Commit Graph

772 Commits

Author SHA1 Message Date
YeonGyu-Kim 1deb13230f Merge pull request #2237 from iyoda/refactor/model-resolution-dedup
refactor(shared): deduplicate model resolution utility functions
2026-03-02 23:27:49 +09:00
YeonGyu-Kim af70ec7449 Merge pull request #2198 from acamq/feat/no-auto-commit-work-plan
feat(start-work): add auto_commit config option
2026-03-02 23:27:44 +09:00
YeonGyu-Kim 930ec52fd1 fix(preemptive-compaction): allow re-compaction after context grows and use model-specific limits
compactedSessions permanently blocked re-compaction after first success,
causing unbounded context growth (e.g. 500k on Kimi K2.5 with 256k limit).

- Clear compactedSessions flag on new message.updated so compaction can
  re-trigger when context exceeds threshold again
- Use modelContextLimitsCache for model-specific context limits instead
  of always falling back to 200k for non-Anthropic providers
2026-03-02 23:07:39 +09:00
IYODA Atsushi 51d53f9976 refactor(shared): deduplicate model resolution utility functions
Extract normalizeModel() (3 identical copies) and normalizeModelID()
(2 identical copies) into canonical src/shared/model-normalization.ts.
Delete dead-end duplicate model-name-matcher.ts. Update all consumers.
2026-03-02 16:38:22 +09:00
YeonGyu-Kim 5f0aec81e5 perf(read-image-resizer): decode only first 32KB of base64 for dimension parsing
Previously decoded entire image buffer to read headers. Now slices base64
to 32KB prefix before decoding — sufficient for PNG/GIF/WebP/JPEG headers.
Dramatically reduces memory allocation for large images.
2026-03-02 15:20:31 +09:00
YeonGyu-Kim 7224433790 fix(hooks): prevent SSRF via URL scheme validation and extend disable mechanism to HTTP hooks
- Restrict HTTP hook URLs to http: and https: schemes only (blocks file://, data://, ftp://)
- Extend hook disable config to cover HTTP hooks by matching against hook URL identifier
- Update all 5 hook executors (pre-tool-use, post-tool-use, stop, pre-compact, user-prompt-submit)
- Add 6 new tests for URL scheme validation (file, data, ftp rejection + http, https, invalid URL)
2026-03-02 15:20:31 +09:00
YeonGyu-Kim f1174cd47d Merge pull request #2210 from code-yeongyu/fix/c1-synthetic-part-id
fix(context-injector): use deterministic synthetic part ID for cache stability
2026-02-28 13:40:16 +09:00
YeonGyu-Kim ae5c7a4fa7 Merge pull request #2151 from ualtinok/fix/agent-usage-reminder-subagent-exclusion
[Bug]: agent-usage-reminder hook sends circular reminders to explore/librarian subagents
2026-02-28 13:37:14 +09:00
YeonGyu-Kim 4b69667aa8 feat(hooks): add read-image-resizer hook
Intercepts Read tool output with image attachments and resizes to comply with Anthropic API limits (≤1568px long edge, ≤5MB). Only activates for Anthropic provider sessions and appends resize metadata (original/new resolution, token count) to tool output.

Ultraworked with [Sisyphus](https://github.com/code-yeongyu/oh-my-opencode)

Co-authored-by: Sisyphus <clio-agent@sisyphuslabs.ai>
2026-02-28 13:21:40 +09:00
YeonGyu-Kim b2183d360f style(hooks): add blank line between interpolateEnvVars and resolveHeaders 2026-02-28 12:05:08 +09:00
YeonGyu-Kim 56c12c360a fix(hooks): resolve cubic review issues
- Replace two-pass env interpolation with single-pass combined regex to
  prevent re-interpolation of $-sequences in substituted header values
- Convert HookEntry to discriminated union so type: "http" requires url,
  preventing invalid configs from passing type checking
- Add regression test for double-interpolation edge case
2026-02-28 12:00:02 +09:00
YeonGyu-Kim d960775da3 feat(hooks): add HTTP hook handler support
Add type: "http" hook support matching Claude Code's HTTP hook specification.
HTTP hooks send POST requests with JSON body, support env var interpolation
in headers via allowedEnvVars, and configurable timeout.

New files:
- execute-http-hook.ts: HTTP hook execution with env var interpolation
- dispatch-hook.ts: Unified dispatcher for command and HTTP hooks
- execute-http-hook.test.ts: 14 tests covering all HTTP hook scenarios

Modified files:
- types.ts: Added HookHttp interface, HookAction union type
- config.ts: Updated to accept HookAction in raw hook matchers
- pre-tool-use/post-tool-use/stop/user-prompt-submit/pre-compact:
  Updated all 5 executors to dispatch HTTP hooks via dispatchHook()
- plugin-loader/types.ts: Added "http" to HookEntry type union
2026-02-28 11:38:34 +09:00
acamq d38d98ba02 feat(start-work): add auto_commit config option
Add start_work.auto_commit configuration option to allow users to
disable the automatic commit step in the /start-work workflow.

When auto_commit is false:
- STEP 8: COMMIT ATOMIC UNIT is removed from orchestrator reminder
- STEP 9: PROCEED TO NEXT TASK becomes STEP 8

Resolves #2197
2026-02-27 13:38:52 -07:00
YeonGyu-Kim 50573da6c5 fix(atlas): use start-work session agent for continuation gating
Prefer the in-memory session agent set by /start-work when validating idle continuation eligibility, so stale message storage agent values do not block boulder continuation.
2026-02-27 15:01:44 +09:00
YeonGyu-Kim 5f7bdf6cdd test(ralph-loop): harden completion detector PluginInput mock 2026-02-27 03:12:22 +09:00
YeonGyu-Kim cf3aed301c fix(ralph-loop): persist session message count at loop start 2026-02-27 03:08:30 +09:00
YeonGyu-Kim e034930194 fix(ralph-loop): scope completion detection to messages since loop start 2026-02-27 03:05:14 +09:00
ismeth 05f1e1fedb fix(agent-usage-reminder): skip reminders for non-orchestrator subagents 2026-02-26 17:05:33 +01:00
YeonGyu-Kim 4b946c8883 Merge pull request #1963 from MoerAI/fix/multi-issue-1888-1693-1891
fix: resolve issues #1888, #1693, #1891
2026-02-26 23:13:00 +09:00
YeonGyu-Kim ec38875fd8 Merge pull request #2145 from code-yeongyu/fix/issue-1915-windows-spawn-hide
fix(windows): add windowsHide to Bun.spawn calls to prevent stray terminal windows
2026-02-26 23:12:57 +09:00
YeonGyu-Kim 193ec5cd19 Merge pull request #2146 from code-yeongyu/fix/issue-2065-1968-model-updates
fix(models): update Gemini 3→3.1 Pro and add Kimi K2.5 to writing category
2026-02-26 21:07:23 +09:00
YeonGyu-Kim adf7980527 Merge pull request #2143 from code-yeongyu/fix/issue-2017-stop-continuation-cancel
fix(stop-continuation): wire backgroundManager to cancel running tasks on stop
2026-02-26 21:07:20 +09:00
YeonGyu-Kim bf2c04928c Merge pull request #2140 from code-yeongyu/fix/issue-2025-blocked-todo-continuation
fix(todo-continuation): exclude blocked todos from incomplete count to prevent infinite loops
2026-02-26 21:06:55 +09:00
YeonGyu-Kim 7a88f151df fix(models): update Gemini 3 to 3.1 Pro and add Kimi to writing category fallback
Closes #2065

Closes #1968
2026-02-26 21:01:26 +09:00
YeonGyu-Kim 06580aa548 fix(windows): add windowsHide to Bun.spawn calls to prevent stray terminal windows
Closes #1915
2026-02-26 21:00:40 +09:00
YeonGyu-Kim da74ccc356 fix(stop-continuation): wire backgroundManager to cancel running tasks on stop
Closes #2017
2026-02-26 20:59:35 +09:00
YeonGyu-Kim 5892d0fbb1 fix(ralph-loop): add inFlight guard and improve completion detection to prevent infinite loops
Closes #2084
2026-02-26 20:58:55 +09:00
YeonGyu-Kim 38466ec676 fix(todo-continuation): exclude blocked todos from incomplete count to prevent infinite loops
Closes #2025
2026-02-26 20:58:48 +09:00
YeonGyu-Kim 6de1b3ba7c fix(compaction): add timeout and ensure cleanup to prevent indefinite hangs on rate limit
Closes #2062
2026-02-26 20:58:01 +09:00
YeonGyu-Kim 502bcf76dd Merge pull request #2095 from code-yeongyu/fix/issue-1934-exit-code-130-timeout
fix(run): add event watchdog and secondary timeout to prevent infinite hang in CI
2026-02-26 20:48:46 +09:00
YeonGyu-Kim bc2cc541cf fix(atlas): replace permanent failure lockout with 5-minute backoff
Ultraworked with [Sisyphus](https://github.com/code-yeongyu/oh-my-opencode)

Co-authored-by: Sisyphus <clio-agent@sisyphuslabs.ai>
2026-02-26 20:20:45 +09:00
YeonGyu-Kim 3b5d0dbb47 refactor(think-mode): migrate hook from chat.params to chat.message and remove thinking config injection
Drop provider-specific thinking config injection (THINKING_CONFIGS, getThinkingConfig,
resolveProvider) and instead rely on the provider to handle thinking based on the variant field.
Hook now fires on chat.message using model from input rather than from the message object.

🤖 Generated with assistance of [OhMyOpenCode](https://github.com/code-yeongyu/oh-my-opencode)
2026-02-26 20:08:44 +09:00
YeonGyu-Kim 81022de9db feat(atlas): inject worktree_path into boulder continuation
Ultraworked with [Sisyphus](https://github.com/code-yeongyu/oh-my-opencode)

Co-authored-by: Sisyphus <clio-agent@sisyphuslabs.ai>
2026-02-26 20:08:44 +09:00
YeonGyu-Kim e4eeac81fc feat(start-work): add --worktree flag support in hook
Ultraworked with [Sisyphus](https://github.com/code-yeongyu/oh-my-opencode)

Co-authored-by: Sisyphus <clio-agent@sisyphuslabs.ai>
2026-02-26 20:08:44 +09:00
YeonGyu-Kim a0fbe85c63 feat(start-work): add --worktree flag parsing from user request
Ultraworked with [Sisyphus](https://github.com/code-yeongyu/oh-my-opencode)

Co-authored-by: Sisyphus <clio-agent@sisyphuslabs.ai>
2026-02-26 20:08:44 +09:00
YeonGyu-Kim 8c3644ae75 feat(start-work): add worktree path detection
Ultraworked with [Sisyphus](https://github.com/code-yeongyu/oh-my-opencode)

Co-authored-by: Sisyphus <clio-agent@sisyphuslabs.ai>
2026-02-26 20:08:44 +09:00
YeonGyu-Kim 3a167b84e5 Merge pull request #2111 from code-yeongyu/fix/background-notification-idle-queue
fix(background-agent): queue notifications for idle parent sessions
2026-02-25 16:30:09 +09:00
YeonGyu-Kim b5d92c76a5 fix(background-agent): queue notifications for idle parent sessions
When a background task completes and the parent session is waiting for
user input, promptAsync() fails with an aborted error. Previously the
notification was silently dropped — lost forever.

Fix: queue the notification text in-memory on the BackgroundManager
when promptAsync fails with an aborted/idle error. On the user's next
message to that session, the queued notifications are injected into the
chat context before the agent sees the message.

- BackgroundManager: add pendingNotifications map + queuePendingNotification()
  and injectPendingNotificationsIntoChatMessage() methods
- background-notification hook: add chat.message handler that calls injection
- chat-message.ts: wire backgroundNotificationHook.chat.message into the
  message processing chain
- Add tests covering queue-on-abort and next-message delivery
2026-02-25 16:26:31 +09:00
YeonGyu-Kim 130239f313 Merge pull request #2110 from code-yeongyu/fix/boulder-continuation-agent-check
fix(atlas): boulder continuation deadlock after /start-work + 30s→5s cooldown
2026-02-25 16:22:58 +09:00
YeonGyu-Kim e9b5518dfa test(atlas): update agent check tests to match fixed behavior
- Rename test to 'should inject when last agent is sisyphus and boulder targets atlas
  explicitly' and flip expectation to toHaveBeenCalled() - the old assertion was
  testing the buggy deadlock behavior
- Add 'should not inject when last agent is non-sisyphus and does not match boulder
  agent' to verify hephaestus (unrelated agents) are still correctly skipped
2026-02-25 16:18:59 +09:00
YeonGyu-Kim 624d345bb9 Merge pull request #2108 from code-yeongyu/fix/issue-2100-reset-strategy-race-condition
fix(ralph-loop): fix race condition in --strategy=reset
2026-02-25 16:16:53 +09:00
YeonGyu-Kim e6a3bbfc35 fix(atlas): allow Sisyphus as last agent when boulder targets atlas explicitly
The boulder continuation in event-handler.ts skipped injection whenever
the last agent was 'sisyphus' and the boulder state had agent='atlas'
set explicitly. The allowSisyphusWhenDefaultAtlas guard required
boulderAgentWasNotExplicitlySet=true, but start-work-hook.ts always
calls createBoulderState(..., 'atlas') which sets the agent explicitly.

This created a chicken-and-egg deadlock: boulder continuation needs
atlas to be the last agent, but the continuation itself is what switches
to atlas. With /start-work, the first iteration was always blocked.

Fix: drop the boulderAgentWasNotExplicitlySet constraint so Sisyphus is
always allowed when the boulder targets atlas (whether explicit or default).

Also reduce todo-continuation-enforcer CONTINUATION_COOLDOWN_MS from
30s to 5s to match atlas hook cooldown and recover interruptions faster.
2026-02-25 14:16:17 +09:00
YeonGyu-Kim 5cc11789cd fix(ralph-loop): prevent race condition in reset strategy between session ID update and TUI switch
Fixes #2100
2026-02-25 14:01:27 +09:00
YeonGyu-Kim cddbd502b4 fix(no-hephaestus-non-gpt): add allow_non_gpt_model config opt-out
Fixes #2054
2026-02-25 14:01:26 +09:00
YeonGyu-Kim f69ae16036 docs(agents): update hook counts 44→46, add hashline-edit documentation
- Update root AGENTS.md: hook count 44→46, commit fcb90d92, generated 2026-02-24
- Update src/AGENTS.md: core hooks 35→37, session hooks 21→23
- Update src/hooks/AGENTS.md: 46 hooks total, add modelFallback/noSisyphusGpt/noHephaestusNonGpt/runtimeFallback, jsonErrorRecovery moved to tool-guard (tier 2)
- Create src/tools/hashline-edit/AGENTS.md (93 lines): documents three-op model, LINE#ID format, execution pipeline
- Refresh timestamps: 2026-02-21→2026-02-24 on 28 files
- Update plugin/AGENTS.md hook composition counts

🤖 Generated with assistance of [OhMyOpenCode](https://github.com/code-yeongyu/oh-my-opencode)
2026-02-25 00:02:05 +09:00
YeonGyu-Kim f2107b2e66 fix(hooks): stabilize session notification checks in parallel tests
Use sender-module indirection and an optional main-session filter guard to keep notification assertions deterministic across concurrent test execution.

Ultraworked with [Sisyphus](https://github.com/code-yeongyu/oh-my-opencode)

Co-authored-by: Sisyphus <clio-agent@sisyphuslabs.ai>
2026-02-24 21:43:47 +09:00
YeonGyu-Kim 74b5547b97 fix(hashline-read-enhancer): skip hashifying OpenCode-truncated lines 2026-02-24 18:21:05 +09:00
YeonGyu-Kim 1278b1e59d fix(hashline-read-enhancer): guard against overwriting error output with success message 2026-02-24 17:52:04 +09:00
YeonGyu-Kim 71259916d6 fix(hashline-edit): address Oracle review feedback
- Extract WRITE_SUCCESS_MARKER constant to couple guard and output string
- Remove double blank line after parseLineRefWithHint
- Add comment clarifying normalized equals ref.trim() in error paths
2026-02-24 17:41:30 +09:00
YeonGyu-Kim 5f4e928bc0 fix(hashline-edit): widen non-numeric prefix detection and remove duplicate try-catch
- Replace regex /^([A-Za-z_]+)#.../ with indexOf-based prefix check to catch
  line-ref#VK and line.ref#VK style inputs that were previously giving generic errors
- Extract parseLineRefWithHint helper to eliminate duplicated try-catch in
  validateLineRef and validateLineRefs
- Restore idempotency guard in appendWriteHashlineOutput using new output format
- Add tests for LINE42 extraction, line-ref hint, line.ref hint, and guard behavior

Ultraworked with [Sisyphus](https://github.com/code-yeongyu/oh-my-opencode)

Co-authored-by: Sisyphus <clio-agent@sisyphuslabs.ai>
2026-02-24 17:32:44 +09:00