mammoth-cli 2.0.69__tar.gz → 2.0.71__tar.gz

{mammoth_cli-2.0.69 → mammoth_cli-2.0.71}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: mammoth-cli
-Version: 2.0.69
+Version: 2.0.71
 Summary: Command-line interface for the Mammoth Analytics platform
 Project-URL: homepage, https://mammoth.io
 Project-URL: repository, https://github.com/EdgeMetric/mammothsdk

{mammoth_cli-2.0.69 → mammoth_cli-2.0.71}/mammoth_cli/__init__.py

@@ -6,7 +6,7 @@ not implement a second HTTP client and does not call private SDK members.
 
 from __future__ import annotations
 
-__version__ = "2.0.69"
+__version__ = "2.0.71"
 
 # The versioned machine output envelope contract.
 SCHEMA_VERSION = 1

{mammoth_cli-2.0.69 → mammoth_cli-2.0.71}/mammoth_cli/bundled_skill/mammoth-cli/SKILL.md

@@ -1,6 +1,6 @@
 ---
 name: mammoth-cli
-version: 2.0.69
+version: 2.0.71
 description: "Use Mammoth Analytics from a terminal: install or authenticate the CLI, discover its live command contract, and safely manage projects, data, views, pipelines (join, merge, pivot, filter, clean), dashboards, exports, and handoffs."
 ---
 
@@ -81,7 +81,7 @@ mammoth view transform --help              # every data transformation
 | Change a column's type | `view transform convert-type` |
 | Rename a column, sort the rows (view settings, not tasks) | `view transform rename-columns`, `view transform sort` |
 | Rank, running total, previous row | `view transform window` |
-| A dashboard | `dashboard create-blank` ([dashboards](references/recipes/dashboards.md)) |
+| A dashboard | `dashboard v3 generate`, then `chat edit` ([dashboards](references/recipes/dashboards.md)) |
 | Deliver the rows | `view export csv`, `view export postgres` (and other destinations), `view export dataset` |
 | Run something on a schedule or on new data (refresh, append, alert) | `automation create`: a condition (`at_specific_time`, new file in a folder, ...) and tasks (`run_data_retrieval`, `append_data`, `send_an_alert`, `pull_cloud_files`). `schedule create` only pulls a connector's data. A dataset made from an uploaded file has no source to refresh from; say so. Full recipe, including a known `automation get` backend caveat: [recurring work](references/recipes/scheduling.md) |
 
@@ -122,11 +122,10 @@ dashboard from t_a and t_b"). Work it out from the data before you build:
 6. If the data has money, add it before you make the dashboard: `revenue`
    (`math`, `qty * price`, `new_column`), the `fix` in `before_dashboard`.
    A dashboard sees only the columns the view had when it was made.
-7. Build the dashboard from the joined view
-   ([dashboards](references/recipes/dashboards.md)) and chart the sum of
-   revenue. Do not sum a unit price. `create-blank`, `canvas save` and
-   `pages add` return `deliverable_check`: fix each warning, or say in the
-   report why not.
+7. Build the dashboard from the joined view with one sentence
+   ([dashboards](references/recipes/dashboards.md)) that charts the sum of
+   revenue. Do not sum a unit price. Fix each `deliverable_check` warning,
+   or say in the report why not.
 
 For the defaults that most pipelines use (new column or overwrite, `LEFT`
 joins, date steps), see

{mammoth_cli-2.0.69 → mammoth_cli-2.0.71}/mammoth_cli/bundled_skill/mammoth-cli/references/recipes/dashboards.md

@@ -2,18 +2,23 @@
 
 `dashboard create` (the legacy AI-generation engine) does not exist as a
 command: it has no handler in current apiv2 (always 404s; historically HTTP
-409 `4DASH012 DASHBOARD_LEGACY_CREATION_RETIRED`). Create with `create-blank`
-(or `v3 generate` when the task allows an AI route).
+409 `4DASH012 DASHBOARD_LEGACY_CREATION_RETIRED`). Build a board from one
+sentence with `v3 generate`, and change it with `chat edit`; author a canvas by
+hand (`create-blank`, `canvas save`, `pages add`) only when asked to.
 `dashboard source list` is an observed blocker on release (see
 capabilities); verify the view binding with `dashboard get DASHBOARD_ID`
 (`data.dataview_id`) instead.
 
 ```bash
-mammoth schema get dashboard.create-blank
-mammoth dashboard create-blank --input INPUT_JSON --project PROJECT_ID
+mammoth dashboard v3 generate --input '{"body": {"params": {"intent": "Monthly revenue by region, with top customers", "dataview_id": VIEW_ID}}}'
+mammoth dashboard chat edit DASHBOARD_ID --input '{"body": {"params": {"prompt": "Add a chart of orders by status"}}}'
 mammoth dashboard get DASHBOARD_ID
 ```
 
+`v3 generate` waits for the bake (about 30 s) and returns the baked canvas;
+`chat edit` returns `changed` and a one-line `message`. Quote one number from
+the board before you report it done.
+
 Discover page/widget/publish routes and verify the binding, draft/published
 data and terminal jobs. Use only returned IDs and schema confirmation policy;
 do not invent dashboard JSON or opaque task specs.

{mammoth_cli-2.0.69 → mammoth_cli-2.0.71}/mammoth_cli/commands/schema.py

@@ -281,6 +281,14 @@ _COMMAND_DISCOVERY_PURPOSES = {
         "url app post posts events rows straight in"
     ),
     "dataset.create-from-pdf": "pdf table document extract get the table out of into a dataset",
+    "dashboard.v3.generate": (
+        "create build make new generate dashboard board report from a view description "
+        "intent sentence ai combined showing"
+    ),
+    "dashboard.chat.edit": (
+        "add change edit update chart charts kpi to an existing dashboard board "
+        "by description sentence ai show instead"
+    ),
     "dashboard.pdf.export": "pdf of a board dashboard download print meeting export",
     "view.checkpoint.create": (
         "stop halt pause pipeline alert notify flag when rows match a rule condition checkpoint"
@@ -339,6 +347,8 @@ _SCOPE_REQUIREMENTS: dict[str, dict[str, Any]] = {
 }
 
 _MAX_FIND_RESULTS = 20
+# Outranks any word-overlap score: a command named by its full path comes first.
+_NAMED_COMMAND_BOOST = 10_000
 _MAX_FIND_LIMIT = 100
 # How many of a find's top matches carry inline accepted_fields/agent_example.
 # 35% of all eval tool calls were command discovery (schema find -> schema
@@ -1471,7 +1481,13 @@ def find_schemas(
                 "support.* commands act on another workspace as an operator, not the "
                 "caller's own; ordinary workspace work uses the non-support command."
             )
-        if len(matched_terms) == len(terms):
+        # A query that spells out this command's whole path (``aggregate view
+        # data ...``) asks for it by name, whatever goal words ride along.
+        path_tokens = set(_tokens(command_path))
+        named = len(path_tokens) >= 3 and path_tokens <= set(terms)
+        if named:
+            score += _NAMED_COMMAND_BOOST
+        if named or len(matched_terms) == len(terms):
             ranked_matches.append((score, entry))
         else:
             near_misses.append((len(matched_terms), score, entry, matched_terms))

{mammoth_cli-2.0.69 → mammoth_cli-2.0.71}/mammoth_cli/runtime/strict.py

@@ -279,13 +279,23 @@ _UNKNOWN_FIELD_HINTS: dict[str, str] = {
         "from 'mammoth dashboard canvas get DASHBOARD_ID', edited. An --input file "
         "takes the same wrapped document."
     ),
-    "dataset.list": (
-        "project_id is not an --input field; pass the project with the global "
-        "--project option (or omit it to use the active project)."
-    ),
 }
 
 
+def _unknown_field_hint(command_id: str, document: dict[str, Any], accepted: list[str]) -> str:
+    """Name the fix for an unknown ``--input`` key, with the caller's own value."""
+    if command_id in _UNKNOWN_FIELD_HINTS:
+        return _UNKNOWN_FIELD_HINTS[command_id]
+    if "project_id" in document and "project_id" not in accepted:
+        return (
+            "project_id is not an --input field; put --project "
+            f"{document['project_id']} after the command: "
+            f"mammoth {command_id.replace('.', ' ')} ... --project {document['project_id']} "
+            "(or omit it to use the active project)."
+        )
+    return f"Accepted fields: {', '.join(accepted) or '(none)' }."
+
+
 def validate_input_fields(command_id: str, document: dict[str, Any] | None) -> None:
     """Reject unknown ``--input`` keys, then coerce known keys to their type.
 
@@ -334,9 +344,7 @@ def validate_input_fields(command_id: str, document: dict[str, Any] | None) -> N
                 f"{', '.join(unknown)}."
             ),
             exit_status=EXIT_USAGE,
-            hint=_UNKNOWN_FIELD_HINTS.get(
-                command_id, f"Accepted fields: {', '.join(sorted(fields_by_name)) or '(none)' }."
-            ),
+            hint=_unknown_field_hint(command_id, document, sorted(fields_by_name)),
             details={"unknown": unknown, "accepted": sorted(fields_by_name)},
         )
     if is_closed_zero_input(command_id) and document == {}:

{mammoth_cli-2.0.69 → mammoth_cli-2.0.71}/pyproject.toml

@@ -1,6 +1,6 @@
 [project]
 name = "mammoth-cli"
-version = "2.0.69"
+version = "2.0.71"
 description = "Command-line interface for the Mammoth Analytics platform"
 authors = [{name = "Mammoth Analytics", email = "support@mammoth.io"}]
 readme = "README.md"

{mammoth_cli-2.0.69 → mammoth_cli-2.0.71}/spec/manifests/commands/dashboard.yaml

@@ -490,7 +490,7 @@ commands:
   - CT-DASHBOARD-CHAT-EDIT-ERROR
   draft_test: null
   undo_test: null
-  known_restrictions: null
+  known_restrictions: 'Changes an existing board from one sentence (add, remove or change a chart, KPI or page): {"body": {"params": {"prompt": "<the change>"}}}; returns changed, a one-line message and the rebaked canvas. Use it rather than dashboard pages add or canvas save.'
   reviewed_by: primary
 - command_id: dashboard.chat.history
   command_path: dashboard chat history
@@ -749,7 +749,7 @@ commands:
   - CT-DASHBOARD-CREATE-BLANK-ERROR
   draft_test: null
   undo_test: null
-  known_restrictions: 'Input is {"params": {"dataview_id": VIEW_ID, "title": ...}} with --yes; returns {id, sequence}. The blank canvas has one page (p1) and no figures: author it with dashboard canvas get -> edit -> dashboard canvas save or dashboard pages add. One dataview per dashboard.'
+  known_restrictions: 'To build a board from a description, use dashboard v3 generate instead. Input is {"params": {"dataview_id": VIEW_ID, "title": ...}} with --yes; returns {id, sequence}. The blank canvas has one page (p1) and no figures: author it with dashboard chat edit, or by hand with dashboard canvas get -> edit -> dashboard canvas save or dashboard pages add. One dataview per dashboard.'
   reviewed_by: primary
 - command_id: dashboard.data.draft
   command_path: dashboard data draft
@@ -1673,7 +1673,7 @@ commands:
   - CT-DASHBOARD-PAGES-ADD-ERROR
   draft_test: null
   undo_test: null
-  known_restrictions: 'Requires --yes --confirm DASHBOARD_ID. body.params.pages is a list of PageAdd objects: {title, focus: {measure, dim, kpis: [...]}, charts: [{kind, title, dim, measure, agg}]}; the schema example''''s {} entry is not a usable page. The route runs through the LLM control guard: in the ILG simulation it dropped unit.prefix on a money card ("nothing in this data says which currency") that dashboard canvas save had kept, and reports that in data.message.'
+  known_restrictions: 'To add or change content by description, use dashboard chat edit instead. Requires --yes --confirm DASHBOARD_ID. body.params.pages is a list of PageAdd objects: {title, focus: {measure, dim, kpis: [...]}, charts: [{kind, title, dim, measure, agg}]}; the schema example''''s {} entry is not a usable page. The route runs through the LLM control guard: in the ILG simulation it dropped unit.prefix on a money card ("nothing in this data says which currency") that dashboard canvas save had kept, and reports that in data.message.'
   reviewed_by: primary
 - command_id: dashboard.pdf-artifact
   command_path: dashboard pdf-artifact
@@ -4397,7 +4397,7 @@ commands:
   - CT-DASHBOARD-V3-GENERATE-ERROR
   draft_test: null
   undo_test: null
-  known_restrictions: null
+  known_restrictions: 'This is how to make a dashboard: it builds and bakes a whole board from one sentence on one view. Input {"body": {"params": {"intent": "<what the board should show>", "dataview_id": VIEW_ID, "title": "<optional name>"}}}; waits (about 30 s) and returns the baked canvas. Change it afterwards with dashboard chat edit. dashboard create-blank is only for authoring a canvas by hand.'
   reviewed_by: primary
 - command_id: dashboard.video-state
   command_path: dashboard video-state