agent/pi - pi - Penguin

feat(coding-agent): batch package updates

closes #2980

Mario Zechner · 2026-04-20 22:56:26 +02:00

e4f7fe7bb7

fix(tui): avoid cwd-path matches in @ autocomplete

Only use fd --full-path for path queries so plain fuzzy @ autocomplete results no longer depend on whether the cwd path contains the query text.

closes #2778

Mario Zechner · 2026-04-20 22:44:34 +02:00

968430f63e

docs(agent): clarify parallel tool ordering closes #3468

Mario Zechner · 2026-04-20 22:35:43 +02:00

7b1599058b

fix(coding-agent): resolve shell config from session settings

Stop shell resolution from consulting ambient process.cwd() during bash execution and use the active session's shellPath setting instead.

closes #3452

Mario Zechner · 2026-04-20 22:32:11 +02:00

c40efa9bab

fix(coding-agent): remove process-cwd tool singletons and use tool-name allowlists

- switch SDK/CLI tool selection to name-based allowlists
- apply allowlists across built-in, extension, and SDK tools
- remove ambient process.cwd defaults from core tooling and resource helpers
- update tests, examples, and TUI callers for explicit cwd plumbing
- add regression coverage for extension tool filtering

closes #3452
closes #2835

Mario Zechner · 2026-04-20 22:05:28 +02:00

5a4e22ea44

fix(coding-agent): use tool-name allowlists and remove cwd-bound singletons

- treat tools as a global allowlist across built-in, extension, and SDK tools
- remove process-cwd singleton tool usage from SDK and CLI paths
- add regression coverage for extension tool filtering

closes #3452
closes #2835

Mario Zechner · 2026-04-20 22:05:28 +02:00

27c1544839

docs(coding-agent): document async extension factories closes #3469

Mario Zechner · 2026-04-20 22:02:03 +02:00

ed89480f20

chore: approve contributor dljsjr

github-actions[bot] · 2026-04-20 19:23:24 +00:00

029a6e54be

fix(ai): preserve requiresThinkingAsText replay semantics closes #3387

Mario Zechner · 2026-04-20 19:44:38 +02:00

1d488626d9

fix(coding-agent): skip redundant npm package updates

closes #3000

Mario Zechner · 2026-04-20 19:12:52 +02:00

b73212616d

fix(ai,coding-agent): support anthropic-style cache control for openai compatibles closes #3392

Mario Zechner · 2026-04-20 17:12:05 +02:00

3054fd7a3b

fix(ai): add completions session affinity compat closes #3430

Mario Zechner · 2026-04-20 16:55:16 +02:00

4b2caf43a8

fix(ai,coding-agent): preserve non-vision image placeholders closes #3429

Mario Zechner · 2026-04-20 16:52:27 +02:00

2f4f283cc2

feat(coding-agent): add configurable working indicator closes #3413

Mario Zechner · 2026-04-20 16:28:39 +02:00

74139c3f66

fix(coding-agent): handle symlinked session selector paths closes #3364

Mario Zechner · 2026-04-20 16:25:12 +02:00

61579214c8

fix(coding-agent): clarify session ID docs closes #3390

Mario Zechner · 2026-04-20 16:12:32 +02:00

860d67518c

fix(coding-agent): detect Windows pnpm global installs closes #3378

Mario Zechner · 2026-04-20 15:59:57 +02:00

d56e5aea0b

fix(ai): add direct OpenAI completions prompt caching closes #3426

Mario Zechner · 2026-04-20 15:37:29 +02:00

aa1b587b10

fix(tui): restore shifted xterm input

closes #3436

Mario Zechner · 2026-04-20 15:36:10 +02:00

6b55d6859d

fix(coding-agent): reuse session thinking for compaction closes #3438

Mario Zechner · 2026-04-20 15:09:50 +02:00

da6a81d398

fix(coding-agent): add missing typebox dependency

closes #3434

Mario Zechner · 2026-04-20 15:08:34 +02:00

d66ef6dce8

fix(coding-agent): preserve share page tool output indentation closes #3440

Mario Zechner · 2026-04-20 14:46:37 +02:00

5ea8a276eb

feat(coding-agent): make scoped models and tree filter shortcuts configurable (#3343 )

- Add app.models.{save,enableAll,clearAll,toggleProvider,reorderUp,reorderDown}
- Add app.tree.filter.{default,noTools,userOnly,labeledOnly,all,cycleForward,cycleBackward}
- Migrate scoped-models-selector cancel to tui.select.cancel
- Prefer reserved actions on key collision in extension shortcut conflict check

Marek Pazik · 2026-04-20 14:35:30 +02:00

d4e2e563ae

fix(coding-agent): split /clone from /fork UX

closes #2962

Mario Zechner · 2026-04-20 14:33:32 +02:00

d554409b1f

fix(coding-agent): add OpenRouter attribution headers

closes #3414

Mario Zechner · 2026-04-20 14:14:36 +02:00

62c1c4031c

feat(coding-agent): add fork position and duplicate session option (#3431 )

Armin Ronacher · 2026-04-20 14:13:34 +02:00

23569e304b

feat(ai): allow overriding oauth callback bind host closes #3396 (#3409 )

Michael · 2026-04-20 13:43:08 +02:00

ebc60aa825

fix(coding-agent): avoid duplicate symlinked skills in pi config (#3417 )

- dedupe resolved skill entries by canonical path (realpathSync) with raw-path fallback matching loadSkills()

- preserve precedence by applying skill dedupe after resource precedence sorting (first winner retained)

- add tests to verify expected behavior

- closes #3405

Ramiz Wachtler · 2026-04-20 13:23:00 +02:00

2b45243817

fix(ai): replace stale OpenRouter Llama 4 Maverick test model with Llama 4 Scout (#3421 )

* fix(ai): replace stale OpenRouter Llama 4 Maverick test model

- Switch OpenRouter Meta test model from `meta-llama/llama-4-maverick` to `meta-llama/llama-4-scout` in AI tests.

- CI failure body:
  - packages/ai/test/context-overflow.test.ts(531,41): error TS2345: Argument of type "meta-llama/llama-4-maverick" is not assignable to parameter of type "...".
  - packages/ai/test/total-tokens.test.ts(520,40): error TS2345: Argument of type "meta-llama/llama-4-maverick" is not assignable to parameter of type "...".

* chore(ai): add changelog entry for OpenRouter test model swap

- Add Unreleased Fixed entry for switching OpenRouter Meta test model to llama-4-scout

- Reference PR #3421 attribution

Ramiz Wachtler · 2026-04-20 13:21:29 +02:00

cbe0e8304f

chore: approve contributor cmraible

github-actions[bot] · 2026-04-19 20:40:54 +00:00

877279f518

fix(ai): strip JSON Schema meta keys for Cloud Code Assist (#3412 )

Vladyslav Tkachenko · 2026-04-19 16:54:49 +02:00

f732f5e858

fix(coding-agent): use browser-safe share viewer shortcuts (#3374 )

fixes #3358

Andrea Baccega · 2026-04-19 08:49:12 +02:00

9b28e185db

feat(amazon-bedrock): conditionally omit maxTokens from inferenceConfig (#3400 )

When model.maxTokens is 0 (unset/unknown), omit maxTokens from the
Bedrock ConverseStream inferenceConfig instead of sending 0.

Bedrock's inferenceConfig.maxTokens is optional — when omitted, the
model uses its own internal default. This is optimal for Bedrock's
token quota management:

- At request start, Bedrock reserves input_tokens + max_tokens from
  your TPM quota
- For Claude 3.7+ models, output tokens have a 5x burndown rate
- Sending an unnecessarily high maxTokens wastes TPM capacity during
  the reservation window, reducing concurrent request throughput
- Omitting it lets Bedrock use the model default (~4096 for Claude),
  matching expected output sizes and maximizing quota utilization

Also applies the same conditional pattern to temperature — only
include it in inferenceConfig when explicitly set.

Changes:
- simple-options.ts: Use ?? instead of || so explicit 0 is preserved;
  return undefined when model.maxTokens is 0 (unset)
- amazon-bedrock.ts: Spread maxTokens and temperature into
  inferenceConfig only when defined

Fixes #3399

wirjo · 2026-04-19 08:48:43 +02:00

a5fac1ef09

fix(amazon-bedrock): pass model.baseUrl as endpoint to BedrockRuntimeClient (#3402 )

BedrockRuntimeClient was constructed without the endpoint option,
causing model.baseUrl to be silently ignored. This meant custom
Bedrock endpoints (VPC endpoints, proxy setups, custom routing)
were never used — requests always went to the default regional
endpoint.

Pass model.baseUrl as the endpoint config when set. When not set
(the default), the SDK falls back to the standard regional endpoint
constructed from the region config, preserving existing behavior.

Fixes openclaw/openclaw#47899

wirjo · 2026-04-19 08:47:10 +02:00

5a889ef58b

chore: approve contributor Michaelliv

github-actions[bot] · 2026-04-19 06:41:37 +00:00

c6a493c50b

fix(coding-agent): coerce stringified JSON edits in edit tool (#3370 )

Some models (Opus 4.6, GLM-5.1) send the edits parameter as a JSON
string instead of a parsed array. This fails AJV validation with
'must be array' and models fall back to sed/python.

Parse stringified edits in prepareEditArguments before validation.

Danila Poyarkov · 2026-04-18 18:29:23 +02:00

a2ec01e12f

fix manifest glob expansion and add regression test (#3350 )

Neon · 2026-04-18 13:16:39 +02:00

182d4ceea3

Add [Unreleased] section for next cycle

Mario Zechner · 2026-04-18 01:32:50 +02:00

00887125c0

Release v0.67.68

Mario Zechner · 2026-04-18 01:31:38 +02:00

v0.67.68 a1edb8a463

fix(ai): use SDK token auth and omit Bedrock display in GovCloud closes #3359

Mario Zechner · 2026-04-18 01:24:13 +02:00

454b9619c7

fix(ai): sanitize Mistral tool schemas closes #3361

Mario Zechner · 2026-04-18 00:54:46 +02:00

2dddc5ba25

feat(agent,coding-agent): per-tool executionMode override for sequential tool execution (#3345 )

* feat(agent,coding-agent): add per-tool executionMode field to AgentTool and ToolDefinition

Add optional executionMode?: ToolExecutionMode to AgentTool and
ToolDefinition interfaces. Propagate through wrapToolDefinition and
createToolDefinitionFromAgentTool. No behavioral change yet — agent
loop will read this field in a follow-up.

* feat(agent): support per-tool executionMode override for sequential execution

When a tool defines executionMode='sequential', the agent loop
forces sequential execution of all tool calls in that batch,
even if the global config is parallel.

* feat(coding-agent): re-export ToolExecutionMode from @mariozechner/pi-agent-core

Makes the type available to extensions that want to set
executionMode on tool definitions.

* feat(coding-agent): add tic-tac-toe extension example with executionMode: sequential

Demonstrates per-tool executionMode: the agent plays via move/play
tool calls that share a cursor. Without sequential execution, play
can resolve before earlier moves finish, landing on the wrong cell.