@mx-space/cli 0.13.1 → 0.13.2

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "@mx-space/cli",
-  "version": "0.13.1",
+  "version": "0.13.2",
   "description": "Command line interface for mx-space (mx-core) — auth, content, configuration",
   "keywords": [
     "mx-space",
@@ -48,33 +48,33 @@
     "access": "public"
   },
   "dependencies": {
-    "@clack/prompts": "^1.5.1",
+    "@clack/prompts": "^1.6.0",
     "@effect/cli": "0.75.2",
-    "@effect/platform": "0.96.1",
+    "@effect/platform": "0.96.2",
     "@effect/platform-node": "0.107.0",
-    "@haklex/rich-headless": "0.27.0",
-    "@haklex/rich-litexml": "0.27.0",
-    "@haklex/rich-litexml-cli": "0.27.0",
-    "@lexical/headless": "^0.45.0",
-    "better-auth": "^1.6.16",
-    "effect": "3.21.3",
-    "fast-xml-parser": "5.8.0",
+    "@haklex/rich-headless": "0.29.1",
+    "@haklex/rich-litexml": "0.29.0",
+    "@haklex/rich-litexml-cli": "0.29.0",
+    "@lexical/headless": "^0.46.0",
+    "better-auth": "^1.6.20",
+    "effect": "3.21.4",
+    "fast-xml-parser": "5.9.3",
     "get-east-asian-width": "^1.6.0",
-    "lexical": "^0.45.0",
+    "lexical": "^0.46.0",
     "minisearch": "^7.2.0",
     "open": "^11.0.0",
-    "semver": "7.8.4",
-    "@mx-space/api-client": "5.3.1",
-    "@mx-space/editor": "0.1.0"
+    "semver": "7.8.5",
+    "@mx-space/api-client": "5.4.0",
+    "@mx-space/editor": "0.2.0"
   },
   "devDependencies": {
     "@effect/vitest": "0.29.0",
-    "@types/node": "25.9.3",
+    "@types/node": "26.0.0",
     "@types/semver": "7.7.1",
-    "tsdown": "0.22.2",
+    "tsdown": "0.22.3",
     "tsx": "^4.22.4",
     "typescript": "6.0.3",
-    "vitest": "4.1.8"
+    "vitest": "4.1.9"
   },
   "scripts": {
     "package": "rm -rf dist && tsdown",

package/skills/commands-ai.md

@@ -0,0 +1,115 @@
+---
+slug: commands-ai
+title: AI commands
+description: mxs ai — manage AI summary / translation / insights artifacts
+order: 45
+---
+
+# AI commands
+
+`mxs ai` manages every AI artifact the core server produces: per-article **summary**, per-article **translation**, per-article **insights**, plus the i18n **translation entries** dictionary used for category / topic / note metadata.
+
+Article references (`<id>`) accept a Snowflake id, a post slug, or a numeric note nid. The CLI resolves through post first, then note. Record references (`<recordId>`) are raw Snowflakes — they are NOT resolved.
+
+## Generate verbs (async)
+
+`summary regen`, `translate run`, `insights refresh` enqueue a task and poll `GET /tasks/:id` until terminal (`succeeded` / `failed` / `cancelled`, or the spec-2 `completed` / `partial_failed` aliases). Progress lines write to stderr. `--no-wait` returns immediately with `status: pending`. If the server reports `created: false`, an in-flight deduplicated task already exists; the CLI joins it instead of erroring. Polling cadence is 1000 ms by default; override with `MXS_AI_POLL_MS=<ms>` for tests.
+
+| Command                                              | Purpose                                                         |
+| ---------------------------------------------------- | --------------------------------------------------------------- |
+| `mxs ai summary regen <id> [--to <lang>...]`         | Regenerate an article's AI summary.                             |
+| `mxs ai translate run <id> --to <lang>...`           | Translate an article into one or more languages (`--to` required, repeatable). |
+| `mxs ai insights refresh <id> [--to <lang>...]`      | Refresh AI insights for an article.                             |
+
+Common flags on the three generate verbs:
+
+| Flag             | Effect                                                             |
+| ---------------- | ------------------------------------------------------------------ |
+| `--to <lang>`    | Target language code (repeatable). Forwarded as `targetLanguages`. |
+| `--no-wait`      | Return after task creation; print `status: pending` and exit 0.    |
+
+On success the CLI emits an `ai-task` view: `taskId`, `status`, `refId`, `targetLanguages`, `totalTokens`, `totalCost` (cents), `resultIds`.
+
+## Summary read / manage
+
+| Command                                                       | Purpose                                                       |
+| ------------------------------------------------------------- | ------------------------------------------------------------- |
+| `mxs ai summary list [--page <n>] [--size <n>] [--grouped]`   | List summaries (flat or grouped by article).                  |
+| `mxs ai summary get <recordId>`                               | Get one summary by record id.                                 |
+| `mxs ai summary by-article <id> [--lang <l>] [--only-db]`     | Show an article's summary; `--only-db` skips auto-generation. |
+| `mxs ai summary edit <recordId>`                              | Edit the `summary` field via `$EDITOR` (JSON envelope).       |
+| `mxs ai summary delete <recordId> [--force]`                  | Delete a summary record. `--force` required in non-TTY.       |
+
+## Translation read / manage
+
+There is no flat list endpoint for translations server-side. `mxs ai translate list` is always grouped by article.
+
+| Command                                                  | Purpose                                                  |
+| -------------------------------------------------------- | -------------------------------------------------------- |
+| `mxs ai translate list [--page <n>] [--size <n>]`        | List translations (grouped by article).                  |
+| `mxs ai translate get <recordId>`                        | Get one translation by record id.                        |
+| `mxs ai translate by-article <id> [--lang <l>]`          | Show an article's translations (single lang or all).     |
+| `mxs ai translate languages <id>`                        | List languages an article has been translated into.      |
+| `mxs ai translate edit <recordId>`                       | Edit translation fields via `$EDITOR` (JSON envelope).   |
+| `mxs ai translate delete <recordId> [--force]`           | Delete a translation record.                             |
+
+Edit envelope keys: `title`, `text`, `subtitle` (nullable), `summary`, `tags`, `content`. Only included keys are PATCHed.
+
+## Translation entries (i18n dictionary)
+
+The dictionary layer used for category / topic / note metadata translations. `--key-path` accepts ONLY the server-validated set: `category.name`, `topic.name`, `topic.introduce`, `topic.description`, `note.mood`, `note.weather`.
+
+| Command                                                                                     | Purpose                                                |
+| ------------------------------------------------------------------------------------------- | ------------------------------------------------------ |
+| `mxs ai translate entries list [--page <n>] [--size <n>] [--key-path <p>] [--lang <l>]`     | List entries with optional filters.                    |
+| `mxs ai translate entries generate [--key-path <p>...] [--to <lang>...]`                    | Regenerate entries (synchronous on the server).        |
+| `mxs ai translate entries edit <recordId>`                                                  | Edit `translatedText` via `$EDITOR` (JSON envelope).   |
+| `mxs ai translate entries delete <recordId> [--force]`                                      | Delete an entry.                                       |
+
+## Insights read / manage
+
+| Command                                                       | Purpose                                                        |
+| ------------------------------------------------------------- | -------------------------------------------------------------- |
+| `mxs ai insights list [--page <n>] [--size <n>] [--grouped]`  | List insights (flat or grouped by article).                    |
+| `mxs ai insights get <recordId>`                              | Get one insights record by record id.                          |
+| `mxs ai insights by-article <id> [--lang <l>] [--only-db]`    | Show an article's insights; `--only-db` skips auto-generation. |
+| `mxs ai insights edit <recordId>`                             | Edit the `content` field via `$EDITOR` (JSON envelope).        |
+| `mxs ai insights delete <recordId> [--force]`                 | Delete an insights record.                                     |
+
+## Exit codes specific to `ai`
+
+| Code   | When                                                                                     |
+| ------ | ---------------------------------------------------------------------------------------- |
+| `0`    | Task reached `succeeded`/`completed`; or `--no-wait`; or read/edit/delete completed.     |
+| `1`    | `AiTaskCreateFailed` (no `taskId` in response) or `AiTaskFailed` (terminal `failed`/`cancelled`/`partial_failed`). |
+| `7`    | `AiRecordNotFound` (`get`/`edit`/`delete` against a missing record).                     |
+
+Network / auth / validation errors map per the global table — see `safety`.
+
+## Examples
+
+```bash
+# Regenerate an article's summary and wait for it
+mxs ai summary regen my-post --to en --to ja
+
+# Translate to Japanese, fire and forget
+mxs ai translate run my-post --to ja --no-wait
+
+# Refresh insights, then read the result
+mxs ai insights refresh my-post
+mxs ai insights by-article my-post --only-db
+
+# Inventory
+mxs ai summary list --grouped --output json | jq .
+mxs ai translate languages my-post
+
+# Patch a translation record through $EDITOR
+mxs ai translate edit 01HXXX...
+```
+
+## Notes for agents
+
+- Always pass `--output json` (or `--json`) for machine-parseable output. The default `readable` mode is for humans.
+- `created: false` is a normal join-existing path, NOT an error. Do not retry on it.
+- The polling loop has no hard timeout. For very long translation runs across many languages, prefer `--no-wait` plus periodic `mxs ai translate by-article` polls.
+- Edit envelopes are typed per resource. Unknown keys are rejected client-side before any PATCH is sent.

package/skills/overview.md

@@ -25,6 +25,7 @@ The audience is **AI agents**. Pass `--output llm` for raw markdown suitable for
 | Categories                                    | `commands-category`              |
 | Topics                                        | `commands-topic`                 |
 | Snippets (server data / functions)            | `commands-snippet`               |
+| AI summary / translation / insights / entries | `commands-ai`                    |
 | Server-side options                           | `commands-config`                |
 | Authentication                                | `commands-auth`                  |
 | Local profile management                      | `commands-profile`               |