before project hydration completed.
Verification
- Formatting passed across 18,823 files.
- Lint passed with 0 errors; 489 existing warnings remain informational.
- Typecheck passed across all 7 packages.
- Release smoke passed across the 1,448-package dependency graph.
- Build passed with all 5 Turbo tasks successful.
- Full tests passed with 890 test files and 10,247 tests; 4 files and 22 tests remain skipped by platform or integration gates.
Full Changelog: v0.8.0...v0.8.1