agent/pi - pi - Penguin

Release v0.67.2

Mario Zechner · 2026-04-14 23:41:17 +02:00

v0.67.2 5a3e702389

docs: audit unreleased changelog entries

Mario Zechner · 2026-04-14 23:40:29 +02:00

be6ac02238

chore: approve contributor lucasmeijer

github-actions[bot] · 2026-04-14 21:24:41 +00:00

73641c0298

fix: align OpenAI cache affinity and use uuidv7 session ids

Mario Zechner · 2026-04-14 23:20:13 +02:00

018b40c30c

chore: replace OSS weekend with permanent contribution gate

Mario Zechner · 2026-04-14 23:16:11 +02:00

d62d22173a

docs(tui): add changelog entry for PR 2989

Mario Zechner · 2026-04-14 21:00:12 +02:00

8f66938c80

fix(tui): resolve rebase typo closes #2979

Mario Zechner · 2026-04-14 20:41:28 +02:00

5ed4600350

fix(tui): complete super key helper support closes #2979

Mario Zechner · 2026-04-14 20:40:34 +02:00

ddb8454c87

Fix Ctrl+Alt key matching in tmux (#2989 )

Kao Félix · 2026-04-14 20:36:08 +02:00

6cf5098f6b

fix(coding-agent): do not inject bunfs script path into subagent prompts (#3002 )

The presence of `process.argv[1]` in the current invocation does not
imply that the current `pi` process is being invoked as a script.
Specifically, in the case of a bun compiled binary, `process.argv[1]`
will be the virtual path `/$bunfs/root/pi`, which is not a real file on
the filesystem.  This causes the subagent extension to inject this path
into the child `pi` invocation, which is interpreted as part of the
prompt, which confuses the subagent.

This change uses the script only if it is not a bunfs virtual script

Co-authored-by: nathyong <nathyong@noreply.github.com>

nathyong · 2026-04-14 20:33:05 +02:00

462b3d21e3

fix(coding-agent): custom models for built-in providers and list-models error surfacing (#3072 )

* fix(coding-agent): surface models.json load errors on stderr in --list-models

When models.json has validation errors, --list-models silently
discarded custom models and overrides without any user-visible
feedback. Now prints the error to stderr via chalk.yellow warning
before listing available models.

* fix(coding-agent): allow custom models for built-in providers in models.json

Built-in providers (openrouter, anthropic, etc.) already have api and
baseUrl on every model, and auth comes from env vars / auth storage.
Relax validation so custom models under built-in providers don't need
redundant baseUrl, apiKey, or api fields. Inherit them from the first
built-in model for that provider.

fixes #2921

---------

Co-authored-by: Mario Zechner <badlogicgames@gmail.com>

Aliou Diallo · 2026-04-14 20:32:18 +02:00

ec4d413fc9

feat(coding-agent): accept inline extension factories in main() (#3099 )

Mat · 2026-04-14 20:26:07 +02:00

dcb1f422c4

fix(coding-agent): use npm view for update checks on non-default registries (#3164 )

Previously, getLatestNpmVersion() hardcoded a fetch to
registry.npmjs.org, which 404s for packages on alternate
registries (private, Verdaccio, GitHub Packages, etc.).
The error was silently swallowed, so users never saw update
notifications for these packages.

Replace the direct fetch with
via the existing runCommandCapture helper. This uses the same
registry resolution as npm install (reads .npmrc config chain),
respects the npmCommand setting, and runs with this.cwd for
correct .npmrc lookup.

Aliou Diallo · 2026-04-14 20:20:14 +02:00

2d3ed0cfa7

feat(coding-agent): support multiple --append-system-prompt flags (#3171 )

Aliou Diallo · 2026-04-14 20:19:11 +02:00

ac37d4d439

chore: approve contributor jay-aye-see-kay

github-actions[bot] · 2026-04-14 18:03:18 +00:00

6722144b92

chore: approve contributor wirjo

github-actions[bot] · 2026-04-14 18:00:06 +00:00

d5261644d3

fix(ai): strip partialJson from responses tool calls

Mutate persisted tool-call blocks in place on function_call completion,
remove partialJson, and emit the same reference on toolcall_end.
Add regression coverage for persisted block cleanup and event identity.

fixes #3078

Armin Ronacher · 2026-04-14 18:56:53 +02:00

e2b40dfc81

fix(coding-agent): stop bordered loader on dispose

fixes #3093

Armin Ronacher · 2026-04-14 16:48:54 +02:00

64a0a74d37

Add [Unreleased] section for next cycle

Mario Zechner · 2026-04-14 00:32:10 +02:00

efc58fed70

Release v0.67.1

Mario Zechner · 2026-04-14 00:31:22 +02:00

v0.67.1 5326452452

fix(coding-agent): correct changelog heading for startup notes

Mario Zechner · 2026-04-14 00:30:29 +02:00

0fa95393c4

Add [Unreleased] section for next cycle

Mario Zechner · 2026-04-14 00:23:18 +02:00

852efe9596

Release v0.67.0

Mario Zechner · 2026-04-14 00:22:27 +02:00

v0.67.0 023cd4afda

feat(coding-agent): add install telemetry ping controls

Mario Zechner · 2026-04-14 00:21:13 +02:00

7371c30c0a

docs: audit unreleased changelog entries

Mario Zechner · 2026-04-13 23:42:31 +02:00

31eaef590e

chore: fix npm audit vulnerabilities

Mario Zechner · 2026-04-13 23:28:38 +02:00

6828603a6b

docs: enable OSS weekend

Mario Zechner · 2026-04-13 22:59:01 +02:00

50ce1b0fb3

fix(ai): support service tier in codex responses (#2996 )

Markus Ylisiurunen · 2026-04-13 11:04:11 +02:00

f829f80885

fix(tui): preserve column of cursor before it entered a paste marker (#3092 )

Sviatoslav Abakumov · 2026-04-13 11:03:18 +02:00

4dda448a3f

fix(coding-agent): flush queued messages after tree navigation (#3091 )

Messages sent during branch summarization are added to the compaction
queue but never flushed when summarization ends. The user sees the
message in the steering queue and has to manually dequeue and resend it.

The fix is to flush the compaction queue after successful tree
navigation, matching the existing behavior of auto-compaction and
`/compact`. If the user jumps to a user message, the queued message is
sent as a prompt, and the jumped-to text is placed in the editor.

Sviatoslav Abakumov · 2026-04-13 11:03:01 +02:00

3166dae728

fix(web-ui): ignore key events during IME composition (#2982 )

* fix(web-ui): ignore key events during IME composition

* fix(web-ui): ignore key events during IME composition in ModelSelector

xu0o0 · 2026-04-13 10:58:44 +02:00

6fefa6fbab

fix(tui): sync test awaits with throttled render pipeline (#3076 )

Tests assumed renders complete synchronously after requestRender() +
nextTick + flush(), but commit 6f5f37f8 changed requestRender() to
schedule via setTimeout with a 16ms minimum interval. Tests reading
viewport state before doRender() runs saw stale data.

Add VirtualTerminal.waitForRender() helper that waits for the throttled
render to settle (nextTick + 20ms + xterm flush), and replace all
terminal.flush() calls after render-triggering actions with it.

Fixes 8 failing tests across overlay-short-content, tui-render, and
markdown test suites.

Aliou Diallo · 2026-04-13 10:57:48 +02:00

41377ee8ec

chore: approve contributor pmateusz

github-actions[bot] · 2026-04-13 08:41:19 +00:00

fb6ccbeaa1

chore: approve contributor Mic92

github-actions[bot] · 2026-04-13 08:38:18 +00:00

ddb1a407bd

chore: approve contributor dyyz1993

github-actions[bot] · 2026-04-11 15:06:38 +00:00

c779c14e91

chore: approve contributor sudosubin

github-actions[bot] · 2026-04-11 15:05:32 +00:00

b3f2c34759

chore: approve contributor bjesuiter

github-actions[bot] · 2026-04-11 14:56:49 +00:00

b98478560a

fix(tui): replace spread-into-push in Container.render() to prevent stack overflow, closes #2651

Mario Zechner · 2026-04-09 03:33:09 +02:00

3b7448d156

fix: bump antigravity User-Agent header version (#2901 )

resolves #2815

Co-authored-by: Mario Zechner <badlogicgames@gmail.com>

Aadish Verma · 2026-04-09 03:10:05 +02:00

4f7fc9de7e

Add all OpenRouter provider routing fields to OpenRouterRouting (#2904 )

The OpenRouterRouting interface only had only and order, but the runtime
passes the entire object as-is to the OpenRouter API's provider field.
This means all other fields worked at runtime but had no TypeScript
type-checking, no autocomplete, and no config validation.

This adds all fields from the OpenRouter provider selection API
(https://openrouter.ai/docs/guides/routing/provider-selection):
allow_fallbacks, require_parameters, data_collection, zdr,
enforce_distillable_text, ignore, quantizations, sort, max_price,
preferred_min_throughput, preferred_max_latency.

Also fixes the models.md documentation which had an invalid fallbacks
field in the example (not part of the OpenRouter API), replaces it with
a comprehensive example showing all available fields, and clarifies that
openRouterRouting is sent as-is in the provider field.

Zeremonienmeister Berber Aab · 2026-04-09 03:08:45 +02:00

f05f4e8a5f

fix: update google provider to handle gemma 4 thinking levels and route between MINIMAL and HIGH (#2903 )

resolves #2812

Co-authored-by: Mario Zechner <badlogicgames@gmail.com>

Aadish Verma · 2026-04-09 03:08:00 +02:00

ee2483cd34

fix(ai): correct thinking budget for 2.5-flash-lite models (#2861 )

The 2.5-flash-lite model was incorrectly matching the 2.5-flash
case, setting minimal budget to 128. Per Google's docs, flash-lite's
minimum thinking budget is 512, not 128. This caused the error:

  The thinking budget 128 is invalid. Please choose a value between
  512 and 24576.

Add a dedicated 2.5-flash-lite case before the 2.5-flash case so
the more specific match takes priority.

Fixes #2838

Co-authored-by: JasonOA888 <JasonOA888@users.noreply.github.com>

Jason · 2026-04-09 03:03:45 +02:00

b48d802930

feat(coding-agent): set PI_CODING_AGENT=true env var at startup

closes #2868

Mario Zechner · 2026-04-09 03:03:19 +02:00

61015830ed

Update sandbox extension configuration instructions (#2915 )

The current sandbox extension instructions state the glob configuration file must exist at `~/.pi/agent/sandbox.json` whereas it actually must exist at  `~/.pi/agent/extensions/sandbox.json` to be picked up.