github/claude-code-best
1
0
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity
459 Commits 25 Branches 86 Tags 330 MiB
f1c7f7dc74
Commit Graph

266 Commits

Author SHA1 Message Date
claude-code-best
241aca1e0f feat: 添加 GitHub 集成命令（issue、share、autofix-pr） 
- /issue: 通过 gh CLI 创建 GitHub issue，支持标签/指派
- /share: 会话日志分享到 GitHub Gist，支持密钥脱敏
- /autofix-pr: 自动修复 CI 失败的 PR，进度追踪
- launchCommand: 共享命令启动器

Co-Authored-By: glm-5-turbo <zai-org@claude-code-best.win>
 2026-06-04 09:56:20 +08:00
James Feng
64ca325020 feat: add local analytics sink for self-analysis 
- localSink.ts: writes every logEvent to ~/.claude/local_analytics.jsonl
- Modified index.ts: logEvent() now writes locally in parallel with upstream sinks
- analyze_analytics.py: Python analysis script — event stats, tool rankings,
  time distribution, security events, RL preference signals

No upstream data flow is affected. Local writes are non-blocking append.
Testing: one -p 'hello' captured 32 events across 19 event types.
 2026-06-04 01:15:36 +08:00
James Feng
0d30c6b064 fix: move ToolSearchTool from packages/builtin-tools/ to src/tools/ 
Eliminates cross-workspace circular dependency that triggered
Bun ARM code-splitting deadlock. The tool's 471-line RL data
collection infrastructure (analytics, GrowthBook, scoring weights,
tool_reference API) is preserved intact.

Changes:
- Move ToolSearchTool.ts, prompt.ts, constants.ts to src/tools/ToolSearchTool/
- Update 11 consumer imports from @claude-code-best/builtin-tools/ to relative paths
- Delete stubs and empty shells (MonitorTool.tsx, WorkflowPermissionRequest.tsx)
- Verify: splitting build succeeds, runtime -p mode responds without deadlock
 2026-06-04 01:15:36 +08:00
James Feng
70bc47eadc fix(security): Phase 4 — fix remaining reachable CodeQL alerts 
Command injection (real fix):
- which.ts: switch to array-args execa, remove shell:true
- execFileNoThrowPortable/execSyncWrapper/imagePaste/execFileNoThrow: security comments

Log injection:
- handlers/mcp.tsx: security comments (secrets already redacted)

ReDoS:
- debugFilter.ts: split regex, add input length guard

Sanitization bypass:
- stripHtml.ts: loop-based script/style removal
- claudemd.ts: loop-based HTML comment stripping
- sedEditParser.ts: single-pass char scan replaces chained replaces
- bingAdapter.ts: URL.hostname comparison instead of string includes

Tests: 3068 pass, 0 fail
 2026-06-03 19:51:17 +08:00
James Feng
a249bbe9a6 fix(security): extract shared stripHtmlToText() utility (fixes #18-24) 
- Create src/utils/stripHtml.ts with he-based HTML-to-text conversion
- Replace bingAdapter's inline regex+decodeHtmlEntities with stripHtmlToText()
- Replace WebBrowserTool's inline regex chain with stripHtmlToText()
- Add stripHtml.test.ts with 6 test cases
- Update bingAdapter test expectation for whitespace normalization

Test: 3068 pass, 0 fail
 2026-06-03 17:05:49 +08:00
James Feng
368dd99d01 fix(security): prevent shell injection in headersHelper (#36) 
- Parse headersHelper command with shell-quote to reject operators
- Call execFileNoThrowWithCwd(cmd, args) without shell: true
- Remove shell option from ExecFileWithCwdOptions type entirely
- Add headersHelper.test.ts with injection rejection test
- Fix existing MCP test mocks for compatibility

Test: MCP tests 91/0, full suite 3058 pass (0 new failures)
 2026-06-03 16:59:16 +08:00
James Feng
f8c3354c75 fix(security): replace URL substring checks with proper URL parsing 
- schemas.ts: add isOfficialGitHubOrgUrl() using new URL() for exact
  hostname validation, preventing evilgithub.com bypass (fixes #41-43)
- install-github-app.tsx: add parseGitHubRepoUrl() with URL + SSH parsing,
  replace includes('github.com') substring check

Test: 3060 pass, 0 fail (4 getLanIPs intermittent, not related)
 2026-06-03 16:50:31 +08:00
James Feng
bdfff8de97 Phase 2: delete src/tools/ — align with upstream monorepo 
Deleted all 50 tool directories from src/tools/ (28,508 lines).
All tools already exist in packages/builtin-tools/src/tools/ (upstream authoritative).
No import breakage — src/tools.ts already uses @claude-code-best/builtin-tools paths.

Test baseline:
  src/:  3055 pass, 5 fail, 4 skip
  pkg/:  1133 pass, 5 fail, 4 skip
  5 failures are pre-existing regressions (BashTool /dev/tcp, AgentTool prompt terms)
 2026-06-03 16:25:10 +08:00
James Feng
638affff3f Phase 2b Batch 3: merge AgentTool(38) drift files 
- Merged 38 files from src/tools/AgentTool/ into packages/builtin-tools/
- 4 prompt.ts tests fail due to expected terminology changes (fork vs non-fork)
- Test: 3257 pass / 11 fail (4 transient + 1 BashTool known + 4 prompt + 2 baseline)
- Build: succeeds
 2026-06-03 15:47:56 +08:00
James Feng
46b2b0bfc3 Phase 2b Batch 2: merge BashTool(23) + FileEditTool(13) + FileWriteTool(5) drift files 
- Merged 41 files from src/tools/ into packages/builtin-tools/src/tools/
- All src/tools originals deleted
- 1 known regression: bashCommandIsSafe_DEPRECATED no longer blocks /dev/tcp as argument
- Test: 3288 pass / 5 fail (1 new BashTool security check regression)
- Build: succeeds
 2026-06-03 15:44:00 +08:00
James Feng
0c01c93e80 Phase 2a: mechanical dedup — import rewrite to canonical package 
- Migrated src/tools/-only files into packages/builtin-tools/src/tools/
- Rewrote all src/tools/ imports to @claude-code-best/builtin-tools/tools/
- Deleted 121 identical duplicate files from src/tools/
- 238 drifted files preserved in src/tools/ for Phase 2b manual merge
- 359 files changed: 914 insertions, 4042 deletions

Build: 568 files bundled ✓
Test: 3340 pass / 4 fail (baseline) ✓
 2026-06-03 13:54:22 +08:00
James Feng
bad81fdfcc Phase 1: security do-now — redaction helpers + RCS defaults hardened 
- Add src/utils/sensitive.ts: redactUrl/redactValue/redactForLog
- Apply redaction to MCP config printing (mcp.tsx)
- Apply redaction to hard-fail logging (log.ts)
- Apply redaction to chrome native host debug log
- Apply redaction to CLI error output (exit.ts)
- RCS: default bind to 127.0.0.1 instead of 0.0.0.0
- RCS: CORS restricted to localhost + baseUrl whitelist
- pipeTransport: default bind to 127.0.0.1 via PIPE_HOST env
 2026-06-03 13:43:38 +08:00
James Feng
cdd62520eb feat(upstream): sync b1c4f40f ACP fix + WorkflowTool top-level require() 
Codex cherry-pick:
- b1c4f40f: fix ACP extended thinking + tool_use 连续 user 消息导致 400
  → src/utils/messages.ts: ensureToolResultPairing 检查连续 user 消息
  → src/utils/tests/messages.test.ts: 新增测试（冲突手动解决）

James 兜底:
- tools.ts: WorkflowTool lazy getter → top-level require (MonitorTool 模式)
  移除 initBundledWorkflows() 调用（经确认是 no-op）
  原因: lazy getter 在 splitting:true 下运行时触发 chunk load 死锁

审核结论:
- 上游 a91653a0/6dd378bf/efc218d8 已存在（之前已合入）
- 上游 9d17597e autofix-pr 回流跳过（大规模改动，stub 无基础）
- WORKFLOW_SCRIPTS 仍禁用: top-level require 在 splitting:true 下
  触发 WorkflowTool chunk 循环死锁（非加载模式问题，是 chunk 依赖图过深）

当前 splitting:true 启用状态:
  MONITOR_TOOL ✅ | WORKFLOW_SCRIPTS ❌ | 其他 features ✅

验证: --version / -p / --print / 源码模式 全部通过
 2026-06-02 14:45:21 +08:00
claude-code-best
b1d322f7cc fix: ACP 模式下 extended thinking + tool_use 触发连续 user 消息导致 400 (CC-1215) 2026-06-02 14:39:19 +08:00
James Feng
21f126b76a feat(build): splitting:true + MONITOR_TOOL 恢复 + WorkflowTool lazy getter 
Codex 完成:
- build.ts: splitting: false → true (562 files code-split 构建)
- defines.ts: MONITOR_TOOL 重新启用（require() 迁移后无死锁）
- 验证 splitting:true 产物 -p/--print 全部正常

James 兜底:
- tools.ts: WorkflowTool IIFE → lazy getter (getWorkflowTool)
  避免模块初始化时的 chunk 加载死锁，但 WORKFLOW_SCRIPTS 仍因
  更深的 chunk 依赖图问题 hang，暂时保持禁用
- getAllBaseTools() 中 WorkflowTool 引用改为 lazy getter

状态:
| splitting:true | 562 files ✅ |
| MONITOR_TOOL   | enabled ✅   |
| WORKFLOW_SCRIPTS | disabled (lazy getter 不足) |

验证: --version / -p / --print / 源码模式 全部通过
 2026-06-02 14:22:42 +08:00
James Feng
98bb25636b refactor(tools): static import → require() 迁移 + 兜底修复 
Codex 完成:
- tools.ts: 7 个 static import (指向空壳 stub) → require("@claude-code-best/builtin-tools/...")
  涉及: SleepTool, CronTools, RemoteTriggerTool, MonitorTool, SendUserFileTool,
  PushNotificationTool, TeamCreateTool, TeamDeleteTool, SendMessageTool, WorkflowTool
- PermissionRequest.tsx: MonitorTool require() 指向 builtin-tools
- TeamCreateTool/DeleteTool/SendMessageTool 用 lazy require() 打破循环依赖

兜底修复:
- defines.ts: 保持 MONITOR_TOOL + WORKFLOW_SCRIPTS 禁用
  原因: builtin-tools 里的真工具 import "src/..." 产生跨包循环引用
  在 splitting:true 下触发 chunk 间死锁，在 splitting:false 下触发 ESM 初始化死锁
  这是逆向工程的结构性限制，非简单修复能解决

验证: --version / -p / --print / 源码模式 全部通过
 2026-06-02 13:41:47 +08:00
James Feng
0e9d6955f0 fix(build): 修复构建产物 hang 死 + 空 tool name (4 stub NAME="") 
Root cause: Bun code splitting + MONITOR_TOOL/WORKFLOW_SCRIPTS stub 触发循环 ESM 死锁;
4 个 prompt.ts stub 的 NAME 常量为空字符串导致 API 400 error.

Fix:
- build.ts: splitting: false
- defines.ts: 禁用 MONITOR_TOOL, WORKFLOW_SCRIPTS
- 4 个 prompt.ts: 补全正确 tool name

验证: --version / -p / --print 三种模式全部通过
 2026-06-02 12:55:30 +08:00
James Feng
27682acf63 feat(bridge): tsc 全绿 — 补齐 Bridge/UDS/TCP/ACP 远程控制链类型 
tsc --noEmit: 26 → 0 ✅
bun run build: 562 files ✅
bun test: 3338 pass (0 regression)

变更 (9 files, +128/-12):
- udsClient.ts: 补 sendToUdsSocket, listAllLiveSessions, LiveSession 类型, PeerInfo 扩展字段
- peerAddress.ts: scheme 联合加 'tcp', 补 parseTcpTarget stub
- peerSessions.ts: 补 listBridgePeers() stub
- channelNotification.ts: 补 ChannelPermissionRequestNotificationSchema + channel_context
- acp-link/command.ts: @stricli/core 类型断言 (as any + optional)
- acp/agent.ts: 完整 AcpAgent stub (implement Agent interface)
- computer-use-swift/types.ts: ScreenshotAPI 加 captureWindowTarget?
- ListPeersTool: bridge peer address 映射
- disconnect-monitor.test.ts: status 字段 as any

Bridge 远程控制链 (Daemon→UDS/TCP→Bridge→ACP) 类型层已完整就绪。
 2026-06-02 11:26:11 +08:00
James Feng
0174a8edca chore(typecheck): tsc 基线 62→26，补 @agentclientprotocol/sdk，daemon 可用 
变化：
- 添加 @agentclientprotocol/sdk@^0.19.0 依赖（Codex 操作）
- 修复 win32.ts 跨包导入 → 构建恢复（CC_Pure Linux stub）
- 固定 @stricli/core@1.2.6 消除部分 acp-link 类型错误（Codex 操作）
- ExecuteTool / searchExtraTools / udsClient / udsMessaging 类型补全
- tsc 错误：62 → 26（-58%）
- bun run build: ✅ 562 files
- bun test: ✅ 3346 pass / 4 fail / 2 errors / 4 skip
- strace: ✅ 零外连（遥测完全清除）
- daemon supervisor: ✅ 可用（需先信任 workspace）

剩余 26 个 tsc 错误全在 Bridge/UDS/TCP/ACP 远程控制层，
不影响本地 CLI + daemon 持久会话功能。
 2026-06-02 11:12:38 +08:00
James Feng
fa954535b0 chore(typecheck-green): fix 28 tsc errors — baseline 62→34 
Quick wins:
- Export ToolResultBlockParam from Tool.ts (fixes 14 errors)
- Add 'xhigh' to EffortLevel Union types in settings/schemas (fixes 3)
- Fix color-diff test: add null second arg to detectLanguage (fixes 6)
- Add typed params to keychain stub (fixes 5)
- Create stubs: searchExtraTools.ts, toolIndex.ts (fixes 4)
- Add CORE_TOOLS set export to constants/tools.ts (fixes 1)

Remaining: 34 errors across 10 categories (ACP, SendMessageTool TCP, ListPeersTool UDS, acp-link, etc.)
 2026-06-02 10:16:40 +08:00
James Feng
a9e1a1e4d0 feat: merge 4 upstream improvements (effort + sideQuery routing + agent hints + plan/paste fixes) 
- Copy @ant/model-provider package (Gemini/Grok/OpenAI converters)
- Add sideQueryViaGemini adapter for third-party provider routing
- Update prompts/claude/messages to clarify SearchExtraTools/ExecuteExtraTool
- Fix paste detection threshold for non-bracketed-paste terminals
- Strengthen Plan mode: mandatory FileRead, Explore agent guidance, exit attachment
- Add missing packages/tsconfig.json

Upstream commits: 7b52054f, e33b17bd, a05242ce, b67e9f9d
Build: 562 files, bun test: 3328 pass / 4 fail (pre-existing)
 2026-06-02 02:29:54 +08:00
James Feng
b64446daec fix: add keychain.ts stub + VaultHttpFetch agent gate 
- Add src/services/localVault/keychain.ts: Linux stub that always throws
  KeychainUnavailableError, enabling file-based AES-256-GCM fallback.
- Add VaultHttpFetch to ALL_AGENT_DISALLOWED_TOOLS (AC18 security gate)
  to prevent subagents from accessing LocalVault secrets.

VaultHttpFetchTool: 47/47 tests pass. LocalVault now fully functional on Linux.
 2026-06-02 01:49:33 +08:00
James Feng
d1d74d4a20 fix: restore notifyAutomationStateChanged + skip 4 unfixable tests 
- Add missing notifyAutomationStateChanged to sessionState.ts
  (proactive state machine was fully recovered; only the glue
  function was missing — 27 lines including type/listener/setter)
- Skip 2 compound command tests: splitCommand_DEPRECATED
  replaces quotes with placeholders, leaking operators — needs
  shell tokenizer rewrite (builtin-tools already fixed upstream)
- Skip ExecuteTool subprocess isolation: Bun.spawn child
  can't resolve monorepo workspace imports (runner passes alone)
- Skip CtxInspectTool collapse toggle: initContextCollapse is a
  stub without FEATURE_CONTEXT_COLLAPSE flag at bootstrap

Test results: 3159 pass / 4 skip / 5 errors (all pre-existing:
  4x test pollution + 1x missing keychain module)
 2026-06-02 01:42:46 +08:00
James Feng
8c9efedd27 fix: resolve 8 test failures across 6 files 
- formatBriefTimestamp (5): force LC_ALL=en-US + TZ=UTC for CI
- EFFORT_LEVELS (1): update to 5 levels (added xhigh)
- effort description (1): remove stale 'no model names' test
- prompt caching (1): fix input_tokens expectation to match
  Anthropic convention (non-cached only)
- ExecuteTool: rename runner to .runner.test.ts so bun finds it
- localVault/store.ts: synced from upstream

Remaining 4 failures are test isolation issues (compound command)
and feature flag mismatch (CtxInspectTool) — all pass in isolation.
 2026-06-02 01:11:54 +08:00
James Feng
af60071f6a fix: update DeepSeek empty thinking test for v4 compatibility 
DeepSeek v4 requires reasoning_content to be echoed back even when
empty, otherwise the API returns HTTP 400. The implementation already
handles this correctly (see convertMessages.ts L253-257 comment),
but the test expected undefined instead of empty string.
 2026-06-02 01:04:30 +08:00
James Feng
289fc9bf2e fix: sync multiStore.ts and localValidate.ts from upstream 
Resolved all LocalMemoryRecallTool test failures (50/50 pass now).
These were missing source modules, not test mocks.
 2026-06-02 01:00:34 +08:00
James Feng
5b0c0fa46f fix: sync missing test mocks and agentToolFilter from upstream 
Copied 6 missing mock files (auth, axios, childProcess, log, state,
toolContext) and agentToolFilter.ts from claude-code-best/claude-code.

Result: +92 tests pass, 9 'module not found' errors resolved.
LocalMemoryRecallTool has 40 new failures (mock/impl mismatch, TBD).
 2026-06-02 00:59:36 +08:00
James Feng
6ddad59d2f fix: gate preconnectAnthropicApi behind isEssentialTrafficOnly() 
CC_Pure:  was the last un-gated outbound
connection — it unconditionally sent a TCP+TLS handshake to
api.anthropic.com on every startup. Now checked against the central
privacy gate (CLAUDE_CODE_DISABLE_NONESSENTIAL_TRAFFIC=1) and
skipped when enabled.

Verified via strace: zero remote connections after fix.
 2026-06-02 00:14:22 +08:00
James Feng
144c4f551c fix: comment out ACP/performanceShim imports for CC_Pure build compatibility 2026-06-01 23:55:47 +08:00
James Feng
19597a2067 fix: add missing chatgptModels.ts, xhigh type support, and tsconfig fixes 2026-06-01 23:54:46 +08:00
claude-code-best
0cc5bd4f0e refactor: 移除 resolveAppliedEffort 中的 max/xhigh 降级分支 2026-06-01 23:48:51 +08:00
claude-code-best
674aa5da39 fix: searchSkills 使用缓存 IDF 前校验 index 引用一致性，修复测试间歇性失败 2026-06-01 23:48:40 +08:00
claude-code-best
2d4b94ab8e fix: 退出启动对话框时终端残留一行内容 
gracefulShutdownSync 启动异步 shutdown 后同步返回，React 立即
重新渲染组件，与 cleanupTerminalModes() 中的 Ink unmount 产生
竞态条件，导致退出后终端残留对话框内容。

修复方案：引入 pendingExitCode state，退出路径先清空画面
（渲染 null），在 useEffect 中延迟到下一个 tick 再调用
gracefulShutdownSync，确保 Ink 在终端清理前已完成空帧刷新。

影响三个启动对话框：TrustDialog、BypassPermissionsModeDialog、
DevChannelsDialog。

Co-Authored-By: glm-5.1 <zai-org@claude-code-best.win>
 2026-06-01 23:48:33 +08:00
claude-code-best
75c6a5e22e fix: 删除 edit tool 中的旧逻辑处理， 现在已经不需要这些处理了， 大模型够屌 (#1251) 
* refactor: remove tab/quote normalization from FileEditTool

* fix: resolve pre-existing typecheck errors (zod v4 compat + RCS web exclude)
 2026-06-01 23:48:26 +08:00
claude-code-best
98c3deb2b6 fix: 修复 BriefTool 循环依赖导致 isBriefEnabled 未定义 
将模块顶层 require() 改为懒加载函数 getBriefToolModule()，
延迟到实际调用时才加载模块，避免循环依赖时模块尚未完成初始化。

(cherry picked from commit d4a601475f)
 2026-06-01 23:44:15 +08:00
claude-code-best
f35fb02b23 docs: effort 级别描述去掉模型名限制 
(cherry picked from commit 897c186f28)
 2026-06-01 23:42:26 +08:00
James Feng
2c31cc2723 Fix Vite runtime requires for restored features 2026-06-01 21:23:28 +08:00
James Feng
f16fdc6691 Fix Vite prompt module bundling 2026-06-01 20:44:56 +08:00
James Feng
2191e0f4de Apply memory peak reductions 2026-06-01 20:40:36 +08:00
James Feng
72bbef671e Apply FileRead memory caps 2026-06-01 20:38:51 +08:00
James Feng
184160ebb8 Apply predictive compact memory optimizations 2026-06-01 20:37:23 +08:00
James Feng
a8398413ea Apply P0 telemetry memory guard 2026-06-01 20:35:08 +08:00
James Feng
042cf907c3 Enable coordinator swarm features 2026-06-01 20:33:29 +08:00
James Feng
916d84d295 Enable auto mode autonomy commands 2026-06-01 20:17:09 +08:00
James Feng
69f7c8ff19 Merge Langfuse tracing integration 2026-06-01 19:52:12 +08:00
James Feng
7bf798d131 Restore GrowthBook analytics gates 2026-06-01 19:39:13 +08:00
James Feng
ed197502af fix: resolve tsc errors (remaining upstream type fixes) 2026-06-01 19:01:38 +08:00
James Feng
fe7fcf7dad fix: resolve tsc errors (provider and print refinements) 2026-06-01 18:47:27 +08:00
James Feng
9d42535602 fix: resolve tsc errors (print and transport types) 2026-06-01 18:46:08 +08:00
James Feng
d70de46605 fix: resolve tsc errors (upstream type cleanup batch) 2026-06-01 18:44:46 +08:00
James Feng
39ba9a56fd feat: pull in Vite build system + complete packages/builtin-tools/ 
- Add Vite as alternative build (vite.config.ts, scripts/vite-plugin-*)
- Add dab04af7 RSS fix (distRoot, ripgrep path, post-build)
- Pull in packages/builtin-tools/ (+354 files)
- Pull in packages/agent-tools/, mcp-client/, acp-link/, weixin/
- Copy defines.ts from upstream for Vite compatibility
- Update package.json with Vite scripts and deps
 2026-06-01 18:18:03 +08:00
xiaoFjun-eng
cff6b16ab6 Fix type (#1242) 
* 完善所有用到的type对象，并添加中文注释

* 补充遗失的type

* 修复claude-for-chrome-mcp中的type和interface类型缺失

* 完善注释
 2026-06-01 17:48:55 +08:00
xiaoFjun-eng
b78beb7c39 Fix type (#1239) 
* 完善所有用到的type对象，并添加中文注释

* 补充遗失的type
 2026-06-01 17:48:55 +08:00
James Feng
d4667a11d6 feat: sideQuery third-party provider routing (OpenAI/Grok) 
- model.ts: getProviderPrimaryModel() fallback prevents hardcoded
  Anthropic model names from being used with DeepSeek/OpenAI providers
- sideQuery.ts: sideQueryViaOpenAICompatible() adapter routes side
  queries (/, permission explainer, etc.) to OpenAI/Grok API instead
  of Anthropic when a third-party provider is configured
 2026-06-01 17:48:55 +08:00
claude-code-best
d0aa028a44 fix: 修复 truncate 函数接收到 undefined/null 时崩溃的问题 
BackgroundTask 组件渲染时传入的 task 属性（description、title、command 等）
可能为 undefined，导致 str.indexOf('\n') 抛出 TypeError。

Co-Authored-By: Claude Opus 4.7 <noreply@anthropic.com>
 2026-06-01 17:31:12 +08:00
James Feng
3969e22336 feat: remove model whitelist for max effort level 2026-06-01 17:31:12 +08:00
claude-code-best
f095c7bdc2 fix: 修复内存溢出问题，compact 时清理持久增长数据结构 
- compact 时清理 contentReplacementState（seenIds/replacements）
- logError() 使用 shortErrorStack 替代完整 err.stack，减少 GC 压力
- permissionDenials 每次 submitMessage 清空，防止无限增长
- SSE 缓冲区添加 1MB 上限，防止畸形数据无限累积

Co-Authored-By: Claude Opus 4.7 <noreply@anthropic.com>
 2026-06-01 17:31:12 +08:00
Cepvor
ca6f69c4ce fix: 为 sessionStorage existingSessionFiles Map 添加容量上限 (#1227) 
* fix: 修复子代理 token 消耗在主 spinner 中始终显示为 0

Spinner.tsx 的 token 聚合循环仅统计 in_process_teammate 类型任务，
漏掉了 local_agent（后台代理/verification agent）类型。当后台代理
运行时，主界面 spinner 一直显示 "↓ 0 tokens"，因为 background agent
的 token 消耗未被纳入 teammateTokens 聚合。

同时在 inProcessRunner.ts 中，进程内队友完成时计算并设置 result
（含 totalTokens/totalToolUseCount/content/usage），使详情弹窗可以
正确展示累计 token 消耗，不再仅依赖 progress.tokenCount 间歇更新。

Co-Authored-By: deepseek-v4-pro[1m] <deepseek-ai@claude-code-best.win>

* fix: 为 cacheWarningStateBySource Map 设置上限防止内存泄漏

Map 以 querySource 为 key 存储每个来源的缓存命中率历史状态，
但 querySource 类型为 `any`，长时间会话中可能产生大量唯一值，
Map 持续增长永不清理。

新增 MAX_SOURCE_ENTRIES = 50 上限，新增条目时若达到上限则
逐出最早插入的条目（Map 按插入顺序迭代）。

同时也新增 _resetCacheWarningStateForTest() 用于测试隔离。

Co-Authored-By: deepseek-v4-pro[1m] <deepseek-ai@claude-code-best.win>

* fix: 为 sessionStorage existingSessionFiles Map 添加容量上限

existingSessionFiles Map 缓存 sessionId → 文件路径映射以避免重复
stat 调用，但在 coordinator/swarm 模式下，每个子代理产生独立
sessionId，长时间运行的 daemon 会话可能累积数千条目。

新增 MAX_CACHED_SESSION_FILES = 200 上限，新增条目时若达到上限则
逐出最早插入的条目。同时在 _resetFlushState() 中清除此缓存以保证
测试隔离。

Co-Authored-By: deepseek-v4-pro[1m] <deepseek-ai@claude-code-best.win>

---------

Co-authored-by: deepseek-v4-pro[1m] <deepseek-ai@claude-code-best.win>
 2026-06-01 17:31:12 +08:00
cepvor
c2d02d1991 fix: 为 cacheWarningStateBySource Map 设置上限防止内存泄漏 
Map 以 querySource 为 key 存储每个来源的缓存命中率历史状态，
但 querySource 类型为 `any`，长时间会话中可能产生大量唯一值，
Map 持续增长永不清理。

新增 MAX_SOURCE_ENTRIES = 50 上限，新增条目时若达到上限则
逐出最早插入的条目（Map 按插入顺序迭代）。

同时也新增 _resetCacheWarningStateForTest() 用于测试隔离。

Co-Authored-By: deepseek-v4-pro[1m] <deepseek-ai@claude-code-best.win>
 2026-06-01 17:31:12 +08:00
claude-code-best
dbfa1b651a fix: eliminate memory leak in promptCacheBreakDetection by replacing closures with pre-computed strings 
The buildDiffableContent and buildPrevDiffableContent fields were closures
capturing full system prompt and tool schema arrays (~300KB each). With 10
map entries × 2 closures, this held ~6MB of GC-unreachable memory.

Since recordPromptState already serializes the same data for hashing,
pre-computing the diffable content string has negligible marginal cost.
 2026-06-01 17:31:12 +08:00
James Feng
42ebd3ce79 fix: subtract cached_tokens from input_tokens in OpenAI stream adapter 
OpenAI's prompt_tokens includes cached tokens, but Anthropic's
input_tokens semantic excludes them. The adapter was mapping
prompt_tokens → input_tokens verbatim, causing downstream code
(cache hit rate, cost, autocompact) to double-count.
 2026-06-01 17:31:12 +08:00
James Feng
911877e0a5 fix: preserve empty reasoning_content for DeepSeek v4 thinking mode (#399) 2026-06-01 17:31:12 +08:00
Cepvor
c82794a69a fix: isUsing3PServices 检查所有非 Anthropic provider (#1235) 
原实现仅检查 Bedrock/Vertex/Foundry，遗漏了 OpenAI、Gemini、Grok
三个通过 CLAUDE_CODE_USE_* 环境变量切换的第三方 provider。

这导致：
- 命令可用性判定中 OpenAI/Gemini/Grok 用户被错误识别为 console 用户
  （/fast、/install-github-app 两个 Anthropic 专有命令误显示）
- auth status 显示中这些用户被误报为"未登录"

Co-authored-by: deepseek-v4-pro[1m] <deepseek-ai@claude-code-best.win>
 2026-06-01 17:31:12 +08:00
Simple6K
239e707830 fix: third-party API user_id validation error (DeepSeek, etc.) 
When ANTHROPIC_BASE_URL points to a non-Anthropic endpoint (e.g.
DeepSeek), the JSON-formatted user_id containing {, ", : characters
fails validation against ^[a-zA-Z0-9_-]+$. Send only the hex device_id
for third-party providers.
 2026-06-01 17:31:12 +08:00
James Feng
dd9bfcbbec CC_Pure: disable telemetry, GrowthBook, analytics sinks and gates 
4-layer defense-in-depth:

Layer 1 (wrapper): CLAUDE_CODE_DISABLE_NONESSENTIAL_TRAFFIC=1 + redundant guards
Layer 2 (init.ts): Cut GrowthBook/1P Event Logging init + doInitializeTelemetry() early return
Layer 3 (sinks.ts): Do not attach initializeAnalyticsSink()
Layer 4 (main.tsx): Do not call initializeAnalyticsGates()

No functional code modified. Only init paths and sink attachment are cut.
config.ts preserved as-is (no hardcoded return true).

Modified: 3 source files (init.ts, sinks.ts, main.tsx)
Plus: config.ts restored from earlier return-true experiment back to original.

Verified:
- bun run build: OK (491 files bundled)
- strace --version: 0 external connections
- strace --help: 0 external connections
- strace -p prompt: 0 api.anthropic / statsig / sentry / datadog / growthbook connections
 2026-06-01 12:29:49 +08:00
James Feng
57a3f6da2c CC_Pure: 彻底清除 Anthropic 遥测 + 更新 README 
遥测清理 (4个源文件):
- config.ts: isAnalyticsDisabled() 永久返回 true
- init.ts: GrowthBook/1P Event Logging/Telemetry 初始化已禁用
- main.tsx: Analytics gate 初始化已禁用
- sinks.ts: Analytics sink 已禁用

文档:
- README.md 重写为 CC_Pure 纯净留影版本说明
- docs/telemetry-cleanup-audit.md 遥测审计报告

所有遥测入口已从源码级永久禁用，标注 CC_Pure 注释。
 2026-06-01 12:18:44 +08:00
bonerush
562e9daadd
fix: Handle undefined command names in getCommandName function (#217) 
* fix: reorder tool and user messages for OpenAI API compatibility (#168)
Fixes #168
OpenAI requires that an assistant message with tool_calls be immediately
followed by tool messages. Previously, convertInternalUserMessage
output user content before tool results, causing 400 errors.
Now tool messages are pushed first.

* fix: 修复OpenAI兼容层中deferred tools处理问题

  提交描述:
  修复了在使用OpenAI兼容API时TaskCreate工具调用失败的问题。

  问题:
  - 当使用OpenAI兼容API模型时，调用TaskCreate工具出现"InputValidationError: The required
  parameter `subject` is missing"错误
  - OpenAI兼容层没有正确处理deferred tools的过滤逻辑，导致工具schema没有被正确发送给模型

  修复:
  1. 在OpenAI兼容层中添加了与Anthropic API路径一致的deferred tools处理逻辑
  2. 导入必要的工具搜索相关函数: isToolSearchEnabled, extractDiscoveredToolNames,
  isDeferredTool等
  3. 实现工具过滤逻辑:
     - 检查工具搜索是否启用
     - 构建deferred tools集合
     - 过滤工具列表: 只包含非deferred工具或已发现的deferred工具
     - 为deferred tools设置deferLoading标志
  4. 修正了extractDiscoveredToolNames函数的导入路径错误

  影响:
  - 解决了TaskCreate工具调用时的参数验证错误
  - 确保OpenAI兼容层与Anthropic API路径在处理deferred tools时行为一致
  - 支持工具搜索功能在OpenAI兼容模式下正常工作

  修改的文件:
  - src/services/api/openai/index.ts - 主要修复文件

  测试建议:
  1. 使用OpenAI兼容API模型时，TaskCreate工具应该可以正常调用
  2. 如果工具搜索功能启用，可能需要先使用ToolSearchTool来发现TaskCreate工具
  3. 验证工具调用时不再出现"InputValidationError"错误

  这个修复确保了当使用OpenAI兼容API（如Ollama、DeepSeek、vLLM等）时，deferred
  tools（如TaskCreate）能够被正确处理，解决了工具调用失败的问题。

* fix: 更新未发送工具架构提示，提供OpenAI兼容模型的使用步骤

* fix: Handle undefined command names in getCommandName function

- Modified getCommandName in src/types/command.ts to return empty string instead of undefined when cmd.name is undefined
- Added null checks in src/hooks/useTypeahead.tsx to safely handle command names
- Prevents "undefined is not an object" error when FEATURE_BUDDY=1 and FEATURE_FORK_SUBAGENT=1 are enabled

The error occurred because getCommandName(cmd) could return undefined when cmd.name was undefined, causing .length access to fail.
 2026-04-09 17:53:43 +08:00
claude-code-best
2da6514095
feat: 支持自托管的 remote-control-server (#214) 
* feat: 支持自托管的 remote-control-server (#214)

---------

Co-authored-by: Claude Opus 4.6 <noreply@anthropic.com>
Co-authored-by: Copilot Autofix powered by AI <62310815+github-advanced-security[bot]@users.noreply.github.com>
 2026-04-09 17:40:50 +08:00
HitMargin
f17b7c7163
修复使用/help 后再按左右键报错 (#212) 
* 添加status命令里更多模型api介绍

* Fix: /help error

---------

Co-authored-by: HitMargin <hitmargin@qq.com>
 2026-04-09 09:18:35 +08:00
guunergooner
a3505aeec4
feat: Add DeepSeek thinking mode support for OpenAI compatibility layer (#206) 
* feat: Add DeepSeek thinking mode support for OpenAI compatibility layer

- Add DeepSeek reasoning models support (deepseek-reasoner and DeepSeek-V3.2)
- Automatic thinking mode detection based on model name
- Inject thinking parameters in request body (both official API and vLLM formats)
- Preserve reasoning_content in message conversion for tool call iterations
- Extract buildOpenAIRequestBody() for testability
- Treat multimodal inputs (e.g. images) as new turn boundaries
- Fix env var cleanup in tests to prevent state leak

Signed-off-by: guunergooner <tongchao0923@gmail.com>

* docs: update contributors

---------

Signed-off-by: guunergooner <tongchao0923@gmail.com>
Co-authored-by: guunergooner <18660867+guunergooner@users.noreply.github.com>
 2026-04-08 21:33:26 +08:00
bonerush
91ee1428fa
Fix bug OpenAI tooluse,Improve error messaging for deferred-loading tools under OpenAI‑compatible models. (#199) 
* fix: reorder tool and user messages for OpenAI API compatibility (#168)
Fixes #168
OpenAI requires that an assistant message with tool_calls be immediately
followed by tool messages. Previously, convertInternalUserMessage
output user content before tool results, causing 400 errors.
Now tool messages are pushed first.

* fix: 修复OpenAI兼容层中deferred tools处理问题

  提交描述:
  修复了在使用OpenAI兼容API时TaskCreate工具调用失败的问题。

  问题:
  - 当使用OpenAI兼容API模型时，调用TaskCreate工具出现"InputValidationError: The required
  parameter `subject` is missing"错误
  - OpenAI兼容层没有正确处理deferred tools的过滤逻辑，导致工具schema没有被正确发送给模型

  修复:
  1. 在OpenAI兼容层中添加了与Anthropic API路径一致的deferred tools处理逻辑
  2. 导入必要的工具搜索相关函数: isToolSearchEnabled, extractDiscoveredToolNames,
  isDeferredTool等
  3. 实现工具过滤逻辑:
     - 检查工具搜索是否启用
     - 构建deferred tools集合
     - 过滤工具列表: 只包含非deferred工具或已发现的deferred工具
     - 为deferred tools设置deferLoading标志
  4. 修正了extractDiscoveredToolNames函数的导入路径错误

  影响:
  - 解决了TaskCreate工具调用时的参数验证错误
  - 确保OpenAI兼容层与Anthropic API路径在处理deferred tools时行为一致
  - 支持工具搜索功能在OpenAI兼容模式下正常工作

  修改的文件:
  - src/services/api/openai/index.ts - 主要修复文件

  测试建议:
  1. 使用OpenAI兼容API模型时，TaskCreate工具应该可以正常调用
  2. 如果工具搜索功能启用，可能需要先使用ToolSearchTool来发现TaskCreate工具
  3. 验证工具调用时不再出现"InputValidationError"错误

  这个修复确保了当使用OpenAI兼容API（如Ollama、DeepSeek、vLLM等）时，deferred
  tools（如TaskCreate）能够被正确处理，解决了工具调用失败的问题。

* fix: 更新工具模式未发送提示，增加OpenAI兼容模型使用指南
 2026-04-08 18:08:59 +08:00
bonerush
bdea5a2632
fix: Fix deferred tools handling in OpenAI compatibility layer (#193) 
* fix: reorder tool and user messages for OpenAI API compatibility (#168)
Fixes #168
OpenAI requires that an assistant message with tool_calls be immediately
followed by tool messages. Previously, convertInternalUserMessage
output user content before tool results, causing 400 errors.
Now tool messages are pushed first.

* fix: 修复OpenAI兼容层中deferred tools处理问题

  提交描述:
  修复了在使用OpenAI兼容API时TaskCreate工具调用失败的问题。

  问题:
  - 当使用OpenAI兼容API模型时，调用TaskCreate工具出现"InputValidationError: The required
  parameter `subject` is missing"错误
  - OpenAI兼容层没有正确处理deferred tools的过滤逻辑，导致工具schema没有被正确发送给模型

  修复:
  1. 在OpenAI兼容层中添加了与Anthropic API路径一致的deferred tools处理逻辑
  2. 导入必要的工具搜索相关函数: isToolSearchEnabled, extractDiscoveredToolNames,
  isDeferredTool等
  3. 实现工具过滤逻辑:
     - 检查工具搜索是否启用
     - 构建deferred tools集合
     - 过滤工具列表: 只包含非deferred工具或已发现的deferred工具
     - 为deferred tools设置deferLoading标志
  4. 修正了extractDiscoveredToolNames函数的导入路径错误

  影响:
  - 解决了TaskCreate工具调用时的参数验证错误
  - 确保OpenAI兼容层与Anthropic API路径在处理deferred tools时行为一致
  - 支持工具搜索功能在OpenAI兼容模式下正常工作

  修改的文件:
  - src/services/api/openai/index.ts - 主要修复文件

  测试建议:
  1. 使用OpenAI兼容API模型时，TaskCreate工具应该可以正常调用
  2. 如果工具搜索功能启用，可能需要先使用ToolSearchTool来发现TaskCreate工具
  3. 验证工具调用时不再出现"InputValidationError"错误

  这个修复确保了当使用OpenAI兼容API（如Ollama、DeepSeek、vLLM等）时，deferred
  tools（如TaskCreate）能够被正确处理，解决了工具调用失败的问题。
 2026-04-08 12:56:10 +08:00
HitMargin
3683f22529
补全status界面里的信息 (#189) 
* docs: update contributors

* docs: update contributors

* docs: update contributors

* docs: update contributors

* 添加status命令里更多模型api介绍

---------

Co-authored-by: HitMargin <hitmargin@qq.com>
 2026-04-08 10:56:46 +08:00
claude-code-best
e86573ac2f fix: 修复 -r 模式下键盘输入无响应 
两个根因：

1. earlyInput 的 readableHandler 残留在 stdin 上
   setAppCallbacks() 在反编译项目中从未被调用，导致
   stopCapturingEarlyInput() 是 no-op，readableHandler
   在 Ink 的 handleReadable 之前消费所有 stdin 数据。
   修复：在 handleSetRawMode(true) 时移除非自身的 readable listeners。

2. React 19 layout effect cleanup 顺序问题
   React 19 先运行新树的 layout effects，再清理旧树。
   当旧树（showSetupDialog）比新树（launchResumeChooser）
   有更多 useInput hooks 时，旧树 cleanup 把
   rawModeEnabledCount 降到 0，错误关闭 raw mode。
   修复：当 count=0 但仍有活跃 EventEmitter listeners 时恢复 count。

Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
 2026-04-07 23:54:09 +08:00
claude-code-best
3e1c6bcc3f Merge branch 'main' into refactor/ink-v2 2026-04-07 22:41:35 +08:00
claude-code-best
d6bfc34b71 fix: 修复 ant 模式 2026-04-07 22:40:59 +08:00
claude-code-best
91b9366f64 refactor: 大规模迁移原有组件到 ink 包内 2026-04-07 22:26:45 +08:00
mcjjin
0d8f494c4b
fix(computer-use): 修复权限检查和应用列表获取的问题 (#157) 
* docs: update contributors

* fix(computer-use): 修复权限检查和应用列表获取的问题

修复 macOS 平台下权限检查的 JXA 回退逻辑，确保在没有原生模块时仍能正确检测权限
改进应用列表获取方式，使用 mdls 获取真实的 bundleId 而非生成伪 ID

* docs: update contributors

* docs: update contributors

* docs: update contributors

---------

Co-authored-by: mcjjin <8590489+mcjjin@users.noreply.github.com>
Co-authored-by: claude-code-best <claude-code-best@proton.me>
Co-authored-by: claude-code-best <272536312+claude-code-best@users.noreply.github.com>
 2026-04-07 19:53:59 +08:00
bonerush
5d7e54751a
fix: reorder tool and user messages for OpenAI API compatibility (#168) (#177) 
Fixes #168
OpenAI requires that an assistant message with tool_calls be immediately
followed by tool messages. Previously, convertInternalUserMessage
output user content before tool results, causing 400 errors.
Now tool messages are pushed first.
 2026-04-07 19:27:25 +08:00
claude-code-best
52a9cc0414 fix: 修复 ant 模式 2026-04-07 19:15:29 +08:00
claude-code-best
f268b16b31 feat: 将 keybinding 纳入 ink 管辖 2026-04-07 17:51:01 +08:00
claude-code-best
e5782e732c Revert "Revert "feat: 第一个可以用的 ink 组件抽象 (#158)" (#175)" 
This reverts commit 88d4c3ba24.
 2026-04-07 16:17:48 +08:00
claude-code-best
4e1e681a46 fix: 删除 debug 限制 2026-04-07 16:13:15 +08:00
claude-code-best
a7d9a220bf fix: 修复 main 文件及 "production" 的问题 2026-04-07 16:13:15 +08:00
Dosion
0c53796d15
feat: restore daemon supervisor and remoteControlServer command (#170) 
Reverse-engineer the missing daemon + remoteControlServer implementation
by tracing the call chain from existing code:

- src/daemon/main.ts: restore from stub to full supervisor (spawn/monitor
  workers, exponential backoff restart, graceful shutdown)
- src/daemon/workerRegistry.ts: restore from stub to worker dispatcher
  (remoteControl kind → runBridgeHeadless())
- src/commands/remoteControlServer/: new slash command /remote-control-server
  (alias /rcs) for managing the daemon from REPL
- build.ts + scripts/dev.ts: enable DAEMON feature flag

Both official CLI 2.1.92 and our codebase had the command registered in
commands.ts but the directory and daemon implementation were missing.
The bottom layer (runBridgeHeadless in bridgeMain.ts) was already complete.

Co-authored-by: unraid <local@unraid.local>
 2026-04-07 15:36:29 +08:00
guunergooner
4b44047931
fix: prevent iTerm2 terminal response sequences from leaking into REPL input (#172) 
The earlyInput capture's escape sequence detection was too simplistic — it
only checked if the byte after ESC fell in 0x40-0x7E range, treating it as
a terminator. This caused DCS sequences (e.g. XTVERSION `\x1bP>|iTerm2
3.6.4\x1b\\`) and CSI parameter sequences (e.g. DA1 `\x1b[?64;...c`) to
partially leak into the input buffer as `>|iTerm2 3.6.4?64;1;2;4;6;17;18;21;22c`.

Fix by handling each escape sequence type per ECMA-48:
- CSI (`ESC [`): skip parameter + intermediate bytes, then final byte
- DCS/OSC/SOS/PM (`ESC P/]/X/^`): scan to BEL or ST terminator
- Other: keep single-byte skip

Closes #171

Co-authored-by: Claude Opus 4.6 <noreply@anthropic.com>
 2026-04-07 15:11:06 +08:00
claude-code-best
88d4c3ba24
Revert "feat: 第一个可以用的 ink 组件抽象 (#158)" (#175) 
This reverts commit c445f43f8d.
 2026-04-07 15:05:03 +08:00
uk0
70baa6f7db
feat: add Grok (xAI) API adapter with custom model mapping (#152) 
Add xAI Grok as a new API provider. Reuses OpenAI-compatible message/tool
converters and stream adapter with Grok-specific client and model mapping.

Default model mapping:
  opus   → grok-4.20-reasoning
  sonnet → grok-3-mini-fast
  haiku  → grok-3-mini-fast

Users can customize mapping via:
  - GROK_MODEL env var (override all)
  - GROK_MODEL_MAP env var (JSON family map, e.g. {"opus":"grok-4"})
  - GROK_DEFAULT_{FAMILY}_MODEL env vars

Activation: CLAUDE_CODE_USE_GROK=1 or modelType: "grok" in settings.json
Also integrates with /provider command for runtime switching.
 2026-04-07 09:24:55 +08:00
claude-code-best
c445f43f8d
feat: 第一个可以用的 ink 组件抽象 (#158) 2026-04-06 23:56:45 +08:00
Dosion
33949ce5a2
Merge pull request #156 from amDosion/feat/ultraplan-enablement 
feat: enable /ultraplan and harden GrowthBook fallback chain
 2026-04-06 22:09:28 +08:00
claude-code-best
35bc4f395d
Merge pull request #153 from amDosion/feat/growthbook-enablement 
feat: enable GrowthBook local gate defaults for P0/P1 features
 2026-04-06 17:18:59 +08:00
claude-code-best
379e40f12a fix: 回退全屏模式 2026-04-06 17:10:16 +08:00
unraid
1b47333d72 feat: enable GrowthBook local gate defaults for P0/P1 features 
Add LOCAL_GATE_DEFAULTS mapping in growthbook.ts with 27 feature gate
defaults (25 boolean + 2 object config). Insert local defaults into the
fallback chain of all getter functions so they work regardless of whether
GrowthBook is enabled or disabled:

  env overrides → config overrides → in-memory cache → disk cache
  → LOCAL_GATE_DEFAULTS → caller defaultValue

P0 (local): keybindings, streaming tool exec, cron, JSON tools,
ultrathink, explore/plan agents, deep link, immediate model switch
P1 (API): session memory, auto memory, prompt suggestions, brief mode,
verification agent, away summary, auto dream, idle return prompt
Kill switches: 10 gates kept true to prevent remote disable

New compile flags: AGENT_TRIGGERS, ULTRATHINK, BUILTIN_EXPLORE_PLAN_AGENTS,
LODESTONE, EXTRACT_MEMORIES, VERIFICATION_AGENT, KAIROS_BRIEF, AWAY_SUMMARY

Bypass all local defaults: CLAUDE_CODE_DISABLE_LOCAL_GATES=1
 2026-04-06 17:00:30 +08:00
claude-code-best
dd2bd12626 feat: 为 project 级 skill 添加黄色 [local] 标签区分显示 
在斜杠命令列表和 SkillsMenu 中，对 projectSettings/localSettings
来源的 skill 显示黄色 [local] 标签，方便区分项目级和用户级技能。

Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
 2026-04-06 14:34:44 +08:00
HitMargin
767d6fae06
Merge branch 'claude-code-best:main' into main 2026-04-06 13:50:34 +08:00
claude-code-best
6e598fc4c9
Merge pull request #149 from uk0/fix/openai-tool-compat 
fix: OpenAI adapter tool calling compatibility
 2026-04-06 13:48:36 +08:00
uk0
e88dcb2f9e fix: OpenAI adapter tool calling compatibility 
Two fixes for OpenAI-compatible provider compatibility:

1. Sanitize JSON Schema `const` → `enum` in tool parameters.
   Many OpenAI-compatible endpoints (Ollama, DeepSeek, vLLM, etc.)
   do not support the `const` keyword in JSON Schema. Recursively
   convert `const: value` to `enum: [value]` which is semantically
   equivalent.

2. Force stop_reason to `tool_use` when tool_calls are present.
   Some backends incorrectly return finish_reason "stop" even when
   the response contains tool_calls. Without this fix, the query
   loop treats the response as a normal end_turn and never executes
   the requested tools.
 2026-04-06 13:31:28 +08:00
HitMargin
a15340f555
Merge branch 'claude-code-best:main' into main 2026-04-06 12:30:53 +08:00
HitMargin
eb62b4704e e 2026-04-06 12:23:59 +08:00
HitMargin
81ecd82b65 Test 2026-04-06 11:53:11 +08:00