You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
Bump version 0.56.0 -> 0.57.0 across pyproject.toml, uv.lock, README, docs, and installer references
Consume Unreleased into a new 0.57.0 (2026-07-05) CHANGELOG section (ghost input-prompt/TUI stability fixes, native /benchmark command suite, benchmark and skill security hardening) and sync into docs/en/release-notes/changelog.md
Add 0.57.0 entry to docs/en/release-notes/breaking-changes.md (no breaking changes)
Copy file name to clipboardExpand all lines: CHANGELOG.md
+23-43Lines changed: 23 additions & 43 deletions
Display the source diff
Display the rich diff
Original file line number
Diff line number
Diff line change
@@ -15,51 +15,31 @@ GitHub Releases page; `0.8.0` is the new starting line.
15
15
16
16
## Unreleased
17
17
18
-
- Show active subagent tool work in the pinned TUI status tail instead of
19
-
leaving long foreground agent runs on the generic composing spinner.
20
-
- Keep the TUI prompt bar visible while an agent turn is starting so the
21
-
empty composer does not disappear during lazy-load frames.
22
-
- Ported the running agent TUI toward Pi's stable diff-rendered scene model so streamed output keeps the input card visible without prompt jumps.
23
-
- Added publishability-focused benchmark comparison planning for multi-model
24
-
runs, activity metrics, exportable reports, and safe online source discovery.
25
-
- Added an experimental Focus TUI mode for active agent turns, keeping the composer pinned, hiding file activity by default, and rendering live output without terminal scrollback jumps.
26
-
- Stream file write/edit activity in a compact live shelf so changed files update in place during agent runs instead of adding noisy terminal rows.
27
-
- Hardened `/benchmark` local fixture runs: task `max_steps` now caps the
28
-
underlying agent turn, and `/benchmark:swe` requires `--trusted-dataset true`
29
-
because trusted local fixture datasets execute verification commands.
30
-
- Hardened benchmark and active-skill security edges: SWE verification commands
31
-
are shape-validated before execution, benchmark discovery requires explicit
32
-
network opt-in, benchmark run IDs avoid clock collisions, runtime overrides
33
-
restore after setup failures, and active-skill deactivation persistence
34
-
failures are surfaced instead of swallowed.
35
-
- Strengthened the bundled `pythinker-core` benchmark suite with edge-case
36
-
fixtures for atomic rollback, iterable de-duplication, explicit falsey
37
-
metadata values, and safe path joins across absolute, sibling-prefix, parent,
38
-
and symlink escapes.
39
-
40
-
- Keep the terminal input composer pinned to the bottom during agent runs with a fullscreen prompt mode to reduce TUI flicker.
41
-
42
-
- Explicitly invoked skills now remain active across later turns through a
43
-
compact reminder, and can be cleared with a named stop request or "normal mode".
18
+
## 0.57.0 (2026-07-05)
44
19
45
20
-**No more ghost/duplicate input prompt while the agent works.** After
46
-
submitting a prompt, the editable input row is no longer fossilized above the
47
-
stream as a second, ghostly prompt. The top border stays visible while the
48
-
pre-attach race frame still collapses before the running-prompt delegate
49
-
exists, preventing prompt chrome from fossilizing above the spinner. Once the
50
-
running frame owns the prompt, the `❯` marker stays visible while only the
51
-
editable buffer is hidden until the first scrollback commit, avoiding the
52
-
collapsed one-line card under the lazy-load spinner; the full editable row
53
-
remains below the live stream so you can still see where to steer.
54
-
55
-
- Added native `/benchmark` slash command for deterministic local Pythinker
56
-
model evaluation with bundled smoke tasks, replayable artifacts, and branded
57
-
markdown reports.
58
-
- Expanded `/benchmark start` to use a richer default core suite, isolate file
59
-
edits through the active toolset workspace override, and exclude generated
60
-
verification caches from changed-file reports.
61
-
- Added `/benchmark:swe` for native SWE-style JSONL benchmark tasks that run
62
-
through Pythinker's existing model, tool, verification, and artifact path.
21
+
submitting a prompt, the input card no longer fossilizes above the stream as
22
+
a second, ghostly prompt. The prompt bar and top border now stay visible
23
+
through turn-start and lazy-load frames, and running-agent output renders
24
+
through a stable diff-rendered scene (ported from Pi) so the input card no
25
+
longer jumps. An experimental Focus TUI mode (`tui.focus_mode`) pins the
26
+
composer and hides file activity behind a compact live shelf for a cleaner
27
+
fullscreen view, and active subagent tool work now surfaces in the pinned
28
+
status tail instead of a generic composing spinner.
29
+
-**Added a native `/benchmark` command suite.** Run deterministic local
30
+
Pythinker model evaluations with bundled smoke tasks and SWE-style JSONL
31
+
tasks (`/benchmark:swe`), plus compare, export, and source-discovery
32
+
subcommands that produce replayable artifacts, activity metrics, and
33
+
branded markdown reports.
34
+
-**Hardened benchmark and skill security edges.** SWE verification commands
35
+
are shape-validated before execution, benchmark discovery requires explicit
36
+
network opt-in, run IDs avoid clock collisions, runtime overrides restore
37
+
after setup failures, and active-skill deactivation failures now surface
38
+
instead of being swallowed. Explicitly invoked skills stay active across
39
+
later turns through a compact reminder until stopped by name or
40
+
"normal mode".
41
+
42
+
Upgrade with `pythinker update`, `pip install --upgrade pythinker-code==0.57.0`, or use the native installer for your platform from the [Releases page](https://github.com/Pythoughts-labs/pythinker-code/releases/latest).
Copy file name to clipboardExpand all lines: README.md
+22-22Lines changed: 22 additions & 22 deletions
Display the source diff
Display the rich diff
Original file line number
Diff line number
Diff line change
@@ -50,13 +50,13 @@ It speaks the [**Agent Client Protocol (ACP)**](https://github.com/agentclientpr
50
50
51
51
---
52
52
53
-
## 🆕 What's New in 0.56.0
53
+
## 🆕 What's New in 0.57.0
54
54
55
-
-**Windows shell UI recovers from mid-session console blanking.**Child processes spawned by the Shell tool, background tasks, and `!` commands no longer attach to the interactive console, and the prompt renderer forces an absolute repaint after terminal resizes and failed scrollback handoffs instead of diffing against a stale frame.
56
-
-**Workflow progress rendering and lifecycle fixes.**Progress display no longer duplicates agents truncated by the per-phase display cap, leaked `agent()` coroutines are closed when `parallel()` rejects its arguments, and a 1000-agent lifetime backstop now guards against runaway workflow loops.
57
-
-**Stale update-success notice cleared after downgrading.**Restarting into an older Homebrew install no longer leaves a permanent "Restart to apply" banner.
55
+
-**No more ghost/duplicate input prompt while the agent works.**The input card no longer fossilizes above the stream as a second, ghostly prompt; the prompt bar and top border stay visible through turn-start and lazy-load frames; running-agent output renders through a stable diff-rendered scene (ported from Pi); and an experimental Focus TUI mode (`tui.focus_mode`) pins the composer and hides file activity behind a compact live shelf.
56
+
-**Added a native `/benchmark` command suite.**Run deterministic local Pythinker model evaluations with bundled smoke tasks and SWE-style JSONL tasks (`/benchmark:swe`), plus compare, export, and source-discovery subcommands with replayable artifacts and branded markdown reports.
57
+
-**Hardened benchmark and skill security edges.**SWE verification commands are shape-validated before execution, benchmark discovery requires explicit network opt-in, and active-skill deactivation failures now surface instead of being swallowed.
58
58
59
-
Upgrade with `pythinker update`, `pip install --upgrade pythinker-code==0.56.0`, or use the native installer for your platform from the [Releases page](https://github.com/Pythoughts-labs/pythinker-code/releases/latest).
59
+
Upgrade with `pythinker update`, `pip install --upgrade pythinker-code==0.57.0`, or use the native installer for your platform from the [Releases page](https://github.com/Pythoughts-labs/pythinker-code/releases/latest).
60
60
61
61
62
62
---
@@ -162,7 +162,7 @@ matches your OS — no Python, Node, or `uv` prerequisite.
Copy file name to clipboardExpand all lines: docs/en/guides/getting-started.md
+1-1Lines changed: 1 addition & 1 deletion
Display the source diff
Display the rich diff
Original file line number
Diff line number
Diff line change
@@ -44,7 +44,7 @@ On Windows, run the PowerShell bootstrap. It downloads the native installer, ver
44
44
irm https://pythinker.com/install.ps1 | iex
45
45
```
46
46
47
-
You can also download `PythinkerSetup-0.56.0.exe` manually from the [latest release](https://github.com/Pythoughts-labs/pythinker-code/releases/latest).
47
+
You can also download `PythinkerSetup-0.57.0.exe` manually from the [latest release](https://github.com/Pythoughts-labs/pythinker-code/releases/latest).
Copy file name to clipboardExpand all lines: docs/en/release-notes/changelog.md
+23-32Lines changed: 23 additions & 32 deletions
Display the source diff
Display the rich diff
Original file line number
Diff line number
Diff line change
@@ -17,40 +17,31 @@ GitHub Releases page; `0.8.0` is the new starting line.
17
17
18
18
## Unreleased
19
19
20
-
- Added publishability-focused benchmark comparison planning for multi-model
21
-
runs, activity metrics, exportable reports, and safe online source discovery.
22
-
- Stream file write/edit activity in a compact live shelf so changed files update in place during agent runs instead of adding noisy terminal rows.
23
-
- Hardened `/benchmark` local fixture runs: task `max_steps` now caps the
24
-
underlying agent turn, and `/benchmark:swe` requires `--trusted-dataset true`
25
-
because trusted local fixture datasets execute verification commands.
26
-
- Strengthened the bundled `pythinker-core` benchmark suite with edge-case
27
-
fixtures for atomic rollback, iterable de-duplication, explicit falsey
28
-
metadata values, and safe path joins across absolute, sibling-prefix, parent,
29
-
and symlink escapes.
30
-
31
-
- Keep the terminal input composer pinned to the bottom during agent runs with a fullscreen prompt mode to reduce TUI flicker.
32
-
33
-
- Explicitly invoked skills now remain active across later turns through a
34
-
compact reminder, and can be cleared with a named stop request or "normal mode".
20
+
## 0.57.0 (2026-07-05)
35
21
36
22
-**No more ghost/duplicate input prompt while the agent works.** After
37
-
submitting a prompt, the editable input row is no longer fossilized above the
38
-
stream as a second, ghostly prompt. The top border stays visible while the
39
-
pre-attach race frame still collapses before the running-prompt delegate
40
-
exists, preventing prompt chrome from fossilizing above the spinner. Once the
41
-
running frame owns the prompt, the `❯` marker stays visible while only the
42
-
editable buffer is hidden until the first scrollback commit, avoiding the
43
-
collapsed one-line card under the lazy-load spinner; the full editable row
44
-
remains below the live stream so you can still see where to steer.
45
-
46
-
- Added native `/benchmark` slash command for deterministic local Pythinker
47
-
model evaluation with bundled smoke tasks, replayable artifacts, and branded
48
-
markdown reports.
49
-
- Expanded `/benchmark start` to use a richer default core suite, isolate file
50
-
edits through the active toolset workspace override, and exclude generated
51
-
verification caches from changed-file reports.
52
-
- Added `/benchmark:swe` for native SWE-style JSONL benchmark tasks that run
53
-
through Pythinker's existing model, tool, verification, and artifact path.
23
+
submitting a prompt, the input card no longer fossilizes above the stream as
24
+
a second, ghostly prompt. The prompt bar and top border now stay visible
25
+
through turn-start and lazy-load frames, and running-agent output renders
26
+
through a stable diff-rendered scene (ported from Pi) so the input card no
27
+
longer jumps. An experimental Focus TUI mode (`tui.focus_mode`) pins the
28
+
composer and hides file activity behind a compact live shelf for a cleaner
29
+
fullscreen view, and active subagent tool work now surfaces in the pinned
30
+
status tail instead of a generic composing spinner.
31
+
-**Added a native `/benchmark` command suite.** Run deterministic local
32
+
Pythinker model evaluations with bundled smoke tasks and SWE-style JSONL
33
+
tasks (`/benchmark:swe`), plus compare, export, and source-discovery
34
+
subcommands that produce replayable artifacts, activity metrics, and
35
+
branded markdown reports.
36
+
-**Hardened benchmark and skill security edges.** SWE verification commands
37
+
are shape-validated before execution, benchmark discovery requires explicit
38
+
network opt-in, run IDs avoid clock collisions, runtime overrides restore
39
+
after setup failures, and active-skill deactivation failures now surface
40
+
instead of being swallowed. Explicitly invoked skills stay active across
41
+
later turns through a compact reminder until stopped by name or
42
+
"normal mode".
43
+
44
+
Upgrade with `pythinker update`, `pip install --upgrade pythinker-code==0.57.0`, or use the native installer for your platform from the [Releases page](https://github.com/Pythoughts-labs/pythinker-code/releases/latest).
0 commit comments