3682 Commits

This Branch

This Branch

All Branches

• Release v0.68.0
  Mario Zechner · 2026-04-20 23:54:54 +02:00
  v0.68.0 6f345b78d5
• docs(changelog): audit unreleased entries
  Mario Zechner · 2026-04-20 23:53:53 +02:00
  639da8a602
• fix(coding-agent): initialize system prompt options lazily
  Mario Zechner · 2026-04-20 23:22:21 +02:00
  14ad636fb5
• fix(coding-agent): add session_shutdown reasons closes #2863
  Mario Zechner · 2026-04-20 23:11:08 +02:00
  12d7161884
• feat: Expose BuildSystemPromptOptions on before-agent-start event. (#3473)

  Resolves: #3463

  Douglas Stephen · 2026-04-20 23:06:47 +02:00
  1891b9ac01
• chore: remove bjesuiter from approved contributors
  Mario Zechner · 2026-04-20 23:04:52 +02:00
  7c69aadcaf
• 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.

  Aliou Diallo · 2026-04-18 00:45:57 +02:00
  bfa11a50e4
• Add [Unreleased] section for next cycle
  Mario Zechner · 2026-04-17 23:06:48 +02:00
  cbf0d1fbfa
• Release v0.67.67
  Mario Zechner · 2026-04-17 23:06:48 +02:00
  v0.67.67 1c0b628882