--- title: "Add a widget to a section" method: POST path: "/widgets/create" tags: ["Widgets"] --- # Add a widget to a section `POST /widgets/create` ## Request - Base URL: `https://api.reportingninja.com/v1` - URL: `https://api.reportingninja.com/v1/widgets/create` - Auth: HTTP bearer Writes one widget into a section of a report. The widget is defined in the widget DSL, the same vocabulary /v1/reports/get answers for the widgets it can express. Every property of the dsl schema is required, `unfulfilled` included: send null in the optional ones and an empty array in `unfulfilled`. That field belongs to the AI generator, which lists there what it could not express; a client writing the DSL itself sends an empty array. The `data_view` of a dsl takes its values from /v1/capabilities (the integrations block), which is not always the data_views list of /v1/integrations/detail: an integration that reads a single level when it is queried can still be drawn at several levels in a report. A refusal names the accepted values in the problem's `suggestions`. The integration must already have a data source in the report, and integration_id may be omitted only for a dsl whose `visualization` is `text`: a text block is prose and reads no data. A text block written into a TEMPLATE is cloned into the reports based on it, like any other widget of a template. A section holds at most 12 widgets. Without placement the widget is placed automatically. This endpoint accepts if_version and verify_data, and never asks for a confirmation, so it never answers CONFIRMATION_EXPIRED or CONFIRMATION_MISMATCH. With verify_data in a dry run the widget IS run against the provider and the verdict travels in data.verification, but the answer is still valid: true -- a rehearsal reports what it found and refuses nothing; the 422 belongs to the real write. Two callers writing into the same section at once are serialised: the second one gets 409 SECTION_BUSY with a Retry-After. An AI summary is written with ai_summary in place of the dsl, and the two are mutually exclusive: it is refused on a TEMPLATE (APPEND_NOT_ALLOWED reason template_report), it never carries a connection, and an empty widget_keys or section_keys is refused -- leave the key out to have the platform choose what it reads. Refusals: 400 VALIDATION_FAILED (including the problem codes DSL_INVALID and TRANSLATION_FAILED), UNKNOWN_PARAM, APPEND_NOT_ALLOWED or INVALID_VALUE for a placement that does not fit, 403 WRITE_ACCESS_DISABLED, 404 NOT_FOUND, 409 VERSION_MISMATCH or SECTION_BUSY plus the idempotency conflicts, 422 WIDGET_EXECUTION_FAILED when verify_data ran the widget and the provider refused it. Creating a widget does not move the version of the report: that number covers the list of sections. Every write endpoint reads dry_run and idempotency_key; see the Write guarantees section of the introduction. A COMPARATOR is written with the `compare` block of the dsl: `by` accounts draws one series per connected account and names at least two of them in `items`, so such a widget carries no `connection_key` of its own -- which is also how it is written into a report that binds a connection per widget, where the items ARE the connections of the widget and a body sending the pair as well is refused with 400 INVALID_VALUE naming connection_key; `by` filters draws one per saved filter of the account, reads one connection like any ordinary widget and names it in the body; `by` breakdown splits one page metric of facebook_insights into a series per value. Only a graph_basic, a graph_pie, a graph_trend and a table_metrics_summary take one, and which classes and which axes an integration offers depends on the entity level of the data view as well as on the integration. The same account or the same saved filter named twice is refused, and a `label` cannot carry the sequence #%#, which is what the parts of a stored account are joined with. A comparator of connected accounts is not written into a report whose report_level_data_sources is true, which answers 409 NOT_ALLOWED with details.reason report_level_data_sources. ## Request body - WidgetsCreateRequest: Which section to add a widget to, and what the widget shows: a definition, or an AI summary. - `report_key` string, required: The report the section belongs to. - `section_key` string, required: The section the widget is added to. - `integration_id` string: Which integration the widget reads from. The report must already have a data source for it. Required for every widget except one whose dsl has `visualization: text`, which reads no data: OMIT the property for a text block, do not send null. Sending a real integration there is accepted and checked the same way, and never read. - `dsl` WidgetDsl - `name` string, required: Title of the widget, in the requested output language. Plain text, no markup. At most 80 characters. - `visualization` 'kpi' | 'graph_basic' | 'graph_pie' | 'graph_trend' | 'table_metrics_summary' | 'table_dimensions_metrics' | 'text', required: How the widget is drawn. The first six read data; text is a block of prose with no data behind it, and it is the only one that carries content. - `chart_type` 'area' | 'line' | 'column' | 'bar', nullable, required: Which chart a graph_basic (area, line, column or bar) or a graph_trend (area, line or column; a trend draws no bars) is drawn as. Null takes the default of the shape, which is a column chart for a graph_basic and an area chart for a graph_trend, and is the usual answer. A chart drawn along a single time dimension is built as a trend and takes the charts of a trend, so it never carries bar. The four shapes that draw no chart always carry null. - `full_width` boolean, nullable, required: True when the widget takes a row of its own; null, like false, pairs it up. The layout of a section is yours: the widgets you do not mark are laid out two to a row in the order you write them, and the KPIs go four to a row before them. A widget marked true closes the row being filled and the next widget opens a new one. Ignored in a kpi, which is always part of a strip of four. - `content` string, nullable, required: The prose of a text block, and null in every other shape. Markdown of a reduced set and nothing else: blank line between paragraphs, **bold**, lines starting with '- ' for a bullet list and with '1. ' for a numbered one. No headings, no links, no italics, no images, no HTML and no emoji; anything else is printed as the characters it is made of. At most 2000 characters. A text block carries no figures: it is written before the widgets of the report are run, so any number in it would be one nobody measured. - `data_view` string, required: Public id of the data view the widget reads, exactly as listed in the catalogue. Empty string when the integration has no data views, and empty string in a text block, which reads none. - `metrics` union[], required: Public field ids of the metrics, in the order they are shown. At least one in the six shapes that read data; empty in a text block. A column of a table_metrics_summary may also be written as the object {id, filter_key}, which reads that one column through a saved filter of the account. Set by API and MCP clients; a definition being generated always writes the field id on its own. - union - string - object - `id` string, required: Public field id of the metric, the same one the short form carries. - `filter_key` string, nullable, required: Key of a saved filter of the account this column is read through, or null when the column is read like every other one. Only a table_metrics_summary takes it, and only on integrations whose columns carry a filter of their own. Set by API and MCP clients; a definition being generated always writes it as null. - `dimensions` string[], required: Public field ids of the dimensions, in the order they are shown. Empty when the widget has none. - `filter_key` string, nullable, required: Key of a saved filter of the account the whole widget is read through, or null when the widget reads no such filter. It excludes filters: a widget either points at a filter of the account or carries conditions of its own, never both. The filter has to belong to the integration of the widget and, where the filters of that integration are read in a data view, to the data view of the widget. Set by API and MCP clients; a definition being generated always writes it as null. - `compare` object, nullable, required: What the widget is drawn several times over, or null when it compares nothing, which is the usual answer. accounts draws one series per connected account and takes at least two of them; filters draws one per saved filter of the account; breakdown splits one page metric of facebook_insights into a series per value -- the metric itself says what it is broken down by -- and only a graph_trend takes it. A comparator is only written when the widget is created: neither the class of comparator, nor whether it is broken down over time, nor the fact of being one is changed later. Which classes an integration offers depends on the entity level of its data view too. Set by API and MCP clients; a definition being generated always writes it as null. - `by` 'accounts' | 'filters' | 'breakdown', required: What the widget compares. - `items` union, required: What is compared, in the order it is drawn: one object per connected account for accounts, one per saved filter for filters, and null for a breakdown, which compares no list. - object[] - `connection_key` string, required: Connection the account is read through, as the list of connections publishes it. - `account_id` string, required: Account of that connection, as the list of connections publishes it. - `label` string, nullable, required: Title this series is drawn under, or null for the one the platform builds. It cannot contain the sequence #%#, which is what the parts of a stored account are joined with. - `filter_key` string, nullable, required: Key of a saved filter of the account this one account is read through, or null when it is read whole. The same account is never compared twice. - object[] - `filter_key` string, required: Key of a saved filter of the account, as the list of filters publishes it. The same filter is never compared twice. - `over_time` 'none' | 'hour_of_day' | 'day_of_week' | 'date' | 'week' | 'month' | 'year', nullable, required: The axis a table_metrics_summary comparator is broken down along, which is what makes it an over time table. Null, like none, draws one column per compared element and is the usual answer. Only a table_metrics_summary takes it, and which axes a source offers depends on the integration and on the entity level of its data view. Switching the axis on or off is a change of subtype and is not done by an update. - `show_past_serie` boolean, nullable, required: True when a breakdown draws the period before as a second series. Null, like false, draws only the current period, and it is null in the other two comparators. - `filters` object, nullable, required: Filter applied to the widget, or null for no filter. - `combination` 'and' | 'or', required: How the conditions are joined. Only a combination the catalogue declares for this integration. - `conditions` object[], required - `field` string, required: Public id of a filter field of the catalogue. - `operator` string, required: Operator declared by the catalogue for that field or category. - `value` string, required: Value to compare against, always as text: digits for numbers, comma separated for list operators. - `type_of_field` string, nullable, required: dimension or metric, only when the catalogue says this integration requires it; null otherwise. - `sort_by` object, nullable, required: Ordering of the rows, or null to leave the default. - `field` string, required: Public field id, present in metrics or dimensions. - `direction` 'asc' | 'desc', required - `limit` integer, nullable, required: Maximum number of rows, or null for the default. - `date_range` object, nullable, required: Date range of the widget, or null to inherit the range of the report section. - `preset` string, required: One of the preset ids listed in the instructions, or custom when start and end are given. - `x` integer, nullable, required: Number of days, weeks or months of a preset that takes one; null otherwise. - `start` string, nullable, required: First day of a custom range as yyyy-MM-dd; null otherwise. - `end` string, nullable, required: Last day of a custom range as yyyy-MM-dd; null otherwise. - `previous_period` boolean, nullable, required: The comparison with the period immediately before, and each shape draws it its own way: a graph_trend draws it as an extra series, a table_dimensions_metrics as a column with the percentage of change of each metric, and a kpi and a table_metrics_summary as the figure of that period beside the current one. A few sources and views do not offer it and say so in a warning. In the two shapes that draw the figure, null means yes, which is what those widgets show by default; in the other two, null is the usual answer and asks for nothing. - `show_change` boolean, nullable, required: The PERCENTAGE OF CHANGE against the period before, drawn only by a kpi and by a table_metrics_summary, where it is the other half of the comparison: previous_period is the figure of that period and this one is the percentage beside it. The two are independent and neither governs the other. Null means yes, which is what those widgets show by default. The other five shapes carry null: the change asked for in one of them is ignored with a warning. - `settings` object[], nullable, required: Integration settings the catalogue declares, or null when none applies. - `id` string, required: Setting id of the catalogue. - `value` string, required: One of the values the catalogue lists for that setting. - `unfulfilled` string[], required: What was asked for and cannot be expressed with this integration, one short sentence each. Empty when everything asked for is in the widget. - `ai_summary` WidgetAiSummary: An AI summary as the body of a write says it, and as a read publishes it: the reach of the summary, what that reach reads, the settings it is generated with and the brief of this summary. The settings and the brief are a partial write: what a body does not name is left exactly as it is, a setting sent with the value the summary already uses is not written either, and a creation that names none of them is generated with the defaults /v1/capabilities publishes. It is not a visualization of the DSL, and it could not be: what a summary is made of is a selection of widgets or of sections, which is said with keys of this account's report, while a definition is portable and knows no keys. A TEMPLATE takes none: what a template holds is copied into every report built on it, and a summary reads the widgets and the sections of ONE report by key, so each copy would be a summary of nothing; it is refused with APPEND_NOT_ALLOWED and reason template_report. - `scope` 'widget' | 'section' | 'report', required: What the summary reads. `widget` reads the widgets of its own section it names, `section` reads the data widgets of its own section and `report` reads the sections it names. The scope of a summary that exists is never changed: a summary of another scope is placed and counted differently, so it is deleted and written again. - `widget_keys` string[]: Scopes `widget` and `section`: the data widgets of THIS section the summary reads. THREE STATES. Left out (or null) does not touch it: a creation falls back to what the editor preselects -- the first data widgets of the section -- and an update leaves the summary reading exactly what it reads. An EMPTY list is a value and means the summary reads nothing, which is the state the application saves a summary in with nothing ticked: it is accepted and the answer carries a SELECTION_EMPTY warning, because it generates nothing until something is selected. A list of keys replaces the selection. It is required on a CREATION of scope `widget`, which is made of them, and not on an update of one. /v1/capabilities publishes how many each scope may read. - `section_keys` string[]: Scope `report`: the sections of this report the summary reads. The same three states as widget_keys: left out (or null) does not touch it -- a creation reads the first sections of the report, an update keeps what the summary has -- an EMPTY list means it reads nothing and answers a SELECTION_EMPTY warning, and a list of keys replaces the selection. - `position` 'first' | 'last': Scopes `section` and `report`, and only when the summary is being CREATED: where it goes inside the section, alone in its row. `first`, the default, puts it above everything the section carries and moves the rest down; `last` opens a new row after the last one. A summary of scope `widget` is one more widget of its row and takes `placement` instead. - `ai_prompt` string, nullable: The instruction this summary is written with. A read answers the text or null, and never an empty string: an emptied box is stored as nothing, so a text that was emptied and one nobody ever wrote come back the same way. In a body, leaving the property out or sending null keeps what the summary has, an empty string empties it and a text replaces it. - `ai_include_comparison` boolean: Whether the summary is given the period before to compare against. True unless it was turned off. - `ai_auto_refresh` boolean: Whether the summary is written again on its own when the period or the data behind it changes. True unless it was turned off. - `ai_fallback_policy` 'omit' | 'default_text' | 'last_cached': What the end client is shown when the summary could not be produced: `omit` prints nothing, `default_text` prints ai_fallback_text and `last_cached` prints the last summary that worked, labelled with the period it belongs to. `omit` unless it was changed. - `ai_quota_fallback_policy` 'omit' | 'default_text' | 'last_cached': The same choice for the case where the monthly AI allowance of the account is used up, which is a different question: the data is healthy there. `omit` unless it was changed. - `ai_fallback_text` string, nullable: The text both policies print when they are set to `default_text`. It is never required: a policy of `default_text` with no text prints nothing, exactly as the settings window of the application saves it. A read answers the text or null and never an empty string; in a body, null keeps what the summary has and an empty string empties it. - `ai_show_disclaimer` boolean: Whether the summary carries the line saying it was written by AI. True unless it was turned off. - `ai_brief` string, nullable: What the summaries of THIS widget should know about the client. It is the narrowest of three levels and they are added together, not replaced: the level of the account is written in the application, the level of the report travels on /v1/reports/update, and this one belongs to the summary. Three states: leaving the property out -- or sending null, which is what a read answers when there is none -- keeps the brief the summary has, an empty string removes it, and a text replaces it. It is stored apart from the widget: it is not covered by widget_hash, /v1/widgets/restore does not bring it back, and a brief that could not be written comes back as an AI_BRIEF_NOT_SAVED warning of a 200. - `credits_per_generation` integer: What ONE generation of this summary costs the account, in credits. It is spent when the report is opened, viewed or delivered by a schedule, and never when the summary is written. - `credits_note` string: When those credits are spent. - `connection_key` string: The connection this widget reads through, and ONLY for a report whose report_level_data_sources is false, which binds a connection per widget: those reports refuse a widget that brings none, and a report that binds once for the whole report refuses one that brings it. Take it from /v1/connections. A widget that reads no data -- a text block, an AI summary -- takes none. An ai_summary never takes one: it reads the widgets of its report and no data source of its own. - `account_id` string: The account within that connection. It travels with connection_key and is checked against the accounts /v1/connections publishes for it. - `placement` object: Where to put the widget. When absent, it is placed automatically. Only row and column are read here; anything else inside this object is ignored rather than refused. - `row` integer, required: 1-based row. It must be an existing row or the first free one, and it must respect how that row is already filled: a refusal names the first free row in details.first_free_row. - `column` integer: 1-based column. A column that is taken is moved to the first free one and the answer says so with placement_adjusted. - `dry_run` boolean: Validate and answer what WOULD happen, without writing anything. Default false. - `idempotency_key` string: Makes a retry safe: the answer of the first call is replayed for 24 hours to any later call with the same key and the same body. - `if_version` number: The version of the report's sections structure this call expects. A mismatch is HTTP 409 VERSION_MISMATCH and nothing is written. - `verify_data` boolean: Run the widget against the provider before writing it, waiting up to 45 seconds, and publish the verdict as data.verification. A failure verdict refuses the write with HTTP 422. Default false. ## Response `200` The call succeeded. This endpoint reports its errors with a real HTTP status, so a 200 always carries status=ok. - WidgetsCreateResponse: The 200 of this write, in whichever of its shapes the call asked for. - `status` 'ok', required - `data` union, required: The answer of the write. A call sent with dry_run: true answers WriteDryRunData instead. Those two carry the preview of this operation alongside their own fields, so an answer can match more than one of these shapes. - WidgetsCreateData: The widget that was created, where it landed and, when verify_data was asked for, the verdict of running it. - `report_key` string, required - `section_key` string, required - `name` string, nullable, required - `integration_id` string, required - `row` integer, nullable, required - `column` integer, nullable, required - `placement_adjusted` boolean, required: True when the column asked for was taken and the widget went to the first free one. - `widget_key` string, required - `created_keys` string[], required: Every key this call created, flat, in the order they were written. - `verification` Verification: The verdict of a verify_data run. - `verdict` 'success' | 'success_empty' | 'not_verifiable' | 'failure', required: What happened when the widget was run against the provider. `success`: rows came back. `success_empty`: the provider answered with no rows. `not_verifiable`: it could not be run right now. `failure`: the provider refused it, and the widget is not written. - `message` string, nullable: What the provider said, when it said anything. - `reason` string, nullable: Why the run did not happen, for not_verifiable. `busy` means the verification capacity is in use and the call did not wait for it. - `unfulfilled` string[], required: What was asked for and could not be expressed; the API returns an empty list. - `warnings` Problem[], required: Things worth knowing about a write that DID happen. They are never errors. - `path` string, required: Dotted path of the offending field inside the request body, empty for a problem about the body as a whole. - `code` 'MISSING_REQUIRED_FIELD' | 'INVALID_VALUE' | 'UNKNOWN_PARAM' | 'INVALID_DATE_RANGE' | 'INTEGRATION_NOT_SUPPORTED' | 'CONNECTION_NOT_FOUND' | 'ACCOUNT_NOT_FOUND' | 'ONE_CONNECTION_PER_INTEGRATION' | 'DATE_PRESET_NOT_IN_REPORTS' | 'TEMPLATE_NOT_FOUND' | 'TEMPLATE_PARAMETERIZED' | 'ROW_RULE' | 'POSITION_CHANGED' | 'SECTION_CHANGED' | 'TOO_MANY_PROBLEMS' | 'DATA_SOURCES_PARTIALLY_BOUND' | 'WELCOME_EMAIL_FAILED' | 'REPORT_WILL_BE_SHARED_PUBLICLY' | 'MAY_BE_DELIVERED_AGAIN_TODAY' | 'SUMMARY_SECTION_SKIPPED' | 'COVER_FILE_REMOVED' | 'COVER_TYPE_SET' | 'LOGO_REMOVED' | 'READ_ONLY_FIELD_IGNORED' | 'AI_BRIEF_NOT_SAVED' | 'SPENDS_AI_CREDITS' | 'DROPS_ACCOUNT_CACHE' | 'CHART_TYPE_NOT_FOR_THIS_SHAPE' | 'CHART_TYPE_NOT_FOR_TREND' | 'CONTENT_NOT_FOR_THIS_SHAPE' | 'FULL_WIDTH_NOT_FOR_KPI' | 'DATA_VIEW_FIELDS_PER_ACCOUNT' | 'DATA_VIEW_NOT_AVAILABLE' | 'DATA_VIEW_NOT_EMPTY' | 'DATA_VIEW_NOT_OF_THE_SECTION' | 'DATA_VIEW_REQUIRED' | 'DATA_VIEW_UNKNOWN' | 'DATE_RANGE_CUSTOM_INCOMPLETE' | 'DATE_RANGE_CUSTOM_INVALID' | 'DATE_RANGE_CUSTOM_INVERTED' | 'DATE_RANGE_PRESET_UNKNOWN' | 'DATE_RANGE_X_REQUIRED' | 'DIMENSION_DUPLICATED' | 'DIMENSION_IS_A_METRIC' | 'DIMENSION_REQUIRED' | 'DIMENSION_UNKNOWN' | 'DSL_INVALID' | 'FIELD_NOT_IN_AD_TYPES' | 'FILTERS_NOT_SUPPORTED' | 'FILTER_COMBINATION_NOT_SUPPORTED' | 'FILTER_CONDITIONS_MIXED_KINDS' | 'FILTER_FIELD_NOT_A_COLUMN' | 'FILTER_FIELD_NOT_IN_DATA_VIEW' | 'FILTER_FIELD_UNKNOWN' | 'FILTER_LIKE_NEEDS_WILDCARD' | 'FILTER_OPERATOR_INVALID' | 'FILTER_TYPE_OF_FIELD_MISMATCH' | 'FILTER_TYPE_OF_FIELD_NOT_USED' | 'FILTER_TYPE_OF_FIELD_REQUIRED' | 'FILTER_VALUE_EMPTY' | 'FILTER_VALUE_NOT_ALLOWED' | 'FILTER_VALUE_NOT_NUMERIC' | 'GRAPH_EXTRA_DIMENSIONS' | 'GRAPH_TIME_DIMENSION_MIXED' | 'INTEGRATION_UNKNOWN' | 'INTENT_MISSING' | 'KPI_NEEDS_ONE_METRIC' | 'KPI_TAKES_NO_DIMENSIONS' | 'LIMIT_IGNORED_IN_TREND' | 'LIMIT_OUT_OF_RANGE' | 'METRICS_SUMMARY_TAKES_NO_DIMENSIONS' | 'METRIC_DUPLICATED' | 'METRIC_IS_A_DIMENSION' | 'METRIC_UNKNOWN' | 'METRICS_REQUIRED' | 'NAME_EMPTY' | 'NAME_NOT_PLAIN_TEXT' | 'NAME_TOO_LONG' | 'PIE_TAKES_ONE_DIMENSION' | 'PREVIOUS_PERIOD_NOT_AVAILABLE_FOR_THIS_TABLE' | 'COMPARISON_NOT_AVAILABLE_FOR_THIS_WIDGET' | 'PREVIOUS_PERIOD_NOT_FOR_THIS_SHAPE' | 'SHOW_CHANGE_NOT_FOR_THIS_SHAPE' | 'SHOW_CHANGE_NEEDS_PREVIOUS_PERIOD' | 'SETTING_DUPLICATED' | 'SETTING_NOT_FOR_DATA_VIEW' | 'SETTING_NOT_IN_WIDGET' | 'SETTING_UNKNOWN' | 'SETTING_VALUE_INVALID' | 'SORT_FIELD_NOT_SELECTED' | 'TABLE_TIME_DIMENSION_NOT_DRAWABLE' | 'TEXT_CONTENT_REQUIRED' | 'TEXT_CONTENT_TOO_LONG' | 'TEXT_NOT_A_WIDGET_TYPE' | 'TEXT_TAKES_NO_DATA' | 'TOO_MANY_DIMENSIONS' | 'TOO_MANY_METRICS' | 'TRANSLATION_FAILED' | 'TREND_ORDER_IS_THE_TIME_AXIS' | 'TREND_TAKES_NO_DIMENSIONS' | 'VISUALIZATION_NOT_AVAILABLE' | 'VISUALIZATION_NOT_OF_REQUESTED_TYPE' | 'FIELD_NOT_AVAILABLE' | 'FILTER_CONDITIONS_MIXED_LEVELS' | 'FILTER_LEVEL_MISMATCH' | 'FORMULA_TOKEN_UNKNOWN' | 'FORMULA_INVALID' | 'LANGUAGE_FILE_KEYS_IGNORED' | 'SELECTION_EMPTY' | 'FILTER_KEY_AND_FILTERS' | 'FILTER_KEY_NOT_FOR_THIS_SHAPE' | 'KEY_NOT_FOR_GENERATION' | 'COMPARE_NOT_FOR_THIS_SHAPE' | 'COMPARE_NOT_FOR_INTEGRATION' | 'COMPARE_ITEMS_CARDINALITY' | 'COMPARE_METRICS_CARDINALITY' | 'COMPARE_OVER_TIME_NOT_AVAILABLE' | 'COMPARE_METRIC_HAS_NO_BREAKDOWN' | 'COMPARE_KEY_NOT_FOR_THIS_COMPARATOR' | 'COMPARATOR_CLASS_FIXED', required: What is wrong with that field, or what is worth knowing about it. The first group is the vocabulary of the body itself; the eleven that follow it appear only in `warnings[]`, bar TOO_MANY_PROBLEMS, which closes a `problems` list that was truncated; and the rest are the codes of the widget DSL, which travel here when a composition or a widget carries a dsl that does not validate. - `message` string, required: Human-readable explanation. Not stable across versions. - `suggestions` string[]: The accepted values, when the field takes a closed list. - `ai_summary` WidgetAiSummary: An AI summary as the body of a write says it, and as a read publishes it: the reach of the summary, what that reach reads, the settings it is generated with and the brief of this summary. The settings and the brief are a partial write: what a body does not name is left exactly as it is, a setting sent with the value the summary already uses is not written either, and a creation that names none of them is generated with the defaults /v1/capabilities publishes. It is not a visualization of the DSL, and it could not be: what a summary is made of is a selection of widgets or of sections, which is said with keys of this account's report, while a definition is portable and knows no keys. A TEMPLATE takes none: what a template holds is copied into every report built on it, and a summary reads the widgets and the sections of ONE report by key, so each copy would be a summary of nothing; it is refused with APPEND_NOT_ALLOWED and reason template_report. - `scope` 'widget' | 'section' | 'report', required: What the summary reads. `widget` reads the widgets of its own section it names, `section` reads the data widgets of its own section and `report` reads the sections it names. The scope of a summary that exists is never changed: a summary of another scope is placed and counted differently, so it is deleted and written again. - `widget_keys` string[]: Scopes `widget` and `section`: the data widgets of THIS section the summary reads. THREE STATES. Left out (or null) does not touch it: a creation falls back to what the editor preselects -- the first data widgets of the section -- and an update leaves the summary reading exactly what it reads. An EMPTY list is a value and means the summary reads nothing, which is the state the application saves a summary in with nothing ticked: it is accepted and the answer carries a SELECTION_EMPTY warning, because it generates nothing until something is selected. A list of keys replaces the selection. It is required on a CREATION of scope `widget`, which is made of them, and not on an update of one. /v1/capabilities publishes how many each scope may read. - `section_keys` string[]: Scope `report`: the sections of this report the summary reads. The same three states as widget_keys: left out (or null) does not touch it -- a creation reads the first sections of the report, an update keeps what the summary has -- an EMPTY list means it reads nothing and answers a SELECTION_EMPTY warning, and a list of keys replaces the selection. - `position` 'first' | 'last': Scopes `section` and `report`, and only when the summary is being CREATED: where it goes inside the section, alone in its row. `first`, the default, puts it above everything the section carries and moves the rest down; `last` opens a new row after the last one. A summary of scope `widget` is one more widget of its row and takes `placement` instead. - `ai_prompt` string, nullable: The instruction this summary is written with. A read answers the text or null, and never an empty string: an emptied box is stored as nothing, so a text that was emptied and one nobody ever wrote come back the same way. In a body, leaving the property out or sending null keeps what the summary has, an empty string empties it and a text replaces it. - `ai_include_comparison` boolean: Whether the summary is given the period before to compare against. True unless it was turned off. - `ai_auto_refresh` boolean: Whether the summary is written again on its own when the period or the data behind it changes. True unless it was turned off. - `ai_fallback_policy` 'omit' | 'default_text' | 'last_cached': What the end client is shown when the summary could not be produced: `omit` prints nothing, `default_text` prints ai_fallback_text and `last_cached` prints the last summary that worked, labelled with the period it belongs to. `omit` unless it was changed. - `ai_quota_fallback_policy` 'omit' | 'default_text' | 'last_cached': The same choice for the case where the monthly AI allowance of the account is used up, which is a different question: the data is healthy there. `omit` unless it was changed. - `ai_fallback_text` string, nullable: The text both policies print when they are set to `default_text`. It is never required: a policy of `default_text` with no text prints nothing, exactly as the settings window of the application saves it. A read answers the text or null and never an empty string; in a body, null keeps what the summary has and an empty string empties it. - `ai_show_disclaimer` boolean: Whether the summary carries the line saying it was written by AI. True unless it was turned off. - `ai_brief` string, nullable: What the summaries of THIS widget should know about the client. It is the narrowest of three levels and they are added together, not replaced: the level of the account is written in the application, the level of the report travels on /v1/reports/update, and this one belongs to the summary. Three states: leaving the property out -- or sending null, which is what a read answers when there is none -- keeps the brief the summary has, an empty string removes it, and a text replaces it. It is stored apart from the widget: it is not covered by widget_hash, /v1/widgets/restore does not bring it back, and a brief that could not be written comes back as an AI_BRIEF_NOT_SAVED warning of a 200. - `credits_per_generation` integer: What ONE generation of this summary costs the account, in credits. It is spent when the report is opened, viewed or delivered by a schedule, and never when the summary is written. - `credits_note` string: When those credits are spent. - WriteDryRunData: The answer of a call sent with dry_run: true. Nothing was written, no idempotency key was consumed and no confirmation token was burnt. When valid is true, the fields of the endpoint's own preview travel alongside these two; when it is false, `message` and `problems` do instead. The HTTP status is 200 in both cases: a rehearsal that says the body is wrong is a successful rehearsal. Two refusals are NOT translated and arrive with their real status: a 429 (the budgets are spent either way) and any 5xx. - `dry_run` true, required: Always true in this shape. - `valid` boolean, required: Whether the body would be accepted. - `message` string: Why the body would be refused. Present only when valid is false. - `problems` Problem[]: Everything that is wrong with the body. Present only when valid is false. - `path` string, required: Dotted path of the offending field inside the request body, empty for a problem about the body as a whole. - `code` 'MISSING_REQUIRED_FIELD' | 'INVALID_VALUE' | 'UNKNOWN_PARAM' | 'INVALID_DATE_RANGE' | 'INTEGRATION_NOT_SUPPORTED' | 'CONNECTION_NOT_FOUND' | 'ACCOUNT_NOT_FOUND' | 'ONE_CONNECTION_PER_INTEGRATION' | 'DATE_PRESET_NOT_IN_REPORTS' | 'TEMPLATE_NOT_FOUND' | 'TEMPLATE_PARAMETERIZED' | 'ROW_RULE' | 'POSITION_CHANGED' | 'SECTION_CHANGED' | 'TOO_MANY_PROBLEMS' | 'DATA_SOURCES_PARTIALLY_BOUND' | 'WELCOME_EMAIL_FAILED' | 'REPORT_WILL_BE_SHARED_PUBLICLY' | 'MAY_BE_DELIVERED_AGAIN_TODAY' | 'SUMMARY_SECTION_SKIPPED' | 'COVER_FILE_REMOVED' | 'COVER_TYPE_SET' | 'LOGO_REMOVED' | 'READ_ONLY_FIELD_IGNORED' | 'AI_BRIEF_NOT_SAVED' | 'SPENDS_AI_CREDITS' | 'DROPS_ACCOUNT_CACHE' | 'CHART_TYPE_NOT_FOR_THIS_SHAPE' | 'CHART_TYPE_NOT_FOR_TREND' | 'CONTENT_NOT_FOR_THIS_SHAPE' | 'FULL_WIDTH_NOT_FOR_KPI' | 'DATA_VIEW_FIELDS_PER_ACCOUNT' | 'DATA_VIEW_NOT_AVAILABLE' | 'DATA_VIEW_NOT_EMPTY' | 'DATA_VIEW_NOT_OF_THE_SECTION' | 'DATA_VIEW_REQUIRED' | 'DATA_VIEW_UNKNOWN' | 'DATE_RANGE_CUSTOM_INCOMPLETE' | 'DATE_RANGE_CUSTOM_INVALID' | 'DATE_RANGE_CUSTOM_INVERTED' | 'DATE_RANGE_PRESET_UNKNOWN' | 'DATE_RANGE_X_REQUIRED' | 'DIMENSION_DUPLICATED' | 'DIMENSION_IS_A_METRIC' | 'DIMENSION_REQUIRED' | 'DIMENSION_UNKNOWN' | 'DSL_INVALID' | 'FIELD_NOT_IN_AD_TYPES' | 'FILTERS_NOT_SUPPORTED' | 'FILTER_COMBINATION_NOT_SUPPORTED' | 'FILTER_CONDITIONS_MIXED_KINDS' | 'FILTER_FIELD_NOT_A_COLUMN' | 'FILTER_FIELD_NOT_IN_DATA_VIEW' | 'FILTER_FIELD_UNKNOWN' | 'FILTER_LIKE_NEEDS_WILDCARD' | 'FILTER_OPERATOR_INVALID' | 'FILTER_TYPE_OF_FIELD_MISMATCH' | 'FILTER_TYPE_OF_FIELD_NOT_USED' | 'FILTER_TYPE_OF_FIELD_REQUIRED' | 'FILTER_VALUE_EMPTY' | 'FILTER_VALUE_NOT_ALLOWED' | 'FILTER_VALUE_NOT_NUMERIC' | 'GRAPH_EXTRA_DIMENSIONS' | 'GRAPH_TIME_DIMENSION_MIXED' | 'INTEGRATION_UNKNOWN' | 'INTENT_MISSING' | 'KPI_NEEDS_ONE_METRIC' | 'KPI_TAKES_NO_DIMENSIONS' | 'LIMIT_IGNORED_IN_TREND' | 'LIMIT_OUT_OF_RANGE' | 'METRICS_SUMMARY_TAKES_NO_DIMENSIONS' | 'METRIC_DUPLICATED' | 'METRIC_IS_A_DIMENSION' | 'METRIC_UNKNOWN' | 'METRICS_REQUIRED' | 'NAME_EMPTY' | 'NAME_NOT_PLAIN_TEXT' | 'NAME_TOO_LONG' | 'PIE_TAKES_ONE_DIMENSION' | 'PREVIOUS_PERIOD_NOT_AVAILABLE_FOR_THIS_TABLE' | 'COMPARISON_NOT_AVAILABLE_FOR_THIS_WIDGET' | 'PREVIOUS_PERIOD_NOT_FOR_THIS_SHAPE' | 'SHOW_CHANGE_NOT_FOR_THIS_SHAPE' | 'SHOW_CHANGE_NEEDS_PREVIOUS_PERIOD' | 'SETTING_DUPLICATED' | 'SETTING_NOT_FOR_DATA_VIEW' | 'SETTING_NOT_IN_WIDGET' | 'SETTING_UNKNOWN' | 'SETTING_VALUE_INVALID' | 'SORT_FIELD_NOT_SELECTED' | 'TABLE_TIME_DIMENSION_NOT_DRAWABLE' | 'TEXT_CONTENT_REQUIRED' | 'TEXT_CONTENT_TOO_LONG' | 'TEXT_NOT_A_WIDGET_TYPE' | 'TEXT_TAKES_NO_DATA' | 'TOO_MANY_DIMENSIONS' | 'TOO_MANY_METRICS' | 'TRANSLATION_FAILED' | 'TREND_ORDER_IS_THE_TIME_AXIS' | 'TREND_TAKES_NO_DIMENSIONS' | 'VISUALIZATION_NOT_AVAILABLE' | 'VISUALIZATION_NOT_OF_REQUESTED_TYPE' | 'FIELD_NOT_AVAILABLE' | 'FILTER_CONDITIONS_MIXED_LEVELS' | 'FILTER_LEVEL_MISMATCH' | 'FORMULA_TOKEN_UNKNOWN' | 'FORMULA_INVALID' | 'LANGUAGE_FILE_KEYS_IGNORED' | 'SELECTION_EMPTY' | 'FILTER_KEY_AND_FILTERS' | 'FILTER_KEY_NOT_FOR_THIS_SHAPE' | 'KEY_NOT_FOR_GENERATION' | 'COMPARE_NOT_FOR_THIS_SHAPE' | 'COMPARE_NOT_FOR_INTEGRATION' | 'COMPARE_ITEMS_CARDINALITY' | 'COMPARE_METRICS_CARDINALITY' | 'COMPARE_OVER_TIME_NOT_AVAILABLE' | 'COMPARE_METRIC_HAS_NO_BREAKDOWN' | 'COMPARE_KEY_NOT_FOR_THIS_COMPARATOR' | 'COMPARATOR_CLASS_FIXED', required: What is wrong with that field, or what is worth knowing about it. The first group is the vocabulary of the body itself; the eleven that follow it appear only in `warnings[]`, bar TOO_MANY_PROBLEMS, which closes a `problems` list that was truncated; and the rest are the codes of the widget DSL, which travel here when a composition or a widget carries a dsl that does not validate. - `message` string, required: Human-readable explanation. Not stable across versions. - `suggestions` string[]: The accepted values, when the field takes a closed list. - `meta` WriteMeta, required: The meta of a write answer. A replay carries two extra fields so it can be told apart from a first call and still traced back to it. - `request_id` string, required: Id of THIS request, including on a replay. - `timestamp` string, date-time, required - `replayed` boolean: True when this answer is the stored answer of an earlier call with the same idempotency_key and the same body. Absent on a first call. - `replayed_from_request_id` string, nullable: The request_id of the call that actually did the work. Present only on a replay, and null when the stored answer carried no id. ## Other responses - `400`: The request was refused before anything was written. INVALID_REQUEST covers the framework refusals of every endpoint (the Content-Type is not application/json, the body could not be read, the body is not valid JSON). The report endpoints add their own: VALIDATION_FAILED with details.problems, UNKNOWN_PARAM with the path of the field the endpoint does not read, and the single-field refusals MISSING_REQUIRED_FIELD, INVALID_VALUE and APPEND_NOT_ALLOWED. The description of each operation lists the ones it emits. Everything else a body can get wrong arrives as an entry of details.problems rather than as the code of the envelope. - `401`: Invalid or missing Bearer token. - `403`: The account may not perform this write. WRITE_ACCESS_DISABLED means write access has not been enabled for the API key in the REST API tab of the app. PLAN_LIMIT_REACHED means the plan has no room left for another report or template, and is emitted only by the endpoints that create one. - `404`: A key of the request does not name anything this account owns. NOT_FOUND covers reports, sections, widgets, schedules, portal views, brand templates, email templates and language files; TEMPLATE_NOT_FOUND covers a catalog slug, version or section position that stopped being published between the validation and the write. An entity of another account is reported the same way as one that does not exist. - `409`: The request is well formed and the current state does not allow it. Every write can answer IDEMPOTENCY_CONFLICT (the same idempotency_key with a different body) and IDEMPOTENCY_IN_PROGRESS (an identical call is running, with a Retry-After). The endpoints that ask for a confirmation add CONFIRMATION_EXPIRED and CONFIRMATION_MISMATCH; the ones that never confirm cannot answer either. Each operation lists the domain conflicts it adds on top of those. - `413`: The request body is over 1 MB. The cap is applied before authentication, so this answer is refused without the key being read, and it carries the usual error envelope. - `422`: verify_data ran the widget against the provider and the provider refused it. Nothing was created; `details` carries the verdict of the run. - `429`: A budget was spent. RATE_LIMITED is the per-key read budget (300/minute, 5000/day) or the per-account write budget (60/minute, 1000/day). EMAIL_QUOTA_REACHED is the separate budget of 50 calls per account per day that send email, and its Retry-After points at the next UTC midnight. The Retry-After header gives the seconds until the more restrictive window resets. The X-RateLimit-* headers report the read budget, except on the 429 of a write, where they report the write budget. - `500`: Internal server error. - `503`: Writes are paused, or the store the write guarantees rely on is not answering. Nothing was written and the call is worth retrying later. Reads are unaffected. --- [API](https://skmtc.dev/reportingninja/apis/reporting-ninja-api.md) · [All operations](https://skmtc.dev/reportingninja/apis/reporting-ninja-api/llms.txt) · [OpenAPI document](https://skmtc.dev/reportingninja/apis/reporting-ninja-api/revisions/13876c71be34?raw)