@wji.ai/cli 0.5.0 → 0.6.0

package/skill/wji/SKILL.md

@@ -9,6 +9,12 @@ description: Drive the WJI AI platform from the terminal with the `wji` CLI —
 `--json`, so prefer `--json` piped through `jq` when you need to read a value,
 and the human table output when you are showing the user a result.
 
+`--json` prints the **`standard`** detail level: ids, names, status, media urls,
+long text clipped to 200 chars and `content` reduced to a summary marked
+`truncated: true`. Add `--detail full` whenever you need the whole `content`,
+`form` or `settings` (always before a read-modify-write), and `--detail brief`
+when you only want ids. `WJI_DETAIL=full` sets it for a whole session.
+
 ## Before anything else
 
 ```bash
@@ -94,7 +100,7 @@ wji ws fs                                  # whole workspace tree
 wji ws fs --app <appId>                    # one app's tree
 wji fs tree <nodeId> --depth 2
 wji fs ls <folderId>
-wji fs get <nodeId> --json | jq .content
+wji fs get <nodeId> --json --detail full | jq .content      # --detail full: the summary is not the content
 
 wji fs mkdir <parentId> characters --nickname 角色
 wji dataset new <folderId> "Ada" -t character -d "a tired detective"
@@ -211,6 +217,25 @@ balanced default; **Seedance 2.0 Fast / MiniMax H3 Max** for drafts, tests and
 large batches. The ranking is a guide — quote the real number from
 `wji gen cost` / `wji canvas gen` before spending.
 
+**Uncensored (無審核) models.** Use these only when the user asks for
+uncensored output or a censored model refused the content. They carry the
+`uncensored` tag in `wji models list --json`, and their names end in
+`-uncensored`. The Wan 3.0 family has a Ref and a first/last-frame variant of
+each tier, like the censored Wan 3.0.
+
+| Job | Model (`name`) | Inputs |
+| --- | --- | --- |
+| Video, Ref | `wan-3.0-ref-uncensored`, `wan-3.0-prime-ref-uncensored` (480p–1080p); `wan-3.0-pro-ref-uncensored`, `wan-3.0-prime-pro-ref-uncensored` (1080p / 2k / 4k) | ≤ 10 images + ≤ 5 videos + ≤ 5 audios, freely mixed (video/audio ≤ 15 s total) |
+| Video, first/last frame | `wan-3.0-uncensored`, `wan-3.0-prime-uncensored`, `wan-3.0-pro-uncensored`, `wan-3.0-prime-pro-uncensored` | text only, or start and/or end image; 2–30 s |
+| Image → video | `wan-2.7-i2v-uncensored` (720p/1080p, 2–15 s; optional 1 driving audio for lip-sync via `--ref-audio`), `wan-2.2-i2v-uncensored` (cheapest; 480p/720p, 5 or 8 s, optional end frame) | start image |
+| Animate a character | `wan-animate-uncensored` (`--set mode=animate\|replace`) | exactly 1 image + 1 video; output is as long as the video |
+| Text → image | `z-image-uncensored` (long edge ≤ 1536), `z-image-pro-uncensored` (≤ 2560, `--set long_edge=`) | prompt + ratio |
+| Edit an image | `qwen-image-edit-uncensored` | 1–3 images: the **first** is the image to edit |
+| Face / head swap | `face-swap-uncensored`, `head-swap-uncensored` | exactly 2 images, **order matters**: 1st = target picture, 2nd = the face / head to put on it |
+| Video upscale | `flashvsr-uncensored` (`--set resolution=720p\|1080p\|2k\|4k`) | 1 MP4 ≤ 120 s; billed per input second |
+
+"Prime" variants are faster and cost more; "Pro" variants output 1080p–4K.
+
 ### Limits (per account)
 
 | Limit | Value | When exceeded |
@@ -248,7 +273,7 @@ headshot and sheet as image references and the voice as an audio reference to
 a Ref model.
 
 ```bash
-wji fs get <datasetId> --scope full --json | jq '{images: [.images[]? | {id, type}], icon, primaryAudio}'
+wji fs get <datasetId> --json --detail full | jq '{images: [.images[]? | {id, type}], icon, primaryAudio}'
 wji fs set <datasetId> --add-images <headshot> <sheet> <voice> --icon <headshot> --set primaryAudio=<voice>
 ```
 
@@ -549,6 +574,9 @@ wji org list && wji org use <orgId>
 - **Lists are paginated.** `--limit` is capped at 100; `--limit -1` returns at
   most 500 on most endpoints. Page with `--page` instead of assuming you got
   everything.
+- **Lists are trimmed.** `--json` is the `standard` level; read `content`,
+  `form` or `settings` with `--detail full`, and never write back a `content`
+  that carries `truncated: true`.
 - **Report failures honestly.** A failed generation surfaces as a non-zero exit
   with the provider's message — pass it on rather than retrying silently.
 

package/skill/wji/reference/commands.md

@@ -10,7 +10,8 @@ Global flags (accepted before any subcommand):
 | `--org <orgId>` | organization context (`X-Org-Id`) |
 | `--no-org` | force personal context for this call |
 | `--locale <locale>` | response language: `cht`, `chs`, `en` |
-| `--json` | print the raw API response instead of tables |
+| `--json` | print JSON instead of tables (`standard` detail unless `--detail` is given) |
+| `--detail <level>` | `brief` / `standard` / `full` for list and get reads; `full` is the raw API response. Env `WJI_DETAIL` |
 
 Value conventions: `@file` reads a file, `-` reads stdin, `--set key=value`
 repeats and nests on dots, references are `assetId[:label[:weight]]`.
@@ -64,8 +65,8 @@ switch endpoint.
 
 | Command | API |
 | --- | --- |
-| `ws list [-k keyword] [-s all\|me\|shared\|favorite] [--page] [--limit]` | `GET /workspace/search` — `--limit` max 100, `-1` returns at most 500 |
-| `ws get [id]` | `GET /workspace/list/:id` |
+| `ws list [-k keyword] [-s all\|me\|shared\|favorite] [--page] [--limit]` | `GET /workspace/search` — `--limit` max 100, `-1` returns at most 500; sends `detail` |
+| `ws get [id]` | `GET /workspace/list/:id` — sends `detail` (`settings` and full members need `--detail full`) |
 | `ws use <id>` | stores the default workspace locally |
 | `ws create <name> [-d] [-r ratio] [-l lang] [-t type] [-a app...] [--no-apps]` | `POST /workspace/create` — without `-a` the workspace gets the default template `dataset datasetimagegeneration canvas`; `--no-apps` creates it empty (the API itself adds nothing) |
 | `ws update [id] [-n] [-d] [-r] [-l] [--set]` | `PUT /workspace/list/:id` |
@@ -96,10 +97,10 @@ Restriction from the backend: `Dataset`, `Draw`, `DatasetImageGeneration` and
 
 | Command | API |
 | --- | --- |
-| `fs root [key] [-d depth]` | `GET /fsnode/root[/:key]` — key defaults to your drive; pass a workspace id for its root |
-| `fs ls <id> [--limit] [--scope]` | `GET /fsnode/list/:id/children` |
-| `fs tree <id> [-d depth]` | `GET /fsnode/list/:id/tree` |
-| `fs get <id> [--scope]` | `GET /fsnode/list/:id` (comma-separated ids allowed) |
+| `fs root [key] [-d depth]` | `GET /fsnode/root[/:key]` — key defaults to your drive; pass a workspace id for its root; sends `detail` |
+| `fs ls <id> [--limit] [--scope]` | `GET /fsnode/list/:id/children` — sends `detail`; at `standard` each `content` is a summary |
+| `fs tree <id> [-d depth]` | `GET /fsnode/list/:id/tree` — sends `detail` |
+| `fs get <id> [--scope]` | `GET /fsnode/list/:id` (comma-separated ids allowed) — sends `detail`; `--detail full` for the whole `content` |
 | `fs path <id>` | `GET /fsnode/list/:id/path` |
 | `fs mkdir <parent> <name> [--nickname] [--content]` | `POST /fsnode/create` type=folder |
 | `fs touch <parent> <name> [--mimetype] [--asset] [--content]` | `POST /fsnode/create` type=file |
@@ -118,7 +119,7 @@ characters, nicknames at 200, and `fs batch` at 100 items per call.
 | Command | Notes |
 | --- | --- |
 | `dataset new <parent> <name> [-t type] [-d description] [--images] [--nickname]` | creates an fsnode with mimetype `workspace/dataset` |
-| `dataset list <folderId>` | dataset nodes in a folder |
+| `dataset list <folderId>` | dataset nodes in a folder (sends `detail`) |
 | `dataset get <key>` | `GET /dataset/list/:key` (vector-store record) |
 | `dataset rm <key> [-y]` | `DELETE /dataset/list/:key` |
 
@@ -150,6 +151,11 @@ Model parameters worth knowing (`--set`, see `wji models get`):
 `gpt-image-2.5-*` `background=transparent` returns a PNG with alpha;
 `bytedance-seedream-5-pro` `layer_decomposition=true` splits one reference
 image into separate PNG layer assets (exactly 1 reference, `size=1K|2K`).
+`face-swap-uncensored` / `head-swap-uncensored` take exactly 2 references in
+order (target picture, then the face / head); `qwen-image-edit-uncensored`
+edits the first reference; `z-image-*-uncensored` `long_edge=<px>`;
+`wan-animate-uncensored` `mode=animate|replace`; `flashvsr-uncensored`
+`resolution=720p|1080p|2k|4k`.
 
 Uploads of `.html`, `.xhtml`, `.svg`, `.js` and `.xml` are stored as
 `application/octet-stream` with `Content-Disposition: attachment`.
@@ -171,7 +177,7 @@ updates live.
 | `canvas copy [nodeIds...] --to <canvasIds...> [-c] [--selected] [--at x,y] [--same-position] [-y]` | the canvas's own copy-paste: new nodes (no position, `generate.historyId/pendingAssetId` dropped, finished asset shared via `file`), edges among them re-created oldest first so `@image1…` numbering holds, color tags kept and a missing label carried to the target. Edges from nodes outside the copy are left out. Default placement: right of what is already on the target |
 | `canvas nodes [-c] [-k kind] [-n regex] [--tag tags…\|--untagged]` | `GET /fsnode/list/:canvasId/children?limit=-1` + `POST /assets/check-status`. `-n` matches the node name as a case-insensitive regex (`/re/flags` to set flags yourself). `--tag` takes a label (`C01`, case-insensitive), a color (`red`) or a color id (`rose`); several values match any. Filters combine (AND) |
 | `canvas tags [-c] [-a]` | the canvas's color tags: color, label, node count (`-a` also unused, unlabelled colors) |
-| `canvas get <nodeId> [-c]` | same load; one node with references (`@image1 ← id`) and frame slots |
+| `canvas get <nodeId> [-c]` | same load; one node with references (`@image1 ← id`) and frame slots. Canvas commands always load the full canvas; `--detail` trims only the JSON they print (`standard` clips `text` / `generate.prompt` and drops `generate.values`) |
 | `canvas edges [-c] [--node]` | the `canvas/edge` children |
 | `canvas add <kind> [-c] [-n] [-t\|--text-file] [--plain] [--size w,h] [--asset] [--dataset] [-p\|--prompt-file] [-m] [-r] [--resolution] [-d] [--set] [--from ids…] [--at x,y]` | `POST /fsnode/create` (`canvas/node/<kind>`), then one `canvas/edge` per `--from`. Kinds: text, title, comment, image, video, audio, dataset. `add dataset --dataset <fsnodeId>` writes `content {datasetId, datasetName, datasetType, datasetParentName, description}` from the dataset node (must be `workspace/dataset`); it takes no `--from`. Text kinds are Markdown on the web; `--plain` turns every single line break into a Markdown hard break (`"  \n"`, fenced code untouched), and without it the CLI warns when a single break would be joined. `--size` writes `content.width/height` |
 | `canvas set <nodeId> [-c] [-n] [-t\|--text-file] [--plain] [--move x,y] [--size w,h] [-p] [-m] [-r] [--resolution] [-d] [--set]` | read → merge → `PATCH /fsnode/list/:id` with `contentPatch` of only the top-level keys that changed, so a concurrent browser edit of another key survives (falls back to a whole-content write on a backend without `contentPatch`) |
@@ -320,7 +326,7 @@ parameters are range-checked against the model's `min`/`max`.
 ## history
 
 `history list [-w] [--mode image|video|audio] [--fsnode] [--scope] [--scope-id] [--tags] [--bookmarked]` →
-`GET /ai/assets/history/search`
+`GET /ai/assets/history/search` (sends `detail`; `standard` clips `prompt` and summarises `form`)
 
 `history get <id> [--wait] [--timeout] [--interval]` ·
 `history bookmark <id> [--off]` · `history rm <id> [-y]` ·
@@ -329,12 +335,15 @@ parameters are range-checked against the model's `min`/`max`.
 ## models
 
 `models list [-f feature...] [-w] [-k keyword] [--all]` →
-`GET /ai/model/list/:features`
+`GET /ai/model/list/:features` (sends `detail`; at `standard` the `form` keeps
+only ratios, batch, reference caps and `other` options — `--detail full` for the
+whole schema)
 
 The endpoint matches with `features.some(...)`, so an empty feature list returns
 nothing; with no `-f` the CLI asks for every feature instead.
 
-The response is a curated view — no `id`, `rate`, `tags` or `protocol` — and it
+The response is a curated view — no `id`, `rate` or `protocol` (`tags` is
+included; uncensored models carry `uncensored`) — and it
 is de-duplicated by **`nickname`**, which is therefore the unique selector.
 `name` is not unique (`seedance-2.0` exists twice under different providers).
 Pass the `MODEL` column (the nickname) to `-m`.
@@ -467,6 +476,12 @@ Pass `null` to clear a nullable field. Password minimum is 6 characters.
   asset carrying `error` is also a failure.
 - Some GETs are cached per user for 10s, keyed on the full URL including the
   query string.
+- Workspace, fsnode, model and history reads accept `?detail=brief|standard|full`.
+  The API defaults to `full` (the website relies on it); the CLI sends the
+  `--detail` value, or `standard` under `--json`. `standard` clips long text to
+  200 chars, turns media into `{id, url, thumbnail, status, …}` and reduces an
+  fsnode `content` to a summary carrying `truncated: true`; never PATCH such a
+  summary back.
 - Paginated endpoints return `{ data, pagination: { total, page, limit, next,
   hasNext, hasPrev, totalPages } }`. An explicit `limit` is capped at **100**.
   `limit: -1` returns everything only on endpoints that opt in (`fs ls` does);