The "Create a file like clis/<site>/<command>.ts" example predates #928, which
converted the entire adapter layer from TypeScript to JavaScript. The repo now
ships 0 .ts and 1259 .js built-in adapters, so a contributor following the doc
creates a file in the wrong format. Update the example to JavaScript (keeping a
pointer to the still-supported TypeScript path), and fix the adapter test
command, which pointed at src/ rather than the adapter's own clis/ test file.
Co-authored-by: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
* Add ChatGPT Deep Research result extraction
* fix(chatgpt): bind deep research results to requested conversation
* fix(chatgpt): preserve deep research payload failures
---------
Co-authored-by: jackwener <jakevingoo@gmail.com>
* feat(adapters): add 懂车帝 (dongchedi) + 瓜子二手车 (guazi) car adapters
Two no-login PUBLIC adapters for Chinese car platforms. Both read
server-rendered data (no cookies, no signature, no browser) and ship
pure parsers unit-tested against frozen real-data fixtures.
dongchedi (6 commands) — parses __NEXT_DATA__ SSR JSON:
search 车系搜索 + 指导价/经销商价
series 车系概览（品牌/价格/懂车分/销量排名/款型数）
models 款型列表 + 价格
specs 配置概览（尺寸/动力/四驱/悬挂/气囊）
score 懂车分 8 维评分 + 同级对比
koubei 车主口碑/评价正文
(Dongchedi's /motor XHR APIs are ByteDance-signature gated; the SSR
pages expose the same data unsigned, so the adapter reads those.)
guazi (2 commands) — parses m.guazi.com mobile SSR HTML:
browse 分城市在售二手车列表（售价/里程/年份）
car 车源详情（售价/上牌/里程/过户/配置/车况）
(Desktop www.guazi.com is signature-locked; mobile SSR is open. Deep
pagination/filtering uses the signed API and is intentionally omitted.)
Gates green: tsc, doc-coverage --strict, silent-column-drop (new=0),
typed-error-lint (no new), 24 adapter tests passing.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
* feat(adapters): add 汽车之家 (autohome) — brand catalog + 口碑 ratings
Third no-login PUBLIC car adapter (search by brand, not free text).
autohome (2 commands):
brand 按品牌列出全部车系 + 厂商指导价（grade/carhtml/<INITIAL>.html，
中文品牌名→拼音首字母目录页，DL 块按品牌定位）
score 车系口碑评分：总分 + 各维度 + 故障率PPH + 竞品对比
（k.autohome.com.cn/<id> 的 __NEXT_DATA__.baseData，免登录免签名）
Deliberately omitted (would be silently-wrong without a browser running
Autohome's signing/anti-scrape code): free-text keyword search (signature
gated) and full per-trim 参数配置 (rotating CSS font-glyph obfuscation).
Use dongchedi search/specs/koubei for those. Documented in the adapter doc.
Gates green: tsc, doc-coverage --strict (170/170), silent-column-drop
(new=0), typed-error-lint (no new); 31 adapter tests passing across the
three car adapters.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
* fix(car-adapters): fail closed on parser drift
* fix(guazi): fail closed on empty SSR listings
---------
Co-authored-by: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Co-authored-by: jackwener <jakevingoo@gmail.com>
* feat(github): add `github trending` adapter
Add a PUBLIC adapter that lists repositories from
https://github.com/trending — the trending view is a public HTML page with
no official REST API, so the data was previously unreachable through opencli.
The adapter fetches the page server-side (no browser, no auth) and parses
each repo's full name, description, primary language, total stars, forks,
and stars gained in the period.
Flags:
- `--since` daily | weekly | monthly (default daily)
- `--language` filter by language slug, e.g. python, rust, "c++"
- `--limit` 1..25 (GitHub lists at most 25)
Typed errors: ArgumentError for bad --since / --limit, CommandExecutionError
on request/HTTP failure, EmptyResultError when the page yields no repos.
Tests cover parsing (stars/forks/language/description/url, missing language),
limit truncation, language+since URL building, argument validation, the
empty-result path, and non-ok HTTP.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
* refactor(github-trending): rename to standalone `github-trending repos`
Move the trending scraper out of the `github` site namespace into a
dedicated `github-trending` adapter to avoid confusion with the bundled
`gh` external CLI. Command is now `opencli github-trending repos`
(site=github-trending, name=repos), leaving room for a future
`developers` subcommand. The `github` site retains only login/whoami.
Regenerated cli-manifest.json; 8 fixture tests + typecheck pass.
* fix(github-trending): fail closed on parser drift
---------
Co-authored-by: minh <claude@ttfy.cc>
Co-authored-by: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Co-authored-by: jackwener <jakevingoo@gmail.com>
Final head 5f5661c270.
Contract: adds a public read-only Internet Archive adapter with archive search, item, wayback, and snapshots commands. All commands are read access, browser:false, with no login, write, upload, or browser UI side effects. Source of truth is Internet Archive Advanced Search response.docs, /metadata/<identifier>, Wayback available closest snapshot, and CDX JSON header/rows. Fail-closed boundaries: search requires response.docs array, true empty maps to EmptyResultError, rows require stable identifier and numeric downloads; item requires metadata.identifier equals requested identifier and files array, while missing metadata/404 remains empty; wayback distinguishes no closest snapshot true empty from available:true missing URL or 14-digit timestamp malformed CommandExecutionError; snapshots requires top-level CDX array, header array, required columns, and per-row timestamp/original/statuscode/mimetype cells, with malformed shapes typed CommandExecutionError rather than empty snapshot URLs or empty status columns.
Validation: lead+aux final green. GitHub final gate OPEN/non-draft/CLEAN, required checks SUCCESS, adapter/smoke skipped. Local reviewer validation covered current-main merge-tree clean, targeted/full clis/archive 1 file/18 tests, typecheck, build manifest 1231, docs-build, doc coverage 165/165, prod audit clean, node --check touched Archive files/tests, diff-check, typed-error-lint and silent-column-drop no new.
Final head 2279aca321.
Contract: chatgpt model keeps the existing write surface and supports instant, medium, high, extra-high, and pro intelligence levels; thinking remains a backward-compatible alias for high. Unsupported or unknown requested levels fail upfront with ArgumentError. The selection path requires a logged-in ChatGPT composer and native click. The model selector prefers stable test id / exact visible option text, covering current English Instant/Medium/High/Extra High/Pro and Chinese 极速/均衡/高级/超高/专业 labels. Unknown localization falls back to order only when composer-intelligence-picker-content exists and exactly five visible menuitemradio options are present; otherwise it typed-fails with CommandExecutionError instead of treating an ordinary menu or drifted DOM as success. Postcondition re-reads current selector/test id after click; when label recognition is unavailable, it reopens the picker and verifies the target checked index in the five-option intelligence picker. High vs Extra High matching uses longest/exact ordering to avoid substring false success. No ask/send/read/image/history surface changes.
Validation: lead+aux content green. Remote statusCheckRollup is empty, merged under the standing no-check override after final poll confirmed OPEN/non-draft, head unchanged, and MERGEABLE/UNSTABLE with no conflict/dirty/content blocker. Local reviewer validation covered current-main replay clean, focused ChatGPT utils+commands tests 77/77, full ChatGPT 6 files/118, typecheck, build manifest 1227, docs-build, doc coverage 164/164, prod audit clean, node --check touched files/tests, diff-check, typed-error-lint and silent-column-drop no new, and merge-tree clean.
Final head 53686e6a62.
Contract: xiaohongshu ask remains a browser-backed write command. Answer success still requires the same-send message_id/conversation_id plus a finished non-empty answer. Source identity/url/xsec_token are still trusted only from a 24-hex note id, xhsdiscover://item/<id>, or trusted XHS note URL. New source metadata is a minimal optional enrichment: note_type, user_id, and published_at are forwarded from the 点点 source payload and omitted when empty; like_count is parsed only from non-negative safe integers, strict decimal compact counts with 万/w/W/亿 and optional +, or legal pure digit/thousands strings. Malformed values such as 1e2, 0x10, 1..2万, bad comma grouping, negatives, and decimal numbers are omitted rather than coerced into successful counts. No extra note/detail round-trip or new write surface.
Validation: lead+aux final green. GitHub final gate OPEN/non-draft/CLEAN, required checks SUCCESS, adapter/smoke skipped. Local reviewer validation covered focused clis/xiaohongshu/ask.test.js 1 file/15, full XHS adapter 20 files/270, typecheck, build manifest 1227, docs-build, node --check ask/touched tests plus dist/src/main.js, diff-check clean, and local merge-tree clean.
Final head d41552312e.
Contract: xiaohongshu publish gains text-image publishing behind --card-text. --card-text/--images require at least one content source. Normal image suffix/path validation and text-image gif append are pre-navigation ArgumentError. Text-image flow enters 文字配图, writes/verifies each card, waits for a new active empty card before multi-card input, generates previews, then Next enters the standard editor before title/body/topics. Explicit --card-style must resolve/click or typed-fail with CommandExecutionError, with no silent fallback to 基础. After Generate/Next and image append, visible media count postconditions must prove cards and appended images landed. Publish/draft final success requires a success marker or leaving the publish page; warning-like text cannot fake success.
Validation: lead+aux final green on d4155231. GitHub final gate OPEN/non-draft/CLEAN with required checks SUCCESS and adapter/smoke skipped. Local reviewer validation covered publish tests 31/31, full Xiaohongshu 17 files/246 tests, typecheck, build manifest 1225, docs-build, doc coverage 164/164, production audit clean, node --check, diff-check, typed-error-lint/silent-column-drop no new, and merge-tree clean.
Adds Xiaohongshu saved and liked collection read commands.\n\nFinal review contract:\n- saved/liked are read-only current-user collection scrapers; no write, unlike, or favorite side effects.\n- --limit validates strictly in 1..100 before browser navigation.\n- Collection page location is read back after goto and after each scroll before trusting captures/DOM.\n- Location must be exactly https://www.xiaohongshu.com/user/profile/<resolvedUserId>; /login or login-wall text maps to AuthRequiredError; other host/path/profile drift maps to CommandExecutionError.\n- Browser Bridge {session,data} envelopes are unwrapped; malformed location, non-array intercepted requests, and non-array DOM extraction fail closed.\n- API/DOM notes require stable note id and xsec_token; output URLs round-trip to note/detail.\n- Auth/private/empty/malformed/API/parser/selector drift typed boundaries are preserved.\n\nValidation:\n- Lead and aux final green on ef12a197.\n- Final GitHub poll: open, non-draft, head ef12a197, statusCheckRollup empty only; merged under standing no-check override because content review is green and no conflict/dirty/content blocker remains.\n- Focused collection/saved/liked tests 3 files / 19.\n- Full XHS adapter 20 files / 253.\n- typecheck, build/manifest 1227, docs-build.\n- node --check touched collection files + dist/src/main.js.\n- typed-error-lint and silent-column-drop no new issues; listing-id advisory 13; diff-check clean.
Adds targeted Discord desktop-app read/navigation commands with fail-closed identity checks.\n\nFinal review contract:\n- Surface remains read/navigation only: goto, channels, servers, threads, read, and thread-read; no Discord write command is introduced.\n- Browser Bridge evaluate results are unwrapped and shape-guarded at the Node boundary for route state, list rows, messages, and thread lists.\n- True empty list/read/thread-read results map to EmptyResultError; malformed rows or browser output map to CommandExecutionError.\n- Targeted read and thread-read verify message rows bind to the requested channel/thread via channel_id; stale, wrong-channel, parent-channel, or missing identity rows typed-fail.\n- List rows require stable identities: channels require Channel/guild_id/channel_id/url, servers require Server/guild_id/url, threads require Thread/guild_id/channel_id/thread_id/url.\n\nValidation:\n- Lead final and aux final green on e84a7d85.\n- merge-tree clean against base 8ed8ca26.\n- Discord app tests 19/19.\n- build manifest 1225, typecheck, docs-build, doc coverage 164/164, node --check touched files/tests, diff-check.\n- typed-error-lint and silent-column-drop no new issues.\n- GitHub required checks SUCCESS; adapter/smoke skipped.
* fix(daemon): SIGKILL fallback when stale daemon refuses graceful shutdown
When the CLI detects a stale daemon (`daemonVersion !== PKG_VERSION` after
`npm install -g @jackwener/opencli@latest`), it currently asks the daemon to
exit via `POST /shutdown` and waits up to 3 seconds for the port to release.
If the old daemon hangs, refuses /shutdown, or the endpoint is missing
entirely (pre-shutdown-endpoint version), the port stays held and the user
sees `Stale daemon could not be replaced` with a `opencli daemon stop` hint.
99% of "I just upgraded and have to run `opencli doctor` every time" reports
land here: the user upgraded the CLI but the persistent daemon survived from
a previous install, and graceful shutdown is unreliable across versions.
This patch reads the stale daemon's pid from its existing `/status` response
(daemon.ts:252 already exposes `pid: process.pid`) and falls back to
`process.kill(pid, 'SIGKILL')` after graceful shutdown fails, then waits
another 2s for the port to release. Cross-platform: Node's
`process.kill(_, 'SIGKILL')` maps to `TerminateProcess` on Windows, so no
`taskkill` shell-out is needed.
The user-visible "Stale daemon could not be replaced" error only fires when
both graceful shutdown AND SIGKILL fail (cross-user owner / cross-machine
PID — neither is reachable from a normal CLI invocation anyway). The hint
message is updated to reflect that.
Adds 2 tests:
- SIGKILL succeeds → bridge proceeds past the stale block (and eventually
fails the no-extension wait, proving the stale branch was passed cleanly).
- SIGKILL throws EPERM AND waitForDaemonStop still returns false → falls
through to the existing stale-daemon error.
Troubleshooting docs note the new auto-fallback.
* address opus review nits
- bridge.ts: move `await waitForDaemonStop(2000)` out of the try/catch so the
port poll always runs after `process.kill`, even when the kill itself throws
ESRCH (target already dead) or EPERM (cross-user owner).
- browser.test.ts: bump the existing 3 stale-daemon test fixtures from
`pid: 1` to `pid: 999999` so the new SIGKILL fallback no longer fires a
signal at init when the older tests reach the fallback path.
- browser.test.ts: mock `waitForDaemonStop` in those 3 tests too, since the
real implementation now polls for 2s in the fallback path (test runtime
was up to ~6s before; back to ~400ms).
* Revert "chore(skills): remove smart-search (#1683)"
This reverts commit 7a2ab47bf8.
* chore(readme): keep smart-search out of README per @WAWQAQ
Restore smart-search skill files and inner-docs refs, but drop the 6
README mentions (3 EN + 3 ZH). Skill is loadable via:
npx skills add jackwener/opencli --skill smart-search
but no longer surfaced on the README front page.
Adong