MCP operation reference

Human approval is required before this call, and a suggestion with needsReview=true (a lesson you wrote that no person has approved) is refused here: the person accepts it in Model notes. The accepted source id is recorded so it cannot resurface. To write directly without suggestion automation, use set_model_primer.

Entitlement: Free entry point. Paid advice and selected automatic actions still need Pro where stated. Release class: supported. Risk class: standard.

Permission: the current target policy applies. Exact approvals and human-only actions remain as stated in this operation’s description.

Insights are the user'S own data: plain append-only JSONL (.semanticus/knowledge/insights.jsonl for project scope, ~/.semanticus/knowledge/ for global), readable without us, never rewritten. Give deterministic match keys (rule ids, gate signatures, error types, op names, domain tokens) so recall can find it. Write-gated (ssgm): it lands 'pending' (held out of recall) until approve_insight, unless the auto-approve setting is on, which defaults to true for single-user local mode (set knowledge-settings.json {"autoApprove": false} to force review). Whatever you record is never 'approved': it lands 'unreviewed' (or 'pending'), untrusted and needsReview=true, until a person approves it in Model notes. Recall returns it marked untrusted, as data not instruction, and it reaches the Primer only when the person accepts it. Text that looks like an instruction to skip checks or swap models is never auto-approved: it stays pending, lists under status=pending, and keeps its untrusted marker even once a person approves it. fingerprintScoped=true pins it to the current model's shape (needs an open model); false = a user-level lesson that travels across all models.

Entitlement: Free entry point. Paid advice and selected automatic actions still need Pro where stated. Release class: supported. Risk class: standard.

Permission: the current target policy applies. Exact approvals and human-only actions remain as stated in this operation’s description.

From you it never makes an insight 'approved' or trusted: the result is status 'unreviewed', untrusted, recalled only as data, until a person approves it in Model notes. No-op-safe on an insight that is already released.

Entitlement: Free entry point. Paid advice and selected automatic actions still need Pro where stated. Release class: supported. Risk class: standard.

Permission: the current target policy applies. Exact approvals and human-only actions remain as stated in this operation’s description.

It vanishes from the live set and recall. Prefer downvote for a lesson that merely lost relevance.

Entitlement: Free entry point. Paid advice and selected automatic actions still need Pro where stated. Release class: supported. Risk class: high.

Permission: the current target policy applies. Exact approvals and human-only actions remain as stated in this operation’s description.

When the score falls to 0 the insight is materialized out of the live set (the delta trail is kept, nothing is erased). Use it to retire a lesson that stopped applying.

Entitlement: Free entry point. Paid advice and selected automatic actions still need Pro where stated. Release class: supported. Risk class: standard.

Permission: the current target policy applies. Exact approvals and human-only actions remain as stated in this operation’s description.

Provide new text and/or keys; omit one to leave it unchanged.

Entitlement: Free entry point. Paid advice and selected automatic actions still need Pro where stated. Release class: supported. Risk class: standard.

Permission: the current target policy applies. Exact approvals and human-only actions remain as stated in this operation’s description.

It is plain Markdown in a .semanticus/primers sidecar, beside the model (travels in source control) for a disk model, or in the workspace for a live/local connection, and always uses six fixed sections: Overview, Business context, Gotchas, Patterns, Known issues, History. Read this before model work for the declared business context and known traps. Read-only.

Entitlement: Free entry point. Paid advice and selected automatic actions still need Pro where stated. Release class: supported. Risk class: standard.

Permission: the current target policy applies. Exact approvals and human-only actions remain as stated in this operation’s description.

No client parameters.

Filter by scope ('project'|'global'; omit for both) and/or status ('pending' = held out of recall, 'unreviewed' = recalled as untrusted data, 'approved' = a person approved it). Reports the count of any corrupt JSONL lines it skipped. The store never bricks. Read-only.

Entitlement: Free entry point. Paid advice and selected automatic actions still need Pro where stated. Release class: supported. Risk class: standard.

Permission: the current target policy applies. Exact approvals and human-only actions remain as stated in this operation’s description.

Free on any tier. Each proposal carries the source lesson id, capture time and source-run provenance, and needsReview=true when you wrote the lesson and no person has approved it yet (only the person can accept those, in Model notes). This never changes the Primer. Show the proposal to the user, then call accept_primer_suggestion only after explicit approval or reject_primer_suggestion to dismiss it.

Entitlement: Free entry point. Paid advice and selected automatic actions still need Pro where stated. Release class: supported. Risk class: standard.

Permission: the current target policy applies. Exact approvals and human-only actions remain as stated in this operation’s description.

No client parameters.

Dry run by default: confirm=false returns how many live insights would be erased and changes nothing; confirm=true appends a purge marker so everything before it in that scope becomes invisible on replay (the file is not rewritten).

Entitlement: Free entry point. Paid advice and selected automatic actions still need Pro where stated. Release class: supported. Risk class: high.

Permission: the current target policy applies. Exact approvals and human-only actions remain as stated in this operation’s description.

The engine computes the model's fingerprint, reads both scopes' released insights (status approved, or unreviewed and marked untrusted), and returns a deterministically-ranked candidate set (key-term overlap with your query + same-shape fingerprint bonus + importance score + temporal decay), each with why it matched (matchedKeys). This is retrieval, not judgment: you do the semantic ranking over the candidates and decide what applies. Returns the fingerprint too, and says so plainly when there is no prior experience for this shape. Needs an open model.

Entitlement: Free entry point. Paid advice and selected automatic actions still need Pro where stated. Release class: supported. Risk class: standard.

Permission: the current target policy applies. Exact approvals and human-only actions remain as stated in this operation’s description.

The source learning remains in the hidden learning store, while this model records the rejection so the same suggestion does not keep resurfacing.

Entitlement: Free entry point. Paid advice and selected automatic actions still need Pro where stated. Release class: supported. Risk class: standard.

Permission: the current target policy applies. Exact approvals and human-only actions remain as stated in this operation’s description.

The document must keep exactly six second-level sections in this order: Overview, Business context, Gotchas, Patterns, Known issues, History. Stored beside the model in .semanticus/primers so it travels in source control without changing the model definition. Show the user the proposed edit before writing.

Entitlement: Free entry point. Paid advice and selected automatic actions still need Pro where stated. Release class: supported. Risk class: high.

Permission: the current target policy applies. Exact approvals and human-only actions remain as stated in this operation’s description.

An insight that keeps proving useful ranks higher and resists decay.

Entitlement: Free entry point. Paid advice and selected automatic actions still need Pro where stated. Release class: supported. Risk class: standard.

Permission: the current target policy applies. Exact approvals and human-only actions remain as stated in this operation’s description.

Run_interview grades an inline question without saving, and both are free. Fields by tier: 'value': query (a full evaluate, e.g. evaluate row("v", calculate([Total Sales], 'Date'[Year]=2024))) + expectedValue (the trusted number, confirmed by the user or a verified answer) or expectedMatrixJson (rows, order-insensitive); 'paraphrase': scalarExpr + paraphraseExpr (the same question answered two ways as scalar expressions, not evaluate queries) with optional groupBy/filters for a per-context proof; 'refusal': just the question. It asserts the model cannot answer it, so an assistant that produces a number is confidently wrong. fixRuleId maps a failure to the AI-readiness rule that prevents it (a data-table fallback applies when omitted). seedSource records where the question came from ('user' | 'claude' | 'verified-answer'). Validation is fail-loud so a saved question is always runnable.

Entitlement: Free entry point. Paid advice and selected automatic actions still need Pro where stated. Release class: supported. Risk class: standard.

Permission: the current target policy applies. Exact approvals and human-only actions remain as stated in this operation’s description.

The row-count check then goes back to reading the table's own partition, which is where it was before anyone mapped it. Returns false when that table had no mapping. Changes no model object and needs no deploy.

Entitlement: Free entry point. Paid advice and selected automatic actions still need Pro where stated. Release class: supported. Risk class: standard.

Permission: the current target policy applies. Exact approvals and human-only actions remain as stated in this operation’s description.

Use it to retire a question whose business definition changed. Free.

Entitlement: Free entry point. Paid advice and selected automatic actions still need Pro where stated. Release class: supported. Risk class: high.

Permission: the current target policy applies. Exact approvals and human-only actions remain as stated in this operation’s description.

Returns false when the id wasn't present. Ambient checks cannot be deleted: they derive from the model; to clear a failing integrity check, fix the data it measured.

Entitlement: Free entry point. Paid advice and selected automatic actions still need Pro where stated. Release class: supported. Risk class: high.

Permission: the current target policy applies. Exact approvals and human-only actions remain as stated in this operation’s description.

Returns {markdown, json, html, contentHash, note, error}; JSON is the record of truth, html is its deterministic human view, and contentHash detects any later change. The report includes model and environment context, grade with coverage, hard gates, root causes before detailed evidence, measure reconciliation rows only for failing measures, time-intelligence variant evidence, and relationship counts and rates. It never exports a run from a different model: if no current-model run exists, call run_tests and retry. Read-only.

Entitlement: Free entry point. Paid advice and selected automatic actions still need Pro where stated. Release class: supported. Risk class: standard.

Permission: the current target policy applies. Exact approvals and human-only actions remain as stated in this operation’s description.

No client parameters.

The engine re-verifies the signature and html pairing before returning it; invalid evidence is refused with the exact reason. Read-only.

Entitlement: Free entry point. Paid advice and selected automatic actions still need Pro where stated. Release class: supported. Risk class: standard.

Permission: the current target policy applies. Exact approvals and human-only actions remain as stated in this operation’s description.

Returns the run's health plus every check it decided: each saved definition, each relationship check and each table row count, with its verdict, the answer it was agreed against, what it actually got, the difference, its own words, the settings it ran with (the query it asked, the filter it applied, the tolerance it allowed and the SQL source it resolved to, all as they were at execution time, never re-read from today's definition), and its filter-context rows capped at the number the report shows with the uncapped count beside them. This is how a run from weeks ago can still be read: list_test_runs is a summary and deliberately carries none of it. A run recorded before outcome recording was added comes back with its totals and a note saying outcomes were not recorded then. Read-only.

Entitlement: Free entry point. Paid advice and selected automatic actions still need Pro where stated. Release class: supported. Risk class: standard.

Permission: the current target policy applies. Exact approvals and human-only actions remain as stated in this operation’s description.

A relationship check reports how many fact rows point at a missing lookup key; this answers which ones. Returns one row per distinct many-side key value the one-side table does not have, with how many many-side rows carry it, biggest first, capped at 200 (the cap is disclosed when it hides more). The question asked is character-for-character the one the check counts, so the list and the count can never disagree. It reads current data through the live model, so run after a recorded check and the data may have moved since that check was measured; the note says so. Needs a live connection. Agent calls pass the standard QueryData permission gate, because this is the same row-returning query path. Read-only.

Entitlement: Free entry point. Paid advice and selected automatic actions still need Pro where stated. Release class: supported. Risk class: standard.

Permission: the current target policy applies. Exact approvals and human-only actions remain as stated in this operation’s description.

Each item includes type, producer, verdict with coverage, content signature, paths, and verification state. A changed or conflicted record remains listed as invalid and is never trusted or hidden. Read-only. Next: get_evidence(id) to review one; save_evidence(source) to explicitly retain the latest Test report or a terminal Workflow run.

Entitlement: Free entry point. Paid advice and selected automatic actions still need Pro where stated. Release class: supported. Risk class: standard.

Permission: the current target policy applies. Exact approvals and human-only actions remain as stated in this operation’s description.

No client parameters.

Packs are bound to the model they were authored against, so questions is this model's pack only; otherModelQuestions and unattributedQuestions (legacy, saved before binding, in a shared store) are surfaced separately and are not run against the open model. The pack lives in plain JSONL (.semanticus/interview/questions.jsonl beside the model; '~/.semanticus/interview/' for scope 'global'): the user's own data. Reports the count of any corrupt lines it skipped (the store never bricks). Free, read-only.

Entitlement: Free entry point. Paid advice and selected automatic actions still need Pro where stated. Release class: supported. Risk class: standard.

Permission: the current target policy applies. Exact approvals and human-only actions remain as stated in this operation’s description.

Pack templates bind only to objects that exist (a marked date table, a measure to target, related dimensions...); every unbindable template is listed as skipped with the exact missing shape. no candidate carries a trusted answer and none is fabricated: run the candidate's query, have the user confirm the number, then save it with add_interview_question (expectedValue = the confirmed number; the literal blank records a no-value answer; seedSource 'verified-answer' | 'hard-pack'). Free, read-only.

Entitlement: Free entry point. Paid advice and selected automatic actions still need Pro where stated. Release class: supported. Risk class: standard.

Permission: the current target policy applies. Exact approvals and human-only actions remain as stated in this operation’s description.

This row is the table-count check: there is no saved-test kind for a table count, and mapping a table with set_table_source_mapping is the same action as adding one. 'source' says where the coordinates came from: 'mapping' (a person chose a saved SQL source for this table), 'endpoint' (the model's SQL endpoint, with the schema and table read from the table's own partition), 'detected' (read from the table's own partition, when the model has no endpoint) or 'none'. 'canCount' false means 'reason' says in plain words what is missing, for example a table that is not mapped yet, a mapping that does not name a schema and table, an ambiguous source, or a source that has been removed. 'editable' is true for every table that could be counted, so a table with no source at all still has a row to point at a source. The last outcome (lastVerdict, lastMessage, lastModelCount, lastSourceCount, lastRunUtc) comes from the last run in this session and is absent when nothing has been counted yet, which is its own state and never a pass or a failure. Counts that differ without proven matching snapshots stay 'couldn't check'. Read-only. Next: set_table_source_mapping to point a table at a saved source, run_tests to count.

Entitlement: Free entry point. Paid advice and selected automatic actions still need Pro where stated. Release class: supported. Risk class: standard.

Permission: the current target policy applies. Exact approvals and human-only actions remain as stated in this operation’s description.

No client parameters.

The drift-trend substrate: compare health across runs to see a suite degrade before a user does. Read-only.

Entitlement: Free entry point. Paid advice and selected automatic actions still need Pro where stated. Release class: supported. Risk class: standard.

Permission: the current target policy applies. Exact approvals and human-only actions remain as stated in this operation’s description.

Definitions bind by LineageTag when present, or by a stable tests-sidecar identity for a valid tagless object. A vanished or ambiguous target reports missing on the next run, never a silent pass. Ambient checks are not listed here because they derive from the model on every run_tests. Unreadable lines are counted, never thrown. Read-only. list_sql_source_usage shows which checks use each SQL source.

Entitlement: Free entry point. Paid advice and selected automatic actions still need Pro where stated. Release class: supported. Risk class: standard.

Permission: the current target policy applies. Exact approvals and human-only actions remain as stated in this operation’s description.

No client parameters.

Nothing is executed: this writes the snapshot of the run named by runId (its health, every check's outcome with expected/actual/difference, its model, environment, scope and duration) to the run history, and get_test_run opens it later. This is the operation to use for "save the run I am looking at": run_tests(persist=true) also records, but it records the run it just executed, which is a different run against data that may have moved. Only a full run can be recorded, so the stored history compares like with like; a partial run is refused with a plain note. Recording the same run twice keeps one snapshot. Runs are remembered for the last few of this session, so record the one you just ran. Read-only with respect to the model.

Entitlement: Free entry point. Paid advice and selected automatic actions still need Pro where stated. Release class: supported. Risk class: standard.

Permission: the current target policy applies. Exact approvals and human-only actions remain as stated in this operation’s description.

Returns the detected endpoint/database plus every candidate model table with schema/entity, and echoes the effective endpoint after optional overrides. Detection follows the measure's dependency tables; multiple reachable SQL sources are reported ambiguous and never guessed. testConnection=true performs only select 1 against the effective endpoint/database (it does not run or change the human-accepted SQL); agent calls use the standard QueryData approval gate. Suggested next action is returned on refusal/failure. Read-only.

Entitlement: Free entry point. Paid advice and selected automatic actions still need Pro where stated. Release class: supported. Risk class: standard.

Permission: the current target policy applies. Exact approvals and human-only actions remain as stated in this operation’s description.

High precision by design: offline, an erroring query, a truncated comparison, or a missing oracle all come back Unverified: never a fabricated pass, never a fabricated 'wrong'. Pass questionId (from list_interview_questions; the outcome is also recorded on the saved question) or inlineJson (the same fields add_interview_question takes: a one-off, nothing persisted). An inline value-tier question may omit the oracle: it executes and comes back Unverified with the computed number in the detail: the confirm-and-record flow for seed candidates (confirm the number with the user, then add_interview_question with expectedValue; persistence still requires the oracle). Refusal tier: pass abstained=true if you (honestly) declined to answer, or attemptDax with the DAX you produced: producing any answer to an unanswerable question is the failure being caught. On SilentlyWrong the result carries fixRuleId + a plain fixHint naming the readiness fix that prevents it. Needs a live connection for the value/paraphrase tiers (open_live/open_local). A saved questionId that belongs to a different model (or is unattributed) is refused: its oracle was authored against another schema, so grading it here would be misleading.

Entitlement: Free entry point. Paid advice and selected automatic actions still need Pro where stated. Release class: supported. Risk class: standard.

Permission: the current target policy applies. Exact approvals and human-only actions remain as stated in this operation’s description.

Zero-setup ambient checks run on every model: per-relationship integrity (live: six data probes: orphaned foreign keys, duplicate/blank keys, row counts; an invalid relationship reparents orphans onto the hidden blank row so totals still tie while every per-member breakdown is silently wrong: any integrity Fail caps the grade at D) plus SQL-backed table row counts. Role filters and interview questions are not checked here, because reading a role's saved filter text cannot prove which rows a user sees; run_interview grades interview questions. Saved definitions (list_tests) run too: measureReconcile defs execute the full SQL-vs-DAX reconciliation. Verdicts: Pass | Fail | Suspect (downstream of a sibling failure, its root cause named, not a second failure) | NotVerifiable (couldn't check: offline, refused, insufficient coverage. never counted as a pass and never graded). read Health.Grade with Health.CoveragePct: the grade scores only what was decided; coverage says how much that was; RootFailures is the number that matters. Offline runs the checks it can decide without data and reports the rest NotVerifiable: connect_xmla/connect_local for the full run. scope: omit only and sections to run everything. The sections you can run are 'measures' (the saved checks), 'relationships' and 'tableCounts', and each one runs only itself: asking for relationships does not open a SQL connection, and asking for table counts does not probe relationships. An empty only or sections list is refused (it asks for nothing), and so is a section with no evaluator behind it. Agent-origin data probes are governed by the standard QueryData permission gate. persist=true also appends the run to history. Read-only.

Entitlement: Free entry point. Paid advice and selected automatic actions still need Pro where stated. Release class: supported. Risk class: standard.

Permission: the current target policy applies. Exact approvals and human-only actions remain as stated in this operation’s description.

Source='tests' saves the latest current-model Test report; source='workflow' saves a terminal Workflow run (sourceId may be omitted for the latest). The engine regenerates and verifies the artifact itself, then atomically writes canonical JSON plus matching html under .semanticus/evidence/<model-identity>. Re-saving the same source is idempotent. Nothing is written by export alone.

Entitlement: Free entry point. Paid advice and selected automatic actions still need Pro where stated. Release class: supported. Risk class: standard.

Permission: the current target policy applies. Exact approvals and human-only actions remain as stated in this operation’s description.

Kind='measureReconcile': paramsJson carries the ReconcileRequest (sql, groupBy, required blankPolicy, tolerances; see reconcile_measure, which is the same runner one-off), and filterDax narrows the DAX side the way the visual that produced it was narrowed (the SQL is never rewritten: narrow it yourself). kind='measureValue': paramsJson carries the MeasureValueRequest, where filterDax narrows the measure and expressionDax tests DAX the model does not hold yet (a scalar expression, or a complete evaluate query run verbatim) instead of the bound measure's own formula. Pass targetRef; the engine binds by LineageTag when present and otherwise creates a stable tests-sidecar identity so valid tagless measures follow renames. The returned bindingWarning is non-null if a name fallback was unavoidable. ground truth is ai-drafted, human-accepted: show the user the SQL and get acceptance before saving. A saved test is the user's suite, not the assistant's guess.

Entitlement: Free entry point. Paid advice and selected automatic actions still need Pro where stated. Release class: supported. Risk class: standard.

Permission: the current target policy applies. Exact approvals and human-only actions remain as stated in this operation’s description.

Without a mapping the check reads whatever the table's own partition reveals, and a table whose partition does not say enough reports 'not mapped to a SQL source yet' with nothing to fix it. Pass the model table, a sqlSourceId from list_sql_sources, and the schema and table name in the source. The mapping wins over anything detected from the model, replaces any mapping that table already had, and is stored with the model's test data (.semanticus/tests/) so it travels with the model. It supplies where to count, never a verdict: counts that differ without proven snapshot alignment stay 'couldn't check', not a failure. Next: run_tests to count, clear_table_source_mapping to undo it.

Entitlement: Free entry point. Paid advice and selected automatic actions still need Pro where stated. Release class: supported. Risk class: high.

Permission: the current target policy applies. Exact approvals and human-only actions remain as stated in this operation’s description.

Writes nothing to the suite. kind='measureValue' checks a measure against a trusted number; kind='measureReconcile' is the SQL-vs-DAX reconciliation. Offline the outcome is NotVerifiable. To see one measure's values across many contexts without a pass or fail, use probe_measure.

Entitlement: Free entry point. Paid advice and selected automatic actions still need Pro where stated. Release class: supported. Risk class: standard.

Permission: the current target policy applies. Exact approvals and human-only actions remain as stated in this operation’s description.

Kind is the operation: set_dax | set_description | set_measure_format | set_summarize_by | set_column_hidden | set_data_category | rename | set_synonyms | set_relationship_crossfilter | mark_date_table | set_ai_instructions (model-level: objRef is the model, after = the instructions text) | set_ai_data_schema (objRef a measure/column/table, after = 'Excluded' or 'Included') | delete (remove the object) | delete_if_unused (remove only if unused_objects' safe verdict still holds at apply time, recomputed server-side, else the item is skipped with the reason). An identical pending item (same ref+kind+target+after) is deduped: the existing item is returned, not re-added. For a set_dax rewrite, pass verifyGroupBy so apply_plan proves equivalence (and skips the item if results change). A set_dax item without verifyGroupBy is added as 'proposed' (opt-in, like a rename). It will not be applied unless you explicitly approve it via set_plan_item, and if applied it goes in unverified. Nothing is mutated until apply_plan. The change also appears on the Model Plan (get_model_plan) as a line on the item it serves, or on a new item owned by the assistant when no item serves it, so a person reads it in one list with everything else.

Entitlement: Free entry point. Paid advice and selected automatic actions still need Pro where stated. Release class: supported. Risk class: standard.

Permission: the current target policy applies. Exact approvals and human-only actions remain as stated in this operation’s description.

Only approved items are applied. Passing approvedIds narrows the approved set (it never applies rejected/proposed/needs_content items). DAX rewrites carrying a verify matrix are proven equivalent first and skipped if they'd change results or can't be proven; renames apply last. Returns a report: applied/skipped/failed counts + before→after BPA violations and AI-readiness grade. A single undo reverts everything; re-run propose_plan to iterate the tail. This is approve-all over the changes the Model Plan groups under its items: it applies every change you have approved, wherever it sits, in one step. It is the batch tool, so it applies what it can and reports what it could not; to approve one item's changes all-or-nothing instead, quoting the digest of exactly what you read, use model_plan_action(id, 'approve', digest).

Entitlement: Free entry point. Paid advice and selected automatic actions still need Pro where stated. Release class: supported. Risk class: critical.

Permission: the current target policy applies. Exact approvals and human-only actions remain as stated in this operation’s description.

To run the plan, pass no items and the plan's readyDigest from get_model_plan: every row whose doer is semanticus and whose stateWord is ready is applied in plan order, renames last, and a row with no changes worked out yet has them worked out and applied in the same step. It is refused with 'The plan moved since you read it. Read it again.' when the ready rows changed after your read, including a hold, a doer switch or a skip that did not move the model. A row that waits for another row is never in the batch; it turns ready on your next read. The cleanup row is the person's, so it is never in the ready set, and a cleanup line that is ticked only as a suggestion (its changeSet line says suggested: true) is never applied here until the person ticks, unticks or edits it. A row it cannot take (changes out of date, every change dropped) and a removal whose object is in use by the time the batch reaches it are left out and named in 'skipped'; the rest applies. Otherwise pass every item you want applied with the changeSet.digest the plan showed for it. Every kept line of every listed item goes in one batch: one undo puts all of them back. It is all or nothing across the whole list: if any item's digest is not the one on the plan, if any item is out of date, or if any kept line cannot be made, nothing is changed and the reply names the item and the line that stopped it. The one exception is a removal whose object is in use by the time the batch reaches it: it is left out and named in 'skipped', and the rest applies. An item whose changes are already in is not applied a second time. Lines you dropped are left out, as with model_plan_action('approve'), and an item with every line dropped adds nothing. Semanticus Pro: applying several items at once is running the plan, so a free session is refused and nothing is changed. On free, approve each item on its own with model_plan_action('approve'), which is free for a row that is yours or your assistant's to do and is the same check on one item. The answer is short and comes first: 'sentence' (what changed, and what was left out and why), then 'skipped' (what it left out, each with its reason; when that list is too long for the answer, 'skippedNote' counts the rest and says where they are), 'rows' (the new state of the rows it touched), 'readyDigest' (quote it to run the plan again) and 'counts' (rows in each state). It is not the plan: get_model_plan reads that.

Entitlement: Pro. The operation map admits the entry point, but its automatic Plan run refuses Free sessions. Release class: supported. Risk class: critical.

Permission: the current target policy applies. Exact approvals and human-only actions remain as stated in this operation’s description.

Freezes the measured values of the object's blast radius (the lineage-downstream measures, evaluated by reference over the summarizecolumns(groupBy) grid you supply) into a session-held baseline, and returns a captureId. After the edit, compare_baseline re-evaluates the same grid and reports exactly which numbers moved. Pass representative groupBy columns: a grand-total-only baseline is thin evidence (disclosed in the result). Over-cap measures are listed in Skipped, never silently dropped. certified totals (month-end close): pass a label (the period, e.g. "close fy26 · July") to also persist this capture beside the model as a certified baseline: one control total per call (objRef = the measure, filters = its stated context, includeDependents:false), all under the same label. That certified set survives the session, so a later refresh or edit that moves a signed-off number is caught by compare_baseline(label:…). needs a live connection (a baseline is measured values, not metadata).

Entitlement: Free entry point. Paid advice and selected automatic actions still need Pro where stated. Release class: supported. Risk class: standard.

Permission: the current target policy applies. Exact approvals and human-only actions remain as stated in this operation’s description.

Nothing in the model changes: a proposal never wrote anything. This is cancel-all over the changes the Model Plan groups under its items, so every item goes back to having nothing worked out for it. To drop one item's changes instead, use model_plan_action(id, 'cancel').

Entitlement: Free entry point. Paid advice and selected automatic actions still need Pro where stated. Release class: supported. Risk class: standard.

Permission: the current target policy applies. Exact approvals and human-only actions remain as stated in this operation’s description.

No client parameters.

That is an impact, not a skip). Safe=true only when everything is unchanged, nothing is missing/errored, and coverage wasn't truncated, honesty first. important: it reads the live model, so run it after the edit is deployed (deploy_live). An undeployed local edit is not validated (the result discloses session edits since capture). certified totals (month-end close): pass a label instead of a captureId to re-check that period's persisted certified figures. Each is re-evaluated at its stated context and reported held / moved (old→new) / missing / not-checkable, so drift from what was signed off is named. Detection, not prevention: it cannot stop a refresh or an out-of-tool edit, only report what moved. The verdict is recorded in the Verified Edits audit trail. needs the live connection.

Entitlement: Free entry point. Paid advice and selected automatic actions still need Pro where stated. Release class: supported. Risk class: standard.

Permission: the current target policy applies. Exact approvals and human-only actions remain as stated in this operation’s description.

Free, read-only.

Entitlement: Free entry point. Paid advice and selected automatic actions still need Pro where stated. Release class: supported. Risk class: critical.

Permission: the current target policy applies. Exact approvals and human-only actions remain as stated in this operation’s description.

Use it to see what still needs content (status 'needs_content') and what is approved/ready to apply. These are the same changes the Model Plan groups under its items (get_model_plan), as one flat list: this is the change-by-change view, that is the by-goal view, and there is one set of changes behind both.

Entitlement: Free entry point. Paid advice and selected automatic actions still need Pro where stated. Release class: supported. Risk class: standard.

Permission: the current target policy applies. Exact approvals and human-only actions remain as stated in this operation’s description.

No client parameters.

Each row says whether it is Undoable (pass its Id to undo_change toEntry), how many later changes that would also take back (LaterCount), its Verified word, and whether it is from before this session (never undoable). Filters: by, objectRef (one object), before (a row's When, to page back), limit (default 50). Kept=false means the model has nowhere to keep history yet (save it) and it lasts this session. Free. Read-only.

Entitlement: Free entry point. Paid advice and selected automatic actions still need Pro where stated. Release class: supported. Risk class: standard.

Permission: the current target policy applies. Exact approvals and human-only actions remain as stated in this operation’s description.

This tool keeps its older name so existing assistants and scripts still work. The switch only adds an audit record: every DAX write (update_measure, apply_dax_script, and creating a measure, calculated column, calculated table, calculation item or function) is checked offline whether it is on or off, and malformed syntax, arity problems and unknown table, column or measure references are refused before anything is written. When on, each single-object DAX write that passes is also recorded in the Verified Edits audit trail (list_verified_edits). The record states validity only, not an equivalence or drift proof. Session-scoped (resets on reconnect). 'available' is always true: Record verified edits is free.

Entitlement: Free entry point. Paid advice and selected automatic actions still need Pro where stated. Release class: supported. Risk class: critical.

Permission: the current target policy applies. Exact approvals and human-only actions remain as stated in this operation’s description.

No client parameters.

Each record links the previous by hash, so the trail self-checks: ChainIntact=false (with FirstBrokenSeq) means a record was edited, removed, or reordered after it was written. Read-only, free.

Entitlement: Free entry point. Paid advice and selected automatic actions still need Pro where stated. Release class: supported. Risk class: critical.

Permission: the current target policy applies. Exact approvals and human-only actions remain as stated in this operation’s description.

No client parameters.

Review with get_plan, then apply_plan applies the approved items as one undoable transaction. scope null = whole model; 'table:Name' = just that table. Everything you propose here also lands on the Model Plan (get_model_plan), filed under the item whose family and objects it serves, and a change no item serves opens an item of its own owned by the assistant. So this is the flat list of the same changes the Model Plan groups by goal, not a second list of different work.

Entitlement: Free entry point. Paid advice and selected automatic actions still need Pro where stated. Release class: supported. Risk class: standard.

Permission: the current target policy applies. Exact approvals and human-only actions remain as stated in this operation’s description.

Authoring a value moves the item to 'approved', except the opt-in kinds (rename, set_dax without a non-empty verify matrix, set_m), which land on 'proposed' after any value revision (even one made after an approval: the consent covered the old value) and need an explicit approve. e.g. set_plan_item('ci-0007', after='Total revenue recognised in the period.'). Use get_plan to see item ids + grounding first. this is also how you keep and drop one line of a Model Plan item's change set: approved=true keeps the line, approved=false drops it, and the item's changeSet.digest changes either way, so an approval given before you dropped a line can no longer be honoured. The line's planItemId on get_model_plan is the itemId to pass here. Keeping and dropping change nothing in the model and are free on every tier. Touching a line this way is what makes a cleanup line that arrived ticked as a suggestion (suggested: true) the person's own, and apply_plan from your door does not apply an untouched suggested line.

Entitlement: Free entry point. Paid advice and selected automatic actions still need Pro where stated. Release class: supported. Risk class: high.

Permission: the current target policy applies. Exact approvals and human-only actions remain as stated in this operation’s description.

This tool keeps its older name so existing assistants and scripts still work. It does not change what is refused: every DAX write is checked offline either way (malformed syntax, arity problems and unknown table, column or measure references are refused before anything is written). When on, each single-object DAX write that passes (update_measure and creating a measure, calculated column, calculated table, calculation item or function) is also recorded in the Verified Edits audit trail (list_verified_edits). The record states validity only, not an equivalence proof. This is the human's audit switch, free on any tier.

Entitlement: Free entry point. Paid advice and selected automatic actions still need Pro where stated. Release class: supported. Risk class: critical.

Permission: the current target policy applies. Exact approvals and human-only actions remain as stated in this operation’s description.

Ask the person to add it in Semanticus. What it does from their side: add one line to what this model should become. The plan already carries an aim per goal read off the rules; this adds the person's own, in their own words, and it is free on every tier because a target is their intent and not work the product did for them. The target appears in 'targets' on get_model_plan marked source 'you', and as one item in the section you file it under, in state 'target', owned by you, with the sub line 'nothing exists yet'. That item's diff is empty on purpose: nothing in the model matches the target yet, so the plan does not know the change and will not invent one. Where the item points is worked out from your own words by the same local word matching route_ask uses, and when the words name no clear place the item carries no link rather than a guessed one. Typing the same target twice is one target, refiled under whichever goal you named last. Changes nothing in the model. Kept beside the model's other settings, so it travels with the project; a model that was never saved and is not live has nowhere to keep it and says so.

Entitlement: Free entry point. Paid advice and selected automatic actions still need Pro where stated. Release class: supported. Risk class: standard.

Permission: the current target policy applies. Exact approvals and human-only actions remain as stated in this operation’s description.

Refused, with a sentence, while the person has paused tasks for you, while the task is paused, while it waits for another row on the Plan (the sentence names it), or when someone else holds it; the first claim wins. On success the result carries a 'claimToken' (shown once, keep it for complete_request or decline_request) and the task's packet: the words, the note, the brief, the objects, how done is known, and the grounding for each object. While you hold a claim, your edits count toward the task you claimed most recently; claiming a task you already hold switches back to it. Changes nothing in the model. A restart of Semanticus releases every claim, and the task is offered again.

Entitlement: Free entry point. Paid advice and selected automatic actions still need Pro where stated. Release class: supported. Risk class: standard.

Permission: the current target policy applies. Exact approvals and human-only actions remain as stated in this operation’s description.

A task Semanticus can check (words on named objects, a finding that should be gone) then reads 'Applied, checking' until the model proves it; a task only you can answer reads Done. Needs the claimToken claim_request gave you; an old or wrong token is refused and nothing changes. Completing the task clears it as your working-on task. For an ordinary task, make the edits first, then complete: completing changes nothing in the model. For a part of a draft (its brief says what to write), 'text' is the part: Semanticus checks it against the draft in the same call, and a failure keeps your claim, names each failed check and counts a try (the third hands the part to the person). The model changes only when the last part of an approved draft passes: then the draft's change runs as one undo step. For a step of a draft that you do (its brief says so), 'text' is the step's changes: a JSON list of {"op", "args"} from the change ops (create_measure, set_description and the rest), checked and applied as one undo step; send [] when the step needs no change.

Entitlement: Free entry point. Paid advice and selected automatic actions still need Pro where stated. Release class: supported. Risk class: standard.

Permission: the current target policy applies. Exact approvals and human-only actions remain as stated in this operation’s description.

The reason is required and is shown to them word for word; the task then reads 'Needs you' and they can give it to you again or set it aside. Needs the claimToken claim_request gave you. Clears it as your working-on task. Changes nothing in the model. For a part of a draft you may add 'choices': the person sees your reason and one button per choice that names a real object.

Entitlement: Free entry point. Paid advice and selected automatic actions still need Pro where stated. Release class: supported. Risk class: standard.

Permission: the current target policy applies. Exact approvals and human-only actions remain as stated in this operation’s description.

Each item carries the plain reasons it is there, the exact change it would make, what it touches, who does it (runs on its own, needs you, needs your assistant) and where to go to do it by hand. The plan is worked out fresh from the model every time it is asked for, so it can never disagree with the model. Items keep the same id as long as they are about the same thing. With no suggestion service, or on a free session, the same plan comes back in the order the rules found it, with no scores on it; 'orderedBy' says which you got. On Pro the value order is not there at once either: after a change, the first reads come back in rule order while Jev works the plan out ('jevPending' true, and 'orderedBy' says so), and the value order arrives on a later read, once Jev's own rows have landed and the ranking has been asked. Read again rather than assume the order you got is the value order. An item's reasons describe the work and never repeat one rule's message once per finding: one sentence for what was found and why it matters, one naming up to four examples and counting the rest, and one for what fixing it earns, however many findings are behind it. An item's 'diff' is the change itself, one line per case; it is empty when the change depends on a choice only the person can make, and 'touches' then says 'no change until you choose'. An empty diff never means there is nothing to do, and a finding restated is never a diff line. Each item's 'tables' says, per table, how many of the objects it touches are in that table, a short phrase naming that table's own cases, and that table's own change lines, so you can answer 'what is left on Sales' without re-reading the whole plan. 'history' is the plan's story, newest last, each entry carrying a kind (drawn, redrawn, done, started, skipped, unskipped, refused, proposed, approved, cancelled), who did it (you, assistant, plan, engine), the item, the revision and the detail lines. Each item also carries 'state' (target, gap, proposed, applied, done, skipped), which is how far along the propose, review, approve road it is, as distinct from 'status', which is whether the work can be acted on; applied means the change set landed and done means the gap is gone, and applied always comes first because nothing stamps done, the absence of the gap does. An item with changes worked out carries 'changeSet': every line with its before, its after, where it came from, and a 'kept' flag you turn on and off with set_plan_item; a 'digest' over the lines and the model revision they were worked out against, which model_plan_action('approve') makes you quote so you can only ever approve exactly what you read; 'refused', what was looked at and deliberately left out, with the reason; the cleanup row is a suggestion list for keeping the model lean: every unused column is a line, each with a reason, the ones Semanticus suggests arrive kept with suggested: true and the rest arrive dropped, and the person reviews the list before anything is removed; and 'stale' with a plain reason when the model or the file moved after the lines were worked out, in which case the changes are shown and never silently re-applied. Every row also carries three separate facts: 'source' (found, jev, words, you, workflow, repeat), 'doer' (semanticus, you, assistant; Jev is a source and never a doer) and 'stateWord' (done, set_aside, checking, on_it, paused, waiting, needs_you, ready), which the engine works out, with 'stateNote' beside it; also 'doneWhen' and 'doneText' (how done is known), 'objects', 'held', 'order' (your place in the person's order, on their tasks for you), 'note', 'reply', 'claim' and 'originLink'. 'who', 'status' and 'state' stay for one more release and are deprecated. Each row carries reads, writes, footprintKnown and waitsFor; a row with waitsFor waits for those rows. 'assistantQueue' is 'paused' when the person has paused tasks for you. Rows with ids starting 'pt-' are the person's tasks: adding one, pausing, reordering, changing who does it, the switch and notes are theirs to do in Semanticus, and no tool here does them. A plan too large to read in one result comes back short, and no read is ever more than 40,000 characters, on any model: each row with its state and its counts (touches), its detail left out ('detail' lists, per row, whether its detail is shown, how many lines it holds and where its next page starts), what a row waits for as the ids of the first few and a count (waitsForCount), and 'history' cut to the newest few entries with their first lines (detailNote says how many there are in all). On a plan with very many rows the rows themselves are cut to what fits, the reasons first and then the rows at the end of the list, and detailNote says how many are here. Pass item = a row's id to read that row's lines, tables, objects and every row it waits for, a page at a time (offset = that row's detail.next until it is null, max lines per page); item = rows to read the rows a page at a time (offset = the row to start at, max rows); item = history to read the plan's history a page at a time; or item = history/N to read entry N's lines a page at a time, when a history entry ends with '... and N more lines'. A plan that fits comes back whole. Reads only: nothing here changes the model. get_plan is the same changes as one flat change-by-change list, not different work.

Entitlement: Free findings in rule order. Pro adds ranking through the permitted advice points. Release class: supported. Risk class: standard.

Permission: the current target policy applies. Exact approvals and human-only actions remain as stated in this operation’s description.

This is the same list get_model_summary shows under its suggestions, so the two can never disagree. Reads only.

Entitlement: Free entry point. Paid advice and selected automatic actions still need Pro where stated. Release class: supported. Risk class: standard.

Permission: the current target policy applies. Exact approvals and human-only actions remain as stated in this operation’s description.

No client parameters.

The best-practice format findings that remain (a rule such as INTEGER_FORMATTING or PERCENTAGE_FORMATTING that wants an exact pattern, on a measure the plan will not write a format to itself) come last, each with the pattern the rule asks for in its message and no lift: set it with set_measure_format if it suits what the measure holds. Each item names the tool to use (set_description, rename_object, set_synonyms, create_linguistic_schema when the model has no synonym schema yet, set_ai_instructions, or set_measure_format), the object, what the rule asks, the points it adds; the page's 'grounding' holds each object's DAX, table and sibling names once, under the objectRef of its items. A page holds at most 40 items and at most 40,000 characters of JSON, whichever comes first, so a model with long measures gives fewer per page. Total and Kinds count the whole queue on every page; call again with offset = Next until Next is null. Read only: it writes nothing. Semanticus writes no words itself; you author them, and the Plan checks they land.

Entitlement: Free entry point. Paid advice and selected automatic actions still need Pro where stated. Release class: supported. Risk class: standard.

Permission: the current target policy applies. Exact approvals and human-only actions remain as stated in this operation’s description.

Each request carries its id (the task id, 'pt-...'), the person's words, their note, the brief a Send carried, the objects, how done is known, and 'independent': true when it changes nothing another offered task touches, so you can take it beside them; a task Semanticus cannot read says so and is never called independent. 'switch' is 'paused' when the person has paused tasks for you: then nothing is offered, and you carry on with what they ask you directly. Reads only. Take one with claim_request.

Entitlement: Free entry point. Paid advice and selected automatic actions still need Pro where stated. Release class: supported. Risk class: standard.

Permission: the current target policy applies. Exact approvals and human-only actions remain as stated in this operation’s description.

No client parameters.

Five actions. 'propose' works out the exact changes for that item and attaches them to it, and writes nothing: the row then carries every line with its before and after, a digest over them, and the model revision they were worked out against (get_model_plan with item = the row's id reads them); this answer carries the row's state and that digest. 'approve' applies the lines that are kept, all of them or none, in one undoable step, and you must quote the digest the plan showed you, so you can only ever approve the exact diff you read; it refuses if the model moved, if the file on disk moved, or if the lines changed after you read them, and if any kept line cannot be made the whole set is put back and nothing is changed. 'cancel' drops the worked-out changes and writes nothing. 'start' is the older single step: run a safe-fix row straight away with no review. 'done' is the person's: marking an aim or a task done is refused to you with a sentence (they do it in Semanticus); finish a task they gave you with complete_request instead. Keep and drop one line with set_plan_item (approved true or false) between propose and approve. Free on every tier: reading, propose, keep, drop and cancel. Semanticus Pro: start, and approve on a row the engine runs for you (who=auto), because that is the same run by another name. Approve on a row that is yours to do is free. Not every item can propose: descriptions are words and go to your assistant, and a row where the rules have an opinion and you have the answer names the page to decide it on instead of pretending to a change. The way back from an approve is one undo step on this session, not a durable checkpoint, and the result says so. The answer is short and comes first: 'sentence' (what changed, and what was left out and why), then 'skipped' (what it left out, each with its reason; when that list is too long for the answer, 'skippedNote' counts the rest and says where they are), 'rows' (the new state of the rows it touched), 'readyDigest' (quote it to run the plan again) and 'counts' (rows in each state). It is not the plan: get_model_plan reads that.

Entitlement: Free for manual rows. Pro is required for actions on rows Semanticus does for you. Read each action below. Release class: supported. Risk class: standard.

Permission: the current target policy applies. Exact approvals and human-only actions remain as stated in this operation’s description.

What it does from their side: take one target they added off the plan, and the item it became with it. Only the targets marked source 'you' can be removed; the aim per goal that comes from the rules is derived from the model on every call and is not yours or anyone's to delete. Takes either the target's own id or the id of the item it became, because you are shown both. Free, and it changes nothing in the model. To say you finished a target rather than that you no longer want it, use model_plan_action with action 'done' instead, which leaves it on the list marked done.

Entitlement: Free entry point. Paid advice and selected automatic actions still need Pro where stated. Release class: supported. Risk class: high.

Permission: the current target policy applies. Exact approvals and human-only actions remain as stated in this operation’s description.

It only ever names a place to go; it never runs anything and never changes anything. The first pass is word matching and needs no permission: the plain nouns and verbs each page is about are matched locally, nothing is sent anywhere, and a clear winner is answered on the spot with 'asked' false. Only a question the words cannot place is sent on, and only then does permission for this model matter. When a create word is recognised, 'create' names the thing being made. The words place a line on 'askdata' (Ask your data, where ask_data answers it) with no permission needed when it splits a number by something ('units per brand', 'customer count by city') or uses the page's own words ('total sales by region', 'top 10 products'); a line that only names a field ('what were sales last month') is placed only when permission lets the service be asked, and otherwise reads the no-match line. The same words can be a task and a question, and nothing here decides that for the person: whenever a line is placed on 'askdata', the answer carries both readings and neither is hidden. 'reading' names the top one, 'task' or 'question', and its place is on 'tool', 'area', 'confidence' and 'sentence' as usual; 'other' carries the second one, with its own 'kind', 'tool', 'area', 'confidence' and 'line' to show. order: the question reading is first only when the line plainly asks, which is read from the words of the line alone and never from what placed it. A line plainly asks when it ends in a question mark, opens with what, which, how, why, when, who or where (after a softening 'please', 'can you' or 'could we'), opens with 'show me' or 'list', or opens by asking for Ask your data by name ('ask your data please', 'go to ask your data'). Every other line reads the task first with Ask your data as the named second choice, the examples above included: 'total sales by region' reads task first. The task reading names the page the words say the work is done on, or no page at all, which is ordinary for a task and is exactly what the Plan's row shows for the same words; the person's Enter in the Plan's field keeps the words as a task whichever reading came first. confidence belongs to the reading it sits on: 'confidence' is the top reading's and 'other.confidence' the second one's. A reading the words or the engine's own numbers decided is 1, a reading the service decided carries the service's own number, and a task reading that names no page is 0. The service decides only the page it names: one page other than Ask your data (a single reading, on top), or the question reading on Ask your data, which sits on top or on 'other' by the order rule. So where 'asked' is true and the question is second, the service's number is 'other.confidence' and the top reads no tool at 0; that answer has a 'reading', which a line the service left unplaced does not. When nothing matches, 'fallback' is 'menu', 'sentence' is the plain line a person is shown and 'reason' carries why it went no further, which is for you and not for them. This is a Semanticus Pro feature; a session without Pro is refused before any word is matched.

Entitlement: Pro: Ask about this model. Release class: supported. Risk class: standard.

Permission: the current target policy applies. Exact approvals and human-only actions remain as stated in this operation’s description.

It stays on the plan, marked as set aside and showing the reason, so the next person to read it knows it was a decision rather than something missed. The reason is required. Kept beside the model's other settings, so it travels with the project and not with the person. Changes nothing in the model. The answer is short and comes first: 'sentence' (what changed, and what was left out and why), then 'skipped' (what it left out, each with its reason; when that list is too long for the answer, 'skippedNote' counts the rest and says where they are), 'rows' (the new state of the rows it touched), 'readyDigest' (quote it to run the plan again) and 'counts' (rows in each state). It is not the plan: get_model_plan reads that.

Entitlement: Free entry point. Paid advice and selected automatic actions still need Pro where stated. Release class: supported. Risk class: standard.

Permission: the current target policy applies. Exact approvals and human-only actions remain as stated in this operation’s description.

Changes nothing in the model. The answer is short and comes first: 'sentence' (what changed, and what was left out and why), then 'skipped' (what it left out, each with its reason; when that list is too long for the answer, 'skippedNote' counts the rest and says where they are), 'rows' (the new state of the rows it touched), 'readyDigest' (quote it to run the plan again) and 'counts' (rows in each state). It is not the plan: get_model_plan reads that.

Entitlement: Free entry point. Paid advice and selected automatic actions still need Pro where stated. Release class: supported. Risk class: standard.

Permission: the current target policy applies. Exact approvals and human-only actions remain as stated in this operation’s description.

It returns the overall grade and 0-100 score, per-category scores, coverage KPIs and the reasons any hard gate holds the grade down. detail='summary' adds the total finding count and counts by severity, fix kind and top rule, with no findings list, and stays small on any model. detail='full' (the default) also returns the findings in priority order, each tagged SafeFix (a deterministic fix exists), AiContent (you write it) or Proposal (a person should review it). live=true adds the live cardinality rules: it reads each column's distinct-value count (columnstatistics) from the connected model and checks the Copilot index limits (scale-qna-index, the 5,000,000-value index ceiling, and scale-hicard-column). live=true needs a live connection (connect_xmla or connect_local) and is refused without one. The filters narrow only the returned findings; scores and counts always cover the whole model, and a filter given with detail='summary' is refused. Read-only.

Entitlement: Free entry point. Paid advice and selected automatic actions still need Pro where stated. Release class: supported. Risk class: standard.

Permission: the current target policy applies. Exact approvals and human-only actions remain as stated in this operation’s description.

System='bpa' fixes a Best Practice Analyzer violation that bpa_scan marks CanAutoFix, for example a bi-directional relationship made single, or SummarizeBy set to None on a key column. system='air' fixes an AI-readiness SafeFix finding: vis-fk (hide a foreign-key column), fmt-summarize and summarize-dimension (SummarizeBy None), cat-geo (geographic data category), dim-primary-name (mark the column that names each row of a dimension as its row label). A finding with no deterministic fix is refused; get its instruction from get_fix_prompt with the same system, rule and object.

Entitlement: Free entry point. Paid advice and selected automatic actions still need Pro where stated. Release class: supported. Risk class: high.

Permission: the current target policy applies. Exact approvals and human-only actions remain as stated in this operation’s description.

System='air' applies the AI-readiness safe fixes: hide foreign-key columns, set SummarizeBy to None on key, relationship and identifier columns, set geographic data categories, and mark each dimension's row label. It renames nothing and writes no content. system='bpa' applies every Best Practice Analyzer auto-fix, repeating until a pass changes nothing, so a fix that creates another finding is cleared in the same call. For system='bpa', read 'unfixed' before you report the model clean: it lists each auto-fixable finding still present with the reason it survived (including one left alone because the person accepted the same change under another rule), and 'remaining' is how many. The batch did everything it said only when remaining is 0. A data type change, and a key on a table that already has one, are left for the person: they stay in 'unfixed' with the reason, and apply_fix still makes them one at a time. Content findings, and naming findings that need a new name, are left alone; get_fix_prompt covers those. For the readiness fixes plus a queue of content to write, use make_model_ai_ready.

Entitlement: Free entry point. Paid advice and selected automatic actions still need Pro where stated. Release class: supported. Risk class: high.

Permission: the current target policy applies. Exact approvals and human-only actions remain as stated in this operation’s description.

Seeds a culture's linguistic metadata; it does not change the Power BI Q&A setting. Call once before set_synonyms. Pass null culture for the default (en-us). On an existing schema it self-heals: entities bound to hidden or deleted objects are pruned, and any visible-name collisions Copilot can't tell apart are surfaced as a warning.

Entitlement: Free entry point. Paid advice and selected automatic actions still need Pro where stated. Release class: supported. Risk class: high.

Permission: the current target policy applies. Exact approvals and human-only actions remain as stated in this operation’s description.

Read-only. Call before editing instructions so an append/update doesn't clobber what's already there.

Entitlement: Free entry point. Paid advice and selected automatic actions still need Pro where stated. Release class: supported. Risk class: standard.

Permission: the current target policy applies. Exact approvals and human-only actions remain as stated in this operation’s description.

Read-only; the manage view behind the Studio 'Custom rules' panels.

Entitlement: Free entry point. Paid advice and selected automatic actions still need Pro where stated. Release class: supported. Risk class: standard.

Permission: the current target policy applies. Exact approvals and human-only actions remain as stated in this operation’s description.

No client parameters.

System='bpa' covers Best Practice Analyzer violations. system='air' covers AI-readiness content findings such as desc-measure, name-column, fmt-measure and syn-field, and includes grounding (DAX, table, sibling names). Read-only; it changes nothing.

Entitlement: Free entry point. Paid advice and selected automatic actions still need Pro where stated. Release class: supported. Risk class: standard.

Permission: the current target policy applies. Exact approvals and human-only actions remain as stated in this operation’s description.

Call this before set_description.

Entitlement: Free entry point. Paid advice and selected automatic actions still need Pro where stated. Release class: supported. Risk class: standard.

Permission: the current target policy applies. Exact approvals and human-only actions remain as stated in this operation’s description.

They travel with the model, the load is undoable, and they add to the built-in rules without ever replacing them. replace=false merges with the model's existing custom rules by id; replace=true makes them exactly this set. system='bpa' takes the standard BPARules.json schema: an array of {ID, Name, Category, Severity, Scope, Expression, FixExpression?, Description}. system='air' takes AI-readiness rules: {ID, Name?, Category (an existing readiness category: Naming, Descriptions, Synonyms, Relationships, Visibility, Formatting, CopilotLimits, DataAgentConfig or BestPractice), Severity? (Info, Medium, High or Critical; default Medium), Scope (the BPA scope words, such as Measure, Column, Table), Expression (the violation test, in the BPA expression language), AppliesTo? (the population; an empty population leaves the rule dormant), Message?, Description?, FixKind? (None, AiContent or Proposal; SafeFix is refused)}. Readiness rules are checked against the open model before anything is written, a built-in rule id is refused, and a custom rule cannot set a hard gate. Preview first with validate_rule, list what is saved with get_custom_rules, clear with reset_rules. Returns the new rule counts.

Entitlement: Free entry point. Paid advice and selected automatic actions still need Pro where stated. Release class: supported. Risk class: standard.

Permission: the current target policy applies. Exact approvals and human-only actions remain as stated in this operation’s description.

It writes no descriptions, names or synonyms itself: the queue is for you to author with set_description, rename_object or set_synonyms, or to stage for review with propose_plan.

Entitlement: Free entry point. Paid advice and selected automatic actions still need Pro where stated. Release class: supported. Risk class: standard.

Permission: the current target policy applies. Exact approvals and human-only actions remain as stated in this operation’s description.

Undoable.

Entitlement: Free entry point. Paid advice and selected automatic actions still need Pro where stated. Release class: supported. Risk class: standard.

Permission: the current target policy applies. Exact approvals and human-only actions remain as stated in this operation’s description.

Included=false excludes it (marks it Hidden in the lsdl); included=true restores it. Excluding noise (keys, sort/helper columns, intermediate measures) sharpens AI answers without hiding the field from report authors. Auto-enables the linguistic schema if needed; round-trip verified, live, a single undoable step. caveat: for a deployed model the lsdl edit syncs only after a model refresh. Persist with save_model. On Pro the Plan can propose these choices as rows (the ai.schema point, what Copilot sees); this op is the same change done by hand.

Entitlement: Free entry point. Paid advice and selected automatic actions still need Pro where stated. Release class: supported. Risk class: high.

Permission: the current target policy applies. Exact approvals and human-only actions remain as stated in this operation’s description.

Pass empty/null to clear. Max 10,000 chars (the service ignores anything beyond; the call is rejected if over). Auto-enables the linguistic schema if needed; the write is round-trip verified, live in the shared session, and a single undoable step. caveat: for a deployed model an lsdl edit syncs to the service only after a model refresh. It is not instant. Persist to disk with save_model.

Entitlement: Free entry point. Paid advice and selected automatic actions still need Pro where stated. Release class: supported. Risk class: high.

Permission: the current target policy applies. Exact approvals and human-only actions remain as stated in this operation’s description.

Total Sales -> ['revenue','turnover','sales amount']. Auto-enables the linguistic schema if needed. Helps Copilot and data agents map natural language to the right field. The list replaces the object's own synonyms: a word left out is removed. The object's name term and Power BI's own generated words are kept, except what Power BI still records from an old name, which a list that leaves it out clears. Hidden targets are refused, and entities bound to hidden or deleted objects are pruned automatically. Two visible fields that read as the same word do not make a write fail; the readiness scan reports them as a Copilot ambiguity.

Entitlement: Free entry point. Paid advice and selected automatic actions still need Pro where stated. Release class: supported. Risk class: high.

Permission: the current target policy applies. Exact approvals and human-only actions remain as stated in this operation’s description.

Kind='bpa' or 'readiness'; rules = inline JSON (one rule object or an array, the same schema the matching load op accepts). Per rule you get: compile errors from the real expression parser (with the failing position), the would-be Applicable population, the violation count, the first few flagged object names, and a dormant flag (empty population = the rule would never move a score). Save with load_rules (system='bpa' or 'air') once valid.

Entitlement: Free entry point. Paid advice and selected automatic actions still need Pro where stated. Release class: supported. Risk class: standard.

Permission: the current target policy applies. Exact approvals and human-only actions remain as stated in this operation’s description.

Fetches each report's pbir via the Fabric 'Get Report Definition' API, reconciles its field usage to the open model by name, and recomputes safe-to-remove excluding any field a report uses. reportIds empty = every non-paginated report in the workspace. Behaviour is read-only (Semanticus never modifies the report), but getDefinition requires a write-capable Fabric scope (Item.ReadWrite.All / Report.ReadWrite.All) + the Contributor role, so you must pass consent=true to acknowledge that (tell the user first). Per-report failures (paginated/rdl, a report blocked by an encrypted sensitivity label, a download error) are reported as unreadable with a reason, never silently dropped, so 'safe' is never overstated. A model must be open and must be the one these reports bind to.

Entitlement: Free entry point. Paid advice and selected automatic actions still need Pro where stated. Release class: deferred. Risk class: standard.

Permission: the current target policy applies. Exact approvals and human-only actions remain as stated in this operation’s description.

This closes the model-only blind spot where a descriptive column a report displays (but no measure references) wrongly looks unused. Report field references are reconciled to the open model by name (the model must be the one the reports bind to). Offline (local files); the cloud 'Get Report Definition' path will reuse the same parser. Legacy .pbix / rdl are not parsed yet (reported as unreadable → result stays model-only with a caveat).

Entitlement: Free entry point. Paid advice and selected automatic actions still need Pro where stated. Release class: supported. Risk class: standard.

Permission: the current target policy applies. Exact approvals and human-only actions remain as stated in this operation’s description.

Ids empty = every chosen report. Local report folders need no sign-in. A published report needs consent=true: reading its definition asks Fabric for permission that can also edit reports, even though Semanticus only reads it. A report that cannot be read is reported as couldNotBeFullyChecked with the reason, never silently dropped, so 'no use found' is never overstated.

Entitlement: Free entry point. Paid advice and selected automatic actions still need Pro where stated. Release class: supported. Risk class: standard.

Permission: the current target policy applies. Exact approvals and human-only actions remain as stated in this operation’s description.

Use it to trace where a field comes from and what flows from it. note (Phase 1): model-only. Published-report field usage is not yet included (see .Caveat).

Entitlement: Free entry point. Paid advice and selected automatic actions still need Pro where stated. Release class: supported. Risk class: standard.

Permission: the current target policy applies. Exact approvals and human-only actions remain as stated in this operation’s description.

No client parameters.

Returns the transitive tom blast radius, the affected reports and visuals from this model'S report scope (set_report_scope + check_reports), the current-model saved checks to run afterwards, what was checked and every gap as a plain sentence, one verdict token, and the exact next op. scope='modelAndReports' is the honest default: with no report checked it returns Unknown/NeedsReview, never green, and the next op is check_reports. scope='model' deliberately proves only model-internal impact. modelImpact holds the model's own blast radius: every DAX dependent, plus for a column the relationships, hierarchies and sort-by columns that use it, each with its depth, and counts by kind. scope='model' returns only that half. reportPaths folds a local report folder into the scope for this one call. Free and read-only; it omits report contents and does not apply the change.

Entitlement: Free entry point. Paid advice and selected automatic actions still need Pro where stated. Release class: supported. Risk class: standard.

Permission: the current target policy applies. Exact approvals and human-only actions remain as stated in this operation’s description.

State per report: checked | couldNotBeFullyChecked | notChecked | needsChecking (the model changed after the report was read). The choices are saved with the model and come back on reopen as notChecked, because a saved choice is not evidence that anything was read; the check results live in this session. impact_assessment, unused_objects, remove_safe_objects and the apply-time recheck of a delete_if_unused plan item all read this same scope. Free and read-only.

Entitlement: Free entry point. Paid advice and selected automatic actions still need Pro where stated. Release class: supported. Risk class: standard.

Permission: the current target policy applies. Exact approvals and human-only actions remain as stated in this operation’s description.

No client parameters.

The datasetId tells you which reports bind to a given semantic model. Match it to the open model to find the reports that use it. reportType 'PaginatedReport' = rdl (its field usage can't be parsed). Live read against api.powerbi.com using your Entra identity. Read-only. Pair with analyze_cloud_reports to close the safe-to-remove blind spot using real published-report usage.

Entitlement: Free entry point. Paid advice and selected automatic actions still need Pro where stated. Release class: supported. Risk class: standard.

Permission: the current target policy applies. Exact approvals and human-only actions remain as stated in this operation’s description.

The safe set is recomputed server-side and each item re-verified at apply time. An item whose status changed since your scan (something now references it) is skipped with the reason, never deleted stale. refs (optional) narrows the sweep to those candidates from unused_objects/analyze_reports; omit to remove every currently-safe item. A 'visibleAttribute' item is never swept, named or not; its skip reason names the reviewed route that does remove it. reportPaths (optional, local pbir, same forms as analyze_reports) makes the verification report-aware, so a field a report displays is never removed. Returns removed[] / skipped[{ref,reason}] / count; writes one audit record with the evidence; undo_change reverses the whole sweep in one step.

Entitlement: Free entry point. Paid advice and selected automatic actions still need Pro where stated. Release class: supported. Risk class: high.

Permission: the current target policy applies. Exact approvals and human-only actions remain as stated in this operation’s description.

Replaces the whole list. Each choice is either kind='published' (with workspaceId, reportId and name, from list_workspaces + list_reports) or kind='local' (with path: a Power BI project (.pbip) or its .Report folder; pbix is not supported). Saves the choice beside the model; nothing is read here, so every choice comes back as notChecked until check_reports runs. Read the reports with check_reports.

Entitlement: Free entry point. Paid advice and selected automatic actions still need Pro where stated. Release class: supported. Risk class: high.

Permission: the current target policy applies. Exact approvals and human-only actions remain as stated in this operation’s description.

Conservative: a column used by a relationship, hierarchy, sort-by, or RLS is never listed. Verdict per item: 'safe' (nothing references it), 'usedByUnusedOnly' (referenced only by other unused/dead objects, surfaced with BlockedBy), 'caution' (a report reading could not rule it out), 'visibleAttribute' (a visible column of a dimension table with no model reference: people query by it in reports, Q&A and Copilot, so a sweep never takes it. That is not a refusal of the person: a reviewed removal still removes it, through add_plan_item kind 'delete_if_unused' + apply_plan, or delete_object). It reads this model's report scope (set_report_scope, check_reports): with no report chosen it is a model-only sweep, so a field used only by a published report still appears as 'safe'; when every chosen report has been checked, fields those reports use are excluded; while any chosen report is still unchecked, every would-be 'safe' item comes back as 'caution'. .Caveat names the basis the list carries.

Entitlement: Free entry point. Paid advice and selected automatic actions still need Pro where stated. Release class: supported. Risk class: standard.

Permission: the current target policy applies. Exact approvals and human-only actions remain as stated in this operation’s description.

No client parameters.

Lets you bulk-edit many measures/calc columns/calc items in one document and commit together. All-or-nothing: a missing ref or invalid DAX refuses the whole script and writes nothing. Returns {applied[], skipped[]}.

Entitlement: Free entry point. Paid advice and selected automatic actions still need Pro where stated. Release class: supported. Risk class: high.

Permission: the current target policy applies. Exact approvals and human-only actions remain as stated in this operation’s description.

Existing table/role documents and individually scripted measures round-trip through tom's native partial-update path in one undoable batch. Measure ownership is pinned by the emitted // @object ref; changing its header/parent is refused (use rename_object/create_measure for structural changes). A brand-new top-level object is skipped (use its typed create op). Returns {applied[], skipped[]}.

Entitlement: Free entry point. Paid advice and selected automatic actions still need Pro where stated. Release class: supported. Risk class: high.

Permission: the current target policy applies. Exact approvals and human-only actions remain as stated in this operation’s description.

Returns the new ref ('column:Table/Name'). Undoable.

Entitlement: Free entry point. Paid advice and selected automatic actions still need Pro where stated. Release class: supported. Risk class: high.

Permission: the current target policy applies. Exact approvals and human-only actions remain as stated in this operation’s description.

A date dimension or bridge). Returns the table ref ('table:Name'). Undoable.

Entitlement: Free entry point. Paid advice and selected automatic actions still need Pro where stated. Release class: supported. Risk class: high.

Permission: the current target policy applies. Exact approvals and human-only actions remain as stated in this operation’s description.

Add calculation items with create_calculation_item. Requires compatibility level >= 1470. Returns the new ref ('table:Name'). Undoable.

Entitlement: Free entry point. Paid advice and selected automatic actions still need Pro where stated. Release class: supported. Risk class: high.

Permission: the current target policy applies. Exact approvals and human-only actions remain as stated in this operation’s description.

The expression typically uses selectedmeasure(), e.g. 'calculate(selectedmeasure(), datesytd(...))'. Returns the new ref. Undoable.

Entitlement: Free entry point. Paid advice and selected automatic actions still need Pro where stated. Release class: supported. Risk class: high.

Permission: the current target policy applies. Exact approvals and human-only actions remain as stated in this operation’s description.

DataType: String|Int64|Decimal|Double|DateTime|Boolean (default String). sourceColumn is the underlying physical column name (defaults to the column name). Returns the new ref ('column:Table/Name'). Undoable.

Entitlement: Free entry point. Paid advice and selected automatic actions still need Pro where stated. Release class: supported. Risk class: high.

Permission: the current target policy applies. Exact approvals and human-only actions remain as stated in this operation’s description.

Dry run by default: commit=false returns the exact request and changes nothing; commit=true creates it. aiInstructions capped at 15000 chars.

Entitlement: Free entry point. Paid advice and selected automatic actions still need Pro where stated. Release class: experimental. Risk class: high.

Permission: the current target policy applies. Exact approvals and human-only actions remain as stated in this operation’s description.

Needs compatibility level >= 1400. Referenced by name from create_directlake_table or inside an M expression. Returns the data source name. Undoable. (Credentials are supplied out-of-band at refresh. None are stored here.).

Entitlement: Free entry point. Paid advice and selected automatic actions still need Pro where stated. Release class: supported. Risk class: high.

Permission: the current target policy applies. Exact approvals and human-only actions remain as stated in this operation’s description.

Needs compatibility level >= 1604. Bind it to a source created first with create_named_expression (M to the SQL endpoint) or create_data_source (structured), referenced by name. Returns the table ref ('table:Name'). Undoable.

Entitlement: Free entry point. Paid advice and selected automatic actions still need Pro where stated. Release class: supported. Risk class: high.

Permission: the current target policy applies. Exact approvals and human-only actions remain as stated in this operation’s description.

Builds the full Power-BI-Desktop-identical structure (the 3 columns, the ParameterMetadata marker, sort-by, hidden plumbing columns, and the field-switch key). Items are measure/column refs in display order; each label defaults to the object's name. Requires a Power BI-mode model at compatibility level >= 1400. Returns the table ref ('table:Name'). Undoable.

Entitlement: Free entry point. Paid advice and selected automatic actions still need Pro where stated. Release class: supported. Risk class: high.

Permission: the current target policy applies. Exact approvals and human-only actions remain as stated in this operation’s description.

The expression is a lambda, e.g. '(x: int64, y: int64) => x + y'. Names cannot contain spaces. Requires model compatibility level >= 1702 (call set_compatibility_level first if needed). Returns the new ref ('function:Name'). Undoable.

Entitlement: Free entry point. Paid advice and selected automatic actions still need Pro where stated. Release class: supported. Risk class: high.

Permission: the current target policy applies. Exact approvals and human-only actions remain as stated in this operation’s description.

Returns the new ref. Undoable.

Entitlement: Free entry point. Paid advice and selected automatic actions still need Pro where stated. Release class: supported. Risk class: high.

Permission: the current target policy applies. Exact approvals and human-only actions remain as stated in this operation’s description.

Dry run by default: commit=false reports the owned paths, changed files and whether the model will be saved. Pass commit=true after review. Refuses a detached branch, conflicts, or unrelated staged files instead of sweeping them into the commit. The result names the exact checkpoint hash; use restore_history_checkpoint to preview a restore.

Entitlement: Free entry point. Paid advice and selected automatic actions still need Pro where stated. Release class: supported. Risk class: high.

Permission: the current target policy applies. Exact approvals and human-only actions remain as stated in this operation’s description.

From a Fabric SQL endpoint via Sql.Database(...). Bundles the table + its M partition (no orphaned placeholder). Add columns with create_column. Returns the table ref ('table:Name'). Undoable.

Entitlement: Free entry point. Paid advice and selected automatic actions still need Pro where stated. Release class: supported. Risk class: high.

Permission: the current target policy applies. Exact approvals and human-only actions remain as stated in this operation’s description.

Returns the new object ref. Undoable. Pass displayFolder to create it already filed ('Parent\Child' nests). Create + folder is one undo step.

Entitlement: Free entry point. Paid advice and selected automatic actions still need Pro where stated. Release class: supported. Risk class: high.

Permission: the current target policy applies. Exact approvals and human-only actions remain as stated in this operation’s description.

Returns the expression name. Undoable.

Entitlement: Free entry point. Paid advice and selected automatic actions still need Pro where stated. Release class: supported. Risk class: high.

Permission: the current target policy applies. Exact approvals and human-only actions remain as stated in this operation’s description.

Add members with set_perspective_member; delete with delete_object and rename with rename_object. Returns the new ref ('perspective:Name'). Undoable.

Entitlement: Free entry point. Paid advice and selected automatic actions still need Pro where stated. Release class: supported. Risk class: high.

Permission: the current target policy applies. Exact approvals and human-only actions remain as stated in this operation’s description.

CrossFilter: OneDirection|BothDirections (optional). isActive optional. Returns the new ref ('relationship:Name'). Undoable.

Entitlement: Free entry point. Paid advice and selected automatic actions still need Pro where stated. Release class: supported. Risk class: high.

Permission: the current target policy applies. Exact approvals and human-only actions remain as stated in this operation’s description.

ModelPermission defaults to None (Read is auto-applied once you add an RLS filter). Then use set_table_permission to add per-table row filters and set_role_member to add members. Undoable.

Entitlement: Free entry point. Paid advice and selected automatic actions still need Pro where stated. Release class: supported. Risk class: high.

Permission: the current target policy applies. Exact approvals and human-only actions remain as stated in this operation’s description.

Add columns with create_column/create_calculated_column; it needs a partition before it can be deployed/refreshed. Returns the new ref ('table:Name'). Undoable.

Entitlement: Free entry point. Paid advice and selected automatic actions still need Pro where stated. Release class: supported. Risk class: high.

Permission: the current target policy applies. Exact approvals and human-only actions remain as stated in this operation’s description.

Deleting an absent ref is a net-zero no-op. DAX references to a deleted object are not auto-rewritten. They will error; check dependents first. Undoable.

Entitlement: Free entry point. Paid advice and selected automatic actions still need Pro where stated. Release class: supported. Risk class: high.

Permission: the current target policy applies. Exact approvals and human-only actions remain as stated in this operation’s description.

Same rules as delete_object: an absent ref is skipped, DAX references are not auto-rewritten. One undo restores every object that this call removed. Use delete_object for a single ref.

Entitlement: Free entry point. Paid advice and selected automatic actions still need Pro where stated. Release class: supported. Risk class: high.

Permission: the current target policy applies. Exact approvals and human-only actions remain as stated in this operation’s description.

Calculated tables / field parameters), or calculation group (a deep clone). newName optional (collision-safe auto-name if omitted). targetRef optional: a 'table:' ref to clone a measure onto another table, or a calc group's 'table:' ref to clone a calculation item onto another group. Columns/hierarchies stay on their own table; tables duplicate at model scope. Returns the new ref. Undoable.

Entitlement: Free entry point. Paid advice and selected automatic actions still need Pro where stated. Release class: supported. Risk class: standard.

Permission: the current target policy applies. Exact approvals and human-only actions remain as stated in this operation’s description.

All DAX references to it are automatically rewritten (safe rename). Returns the new ref. Undoable. (Report-layer bindings outside the model can't be updated. Review for renames used in reports.)

Entitlement: Free entry point. Paid advice and selected automatic actions still need Pro where stated. Release class: supported. Risk class: high.

Permission: the current target policy applies. Exact approvals and human-only actions remain as stated in this operation’s description.

Needed for newer features. DAX user-defined functions require 1702+. Refuses to lower it. Undoable within the session, but re-deploying an upgraded model to an older runtime may fail.

Entitlement: Free entry point. Paid advice and selected automatic actions still need Pro where stated. Release class: supported. Risk class: high.

Permission: the current target policy applies. Exact approvals and human-only actions remain as stated in this operation’s description.

Target is the open model when you omit targetFile and targetEndpoint (undoable merge), a model file on disk (targetFile), or a published model on an XMLA endpoint (targetEndpoint + targetDatabase). source defaults to the open model; pass sourceFile or sourceGitRef to merge from elsewhere. dry run by default (commit=false reports what would apply, including any deletes). Pass commit=true to write. A write to a file or into the open model needs the review token the preview returned. selectedRefs limits to specific object refs (e.g. ['measure:Sales/Total']); omit to apply all differences. Pushing to a published model runs an always-on drift guard: if a target object changed since the diff, the push is refused unless you pass overrideReason (recorded in the audit trail). Every committed push to a published model first writes a restore point (returned as restorePointId) so it can be undone with rollback_push; if a push contains deletes and the restore point cannot be written, the push is refused. Next: run model_diff first to see the object refs; on a drift refusal, re-run model_diff to see the new state before overriding.

Entitlement: Free entry point. Paid advice and selected automatic actions still need Pro where stated. Release class: supported. Risk class: high.

Permission: the current target policy applies. Exact approvals and human-only actions remain as stated in this operation’s description.

Copy a measure from a teammate's model into yours. dry run by default (commit=false reports what would copy, what already exists (would overwrite), and what can't). Pass commit=true to apply as one undoable batch. Source is a model on disk (sourceFile) or a git ref of the open model's repo (gitRef). refs are object refs like ['measure:Sales/Margin %']. Supported into the open model: measures, calculated columns, and calculation items (into an existing group). For a whole table or calculation group, use apply_model_diff to a file target instead.

Entitlement: Free entry point. Paid advice and selected automatic actions still need Pro where stated. Release class: supported. Risk class: standard.

Permission: the current target policy applies. Exact approvals and human-only actions remain as stated in this operation’s description.

'measure:Sales/Total Sales'), display name, kind, and whether it has children. This is the same reference tree the vs Code compare/cherry-pick picker shows. Use it to enumerate the model's authorable objects in one call, to pick refs for model_diff / apply_model_diff / cherry_pick, or as a cheap structural map of what exists. Defaults to the open model; pass sourceFile to read a model on disk instead. Read-only.

Entitlement: Free entry point. Paid advice and selected automatic actions still need Pro where stated. Release class: supported. Risk class: standard.

Permission: the current target policy applies. Exact approvals and human-only actions remain as stated in this operation’s description.

To change state, pass a branch name and set create=true, checkout=true, or both. If switching changes the on-disk model, the result names the model path to reopen and flags modelReloadNeeded.

Entitlement: Free entry point. Paid advice and selected automatic actions still need Pro where stated. Release class: supported. Risk class: high.

Permission: the current target policy applies. Exact approvals and human-only actions remain as stated in this operation’s description.

Refuses while the open model has unsaved edits. If the on-disk model changes, the result names the model path to reopen and flags modelReloadNeeded.

Entitlement: Free entry point. Paid advice and selected automatic actions still need Pro where stated. Release class: supported. Risk class: high.

Permission: the current target policy applies. Exact approvals and human-only actions remain as stated in this operation’s description.

A relative target stays inside the current workspace, and the target must not already exist. Returns the model path to open with open_model. Uses the user's git credential helper.

Entitlement: Free entry point. Paid advice and selected automatic actions still need Pro where stated. Release class: supported. Risk class: high.

Permission: the current target policy applies. Exact approvals and human-only actions remain as stated in this operation’s description.

Dry run by default (commit=false returns the file set that would be committed and whether the model needs saving first; nothing changes). Pass commit=true to actually commit. It saves the open model's unsaved edits to disk first, stages the model folder, then commits. Returns the commit hash + files.

Entitlement: Free entry point. Paid advice and selected automatic actions still need Pro where stated. Release class: supported. Risk class: high.

Permission: the current target policy applies. Exact approvals and human-only actions remain as stated in this operation’s description.

Pass a file path to scope it; staged=true for the staged diff. Read-only.

Entitlement: Free entry point. Paid advice and selected automatic actions still need Pro where stated. Release class: supported. Risk class: high.

Permission: the current target policy applies. Exact approvals and human-only actions remain as stated in this operation’s description.

Read-only.

Entitlement: Free entry point. Paid advice and selected automatic actions still need Pro where stated. Release class: supported. Risk class: high.

Permission: the current target policy applies. Exact approvals and human-only actions remain as stated in this operation’s description.

Refuses while the open model has unsaved edits. If the on-disk model changes, the result names the model path to reopen and flags modelReloadNeeded.

Entitlement: Free entry point. Paid advice and selected automatic actions still need Pro where stated. Release class: supported. Risk class: high.

Permission: the current target policy applies. Exact approvals and human-only actions remain as stated in this operation’s description.

No client parameters.

Dry run by default (confirm=false reports what would push). Pass confirm=true to push. Uses the user's git credential helper. The engine never handles git credentials.

Entitlement: Free entry point. Paid advice and selected automatic actions still need Pro where stated. Release class: supported. Risk class: high.

Permission: the current target policy applies. Exact approvals and human-only actions remain as stated in this operation’s description.

Also flags whether the open model has unsaved in-memory edits (git_commit saves them to disk first). Read-only.

Entitlement: Free entry point. Paid advice and selected automatic actions still need Pro where stated. Release class: supported. Risk class: high.

Permission: the current target policy applies. Exact approvals and human-only actions remain as stated in this operation’s description.

No client parameters.

Checkpoints are ordinary model-scoped commits, so they survive closing the session and cloning the repository. Returns supported=false for a model outside git; session undo remains the short-term alternative. Read-only.

Entitlement: Free entry point. Paid advice and selected automatic actions still need Pro where stated. Release class: supported. Risk class: standard.

Permission: the current target policy applies. Exact approvals and human-only actions remain as stated in this operation’s description.

What you changed since the last commit). Returns an object-level diff (per table/column/measure/relationship/role/...: Create/Update/Delete) with each object's before/after text. Pass a different gitRef to compare against a branch/commit, or targetFile to compare against a model on disk, or sourceFile to compare two files. Read-only.

Entitlement: Free entry point. Paid advice and selected automatic actions still need Pro where stated. Release class: supported. Risk class: standard.

Permission: the current target policy applies. Exact approvals and human-only actions remain as stated in this operation’s description.

Dry run by default: restore=false names the model paths and safety sequence. Pass restore=true only after review. The engine first commits the current model as a rescue checkpoint, restores the selected model paths, reopens the model, and commits the restored state. Returns all three hashes so the action stays reversible. Live published models use rollback_push and its restore points instead.

Entitlement: Free entry point. Paid advice and selected automatic actions still need Pro where stated. Release class: supported. Risk class: high.

Permission: the current target policy applies. Exact approvals and human-only actions remain as stated in this operation’s description.

Each item: {change: Added|Removed|TypeChanged, column, dataType? (tom type, for Added/TypeChanged), sourceColumn? (physical name for Added, defaults to column)}. Items that can't apply (a calculated column, a name clash, an unknown column) are skipped with a reason. The accepted changes still commit atomically. Returns per-kind counts + the applied/skipped lists.

Entitlement: Free entry point. Paid advice and selected automatic actions still need Pro where stated. Release class: supported. Risk class: high.

Permission: the current target policy applies. Exact approvals and human-only actions remain as stated in this operation’s description.

Each item carries a stable id for per-item selection. By default it probes the live source (like get_source_schema); pass sourceColumns to diff a supplied/synthesized schema instead (the offline/manual path, no connection needed). Calculated columns are never flagged (they're DAX-derived, not source-driven). Returns Reachable=false + a message when the live source can't be read. Feed the accepted items to apply_schema_update.

Entitlement: Free entry point. Paid advice and selected automatic actions still need Pro where stated. Release class: supported. Risk class: standard.

Permission: the current target policy applies. Exact approvals and human-only actions remain as stated in this operation’s description.

Returns enabled=false if no policy is defined. Read-only.

Entitlement: Free entry point. Paid advice and selected automatic actions still need Pro where stated. Release class: supported. Risk class: standard.

Permission: the current target policy applies. Exact approvals and human-only actions remain as stated in this operation’s description.

Returns the source columns mapped to tom types (String|Int64|Decimal|Double|DateTime|Boolean). Needs a reachable source: for an offline snapshot, a non-SQL source, or a native-query partition it returns Reachable=false with a clear message (never throws). Use diff_schema to compare against the model.

Entitlement: Free entry point. Paid advice and selected automatic actions still need Pro where stated. Release class: supported. Risk class: standard.

Permission: the current target policy applies. Exact approvals and human-only actions remain as stated in this operation’s description.

Read-only. Use it to pick a type for refresh_partition.

Entitlement: Free entry point. Paid advice and selected automatic actions still need Pro where stated. Release class: supported. Risk class: standard.

Permission: the current target policy applies. Exact approvals and human-only actions remain as stated in this operation’s description.

No client parameters.

Leaves RangeStart/RangeEnd and the partition M intact. Deterministic, undoable.

Entitlement: Free entry point. Paid advice and selected automatic actions still need Pro where stated. Release class: supported. Risk class: high.

Permission: the current target policy applies. Exact approvals and human-only actions remain as stated in this operation’s description.

Set autoWire=true to create or repair RangeStart/RangeEnd and append the required half-open date filter without replacing existing M steps. Leave it false to validate and refuse when prerequisites are missing. Metadata only; it does not refresh data. Deterministic and undoable. Existing-policy fields are preserved when omitted.

Entitlement: Free entry point. Paid advice and selected automatic actions still need Pro where stated. Release class: supported. Risk class: high.

Permission: the current target policy applies. Exact approvals and human-only actions remain as stated in this operation’s description.

Pass 'question' in plain words ("net sales by store region for 2025, top 10 products") and Jev reads which measures, columns, filters and top N it asks for from the model's names, column types, folders and descriptions; or pass the parts yourself (group_by, measures, filters, top, top_by, lowest, sort_by, descending), which are taken as said. To ask again with one part changed, pass the answer's 'parts' object back as 'parts' (either name style works) with the one part changed, or pass the snake_case parameters; a name ask_data does not take is refused, never ignored. Filter values are matched on this computer against the connected model's values: pass the words ("Coast", "2025") and Semanticus finds the real value, or says none matches. Before running it checks that every column reaches every set of rows each measure counts through active relationships, that each filter value is really in the data and the filters can all hold, that a sort names a split column or a shown measure, and that the DAX checks out, and it caps the rows (max_rows, default 500). Returns outcome (answered, built, needs-you, refused, error), a 'statement' (one plain sentence, the first thing to read, saying what the answer shows: "This shows Net Sales for each Region, where Year is 2025."), a plain reading line such as "Net Sales by Region, Year = 2025, top 10 by Net Sales", who read it (jev, or you when the parts were given), the parts it settled, the DAX, each check, any unsure lines, and the rows. Nothing runs when a part was not sure: 'unsure' says what to settle. When the run itself fails, outcome is error, the rows check says so in one plain line and 'error' carries the server's own words. When Jev was torn between two good readings of one part, the top one runs and 'others' names the other ("Showing Net Sales. You may have meant Units."), with that reading's line and its whole 'parts', ready to pass back as 'parts' to run it instead. Before a number is shown, a question read from words is checked once more: does the reading give the numbers the question asks for? When it does not (a change, a share, things with no sales, a forecast, a period it left out), when Jev is not sure, or when the check could not be made, nothing runs, 'unsure' says what is missing, and 'others' offers the reading as it is (anyway: true) with its 'parts': pass those parts back as 'parts' to run it as given, or ask again in other words. Read-only; needs a connected model with data to match values and run. Returns rows, so the agent policy's QueryData rule applies to the connected target. Every ask is kept in the model's question history (list_data_questions), without its rows. Ask your data is a Semanticus Pro feature.

Entitlement: Pro: Ask your data. Release class: supported. Risk class: standard.

Permission: the current target policy applies. Exact approvals and human-only actions remain as stated in this operation’s description.

Mode='warm' (the default) runs it 'runs' times without clearing the cache and returns the first run, the fastest and median warm runs in milliseconds, and each run. mode='coldWarm' runs it cold (the cache is cleared before each run) and warm, and returns the average, standard deviation, minimum and maximum of total time and storage-engine time for each, plus each run. Cold runs and the storage-engine split need a local Power BI Desktop or an admin XMLA endpoint. Clearing the cache on a shared or cloud endpoint affects everyone using it, so there it needs confirm=true, and without it the result is warm only. clearForCold=false keeps a coldWarm run warm only. Use it before and after a rewrite to show the rewrite is faster. It does not change the model. runs: 1 to 25.

Entitlement: Free entry point. Paid advice and selected automatic actions still need Pro where stated. Release class: supported. Risk class: standard.

Permission: the current target policy applies. Exact approvals and human-only actions remain as stated in this operation’s description.

The engine ambiently snapshots the top measures' values + their whole DAX dependency-cone expressions at checkpoint moments (apply_plan / optimize_measure / deploy_live / save_model, automatic); this op finds the most recent movement of a measure (or since a given point) and attributes it honestly: verdict 'attributed' when exactly one recorded edit sits in the window; 'interval' when several do: candidates come ranked by dependency overlap (formula-changing edits first) and are a ranking, never a causal claim; 'data-suspected' when every formula is identical but the value moved (data/refresh is the suspect. Proving it needs a shadow pre-edit instance, which v1 does not keep); 'inconclusive' when the history can't answer. Formula-caused movement is proven by a deterministic expression diff of the dependency cone between the two points (ExprDiffs). Free. Read-only.

Entitlement: Free entry point. Paid advice and selected automatic actions still need Pro where stated. Release class: supported. Risk class: standard.

Permission: the current target policy applies. Exact approvals and human-only actions remain as stated in this operation’s description.

The Formula Engine builds the logical plan, the Storage Engine the physical plan (with #Records per operator). Use it alongside profile_dax when a query is fe-bound (a tall physical plan / huge intermediate records points at the costly operators). Needs a local Power BI Desktop or an admin XMLA endpoint; degrades to 'unavailable' otherwise. Read-only.

Entitlement: Free entry point. Paid advice and selected automatic actions still need Pro where stated. Release class: supported. Risk class: standard.

Permission: the current target policy applies. Exact approvals and human-only actions remain as stated in this operation’s description.

Non-destructive: it only evicts the cache (the engine rebuilds it on the next query); no data or metadata changes. safety: on a shared/cloud endpoint this evicts the cache for all users (their next queries run cold), so it is refused there unless you pass confirm=true. Needs a local Power BI Desktop or an admin XMLA endpoint.

Entitlement: Free entry point. Paid advice and selected automatic actions still need Pro where stated. Release class: supported. Risk class: standard.

Permission: the current target policy applies. Exact approvals and human-only actions remain as stated in this operation’s description.

Wrap the sub-expressions you want to inspect in evaluateandlog to see what they actually evaluate to. Needs a local Power BI Desktop or admin XMLA. Read-only. The logged intermediates can include row samples of any table, so the agent policy's QueryData rule applies to every call on the connected target (regardless of the outer query shape). An Ask refusal names the approval to request; a Deny means the target's policy forbids it (a human can run it or change the policy).

Entitlement: Free entry point. Paid advice and selected automatic actions still need Pro where stated. Release class: supported. Risk class: standard.

Permission: the current target policy applies. Exact approvals and human-only actions remain as stated in this operation’s description.

Honesty rules: contributors ship only when the parts provably sum to the total (distinct counts / ratios / min-max / semi-additive measures get an explicit refusal, never a misleading breakdown); RLS is listed as roles that would filter these tables (the engine cannot impersonate a role, so never 'computed under role X'). Describe the cell with filterContext: Filters = column+members (a pivot cell's row/col keys are single-member filters), ExtraPredicates = raw DAX boolean lines (page-filter style). offline the dossier degrades to chain/lineage/RLS with Evidence.Available=false. Connect (open_local / open_live) for the value, blank checks, and breakdown. Next steps: probe_measure to test the measure across many contexts; blame_value to ask what changed a number over time. Free, read-only.

Entitlement: Free entry point. Paid advice and selected automatic actions still need Pro where stated. Release class: supported. Risk class: standard.

Permission: the current target policy applies. Exact approvals and human-only actions remain as stated in this operation’s description.

'measure:Sales/Total Sales').

Entitlement: Free entry point. Paid advice and selected automatic actions still need Pro where stated. Release class: supported. Risk class: standard.

Permission: the current target policy applies. Exact approvals and human-only actions remain as stated in this operation’s description.

With an open session the table-identity rules activate (bare-table filter). Token-based, so comments / strings / [bracketed names] never false-fire. Read-only, free, advisory per-expression. The model-scored layer is the AI-readiness BestPractice category (ai_readiness_scan). Run it on a rewrite before applying.

Entitlement: Free entry point. Paid advice and selected automatic actions still need Pro where stated. Release class: supported. Risk class: standard.

Permission: the current target policy applies. Exact approvals and human-only actions remain as stated in this operation’s description.

Rows are never kept. Both the Studio and ask_data write the same history, kept beside the model like the Plan and never committed with it; a model opened live keeps its own, and an unsaved model keeps none. Read-only. Ask your data is a Semanticus Pro feature.

Entitlement: Pro: Ask your data. Release class: supported. Risk class: standard.

Permission: the current target policy applies. Exact approvals and human-only actions remain as stated in this operation’s description.

Each has a ref like 'function:MyFunc'. Read its body with get_dax, edit with update_measure, rename with rename_object.

Entitlement: Free entry point. Paid advice and selected automatic actions still need Pro where stated. Release class: supported. Risk class: standard.

Permission: the current target policy applies. Exact approvals and human-only actions remain as stated in this operation’s description.

No client parameters.

Points with a null value are honest gaps: the formulas were snapshotted but the number wasn't observed there (offline capture, or the edit's impact cone never reached the measure). Truncated=true means retention pruned older points (last 200 checkpoints / 20MB) or unreadable lines were skipped. Free, read-only.

Entitlement: Free entry point. Paid advice and selected automatic actions still need Pro where stated. Release class: supported. Risk class: standard.

Permission: the current target policy applies. Exact approvals and human-only actions remain as stated in this operation’s description.

Give 2 or more candidate DAX rewrites; the engine validates each (a candidate that references the measure itself is rejected), checks each returns identical values to the current body across the summarizecolumns(verifyGroupBy) matrix you supply, benchmarks only the ones that matched (warm wall-clock over that grid, not a server fe/se trace), and, when apply is true (the default), applies the fastest that beats the current body beyond a measured noise band as one undoable edit, recording the proof and benchmark as evidence. apply=false is a dry run that returns the same evidence and changes nothing. Equivalence is only as strong as verifyGroupBy, which must be a JSON list of text column names: a grand-total-only match is downgraded to unverified and never applied, so pass representative group-by columns. Nothing is applied with fewer than 2 valid candidates, without a live connection (open_live or open_local), or when no candidate is proven; a candidate that matched only under degraded fidelity is reported, not applied. Free on every tier.

Entitlement: Free entry point. Paid advice and selected automatic actions still need Pro where stated. Release class: supported. Risk class: standard.

Permission: the current target policy applies. Exact approvals and human-only actions remain as stated in this operation’s description.

Pass the result's shape and the rules' own answer: 'fits' (the kinds the rules leave, in their order) and 'rules_pick'. Kinds are matrix, line, bar, area, table, card and scatter. Jev's chart pick is off in this version until its replay clears its bar, so today this returns the rules' pick with by 'rules' and a note saying why; it does the same whenever Jev is not asked or not sure. Returns kind, by (jev or rules), a fixed reason sentence written by Semanticus, the confidence when Jev answered, and a note. Sends the question's words and the shape only, never rows or values. Read-only; never changes the model. Ask your data is a Semanticus Pro feature.

Entitlement: Pro: Ask your data. Release class: supported. Risk class: standard.

Permission: the current target policy applies. Exact approvals and human-only actions remain as stated in this operation’s description.

Returns long-form rows the caller pivots into a row x column matrix. Use it to sanity-check a measure's values by category/date before trusting it. Read-only; requires a live connection. Returns rows of source grouping values, so the agent policy's QueryData rule applies to the connected target. An Ask refusal names the approval to request; a Deny means the target's policy forbids it (a human can run it or change the policy).

Entitlement: Free entry point. Paid advice and selected automatic actions still need Pro where stated. Release class: supported. Risk class: standard.

Permission: the current target policy applies. Exact approvals and human-only actions remain as stated in this operation’s description.

Accepts a table name or a 'table:Name' ref. Requires a live connection. Reads data rows, so the agent policy's QueryData rule applies to the connected target. An Ask refusal names the approval to request; a Deny means the target's policy forbids it (a human can run it or change the policy).

Entitlement: Free entry point. Paid advice and selected automatic actions still need Pro where stated. Release class: supported. Risk class: standard.

Permission: the current target policy applies. Exact approvals and human-only actions remain as stated in this operation’s description.

DAX context-dependence bugs (missing time-intel guards, blanks, non-additivity) are invisible until you run the measure in different contexts. It replicates a real visual (summarizecolumns with your filter args as the outer/slicer context + a group-by axis) so allselected / selectedvalue / isinscope resolve correctly. A naive one-context test lies about those. Pass a primaryAxis to auto-generate grand/single/multi/empty/boundary scenarios, and/or scenarios[] for cross-filter cases (name columns + members; do not hand-write DAX). Reports behavior across contexts, never 'correct'. You judge it against intent. Read-only; needs a live connection.

Entitlement: Free entry point. Paid advice and selected automatic actions still need Pro where stated. Release class: supported. Risk class: standard.

Permission: the current target policy applies. Exact approvals and human-only actions remain as stated in this operation’s description.

Use it to understand why a query is slow (fe-bound vs se-bound) when optimizing. Needs a local Power BI Desktop or an admin XMLA endpoint; falls back to wall-clock only otherwise. Read-only.

Entitlement: Free entry point. Paid advice and selected automatic actions still need Pro where stated. Release class: supported. Risk class: standard.

Permission: the current target policy applies. Exact approvals and human-only actions remain as stated in this operation’s description.

The engine runs the DAX side (summarizecolumns over your groupBy + the measure, plus an independent grand-total query) over the live XMLA connection and runs your supplied SQL over the Fabric SQL endpoint, full-outer-joins them by the composite grouping key, and judges every cell under an explicit tolerance + blank policy. The engine never authors the SQL. You supply the ground truth. The SQL runs against the source endpoint as your own identity (the engine holds no elevated credential. It can only read what you already can; the endpoint's nature + your warehouse permissions decide what is possible). Whether you (an agent) may run it is governed by the QueryData permission setting on the SQL target (the permissions tab, Ask/Deny/Allow), exactly like preview_table / run_dax. A human runs ungated. groupBy entries are resolved to real model columns (a 'column:Table/Column' ref or a plain 'Table'[Column]); an unresolved entry is refused (this prevents wrong results, not writes). blankPolicy is required ("zero"|"null"|"distinct"). Grade on status (Reconciled|Mismatch|InsufficientCoverage|NothingVerifiable|InputError), not anyMismatch; an uncaveated pass is Reconciled and unverifiable==0 and complete==true, and even then expected-member coverage is unknown (coverageKnown=false; it reconciles the members observed, not that none is missing). A total-only run is InsufficientCoverage by design (a matching total can hide the VertiPaq blank-row trap). Requires a live connection; reads source-row aggregates from both the model and the SQL endpoint. Returns the status, summary, worst offender, coverage facts, per-side execution timestamps (independent connections may see different snapshots), snapshotNote, and the mismatching cells (capped).

Entitlement: Free entry point. Paid advice and selected automatic actions still need Pro where stated. Release class: supported. Risk class: standard.

Permission: the current target policy applies. Exact approvals and human-only actions remain as stated in this operation’s description.

Returns {error} if not connected or the query fails. Read-only. A row-returning query (evaluate <table>, summarizecolumns, filter, values, topn, …) reads source rows, so the agent policy's QueryData rule applies to the connected target. An Ask refusal names the approval to request; a Deny means the target's policy forbids it (a human can run it or change the policy). A scalar probe (evaluate row(...) / evaluate { … }) is a calculation and is never gated.

Entitlement: Free entry point. Paid advice and selected automatic actions still need Pro where stated. Release class: supported. Risk class: standard.

Permission: the current target policy applies. Exact approvals and human-only actions remain as stated in this operation’s description.

Select * from $system.TMSCHEMA_MEASURES) against the connected live model. Read-only.

Entitlement: Free entry point. Paid advice and selected automatic actions still need Pro where stated. Release class: supported. Risk class: standard.

Permission: the current target policy applies. Exact approvals and human-only actions remain as stated in this operation’s description.

The expression is validated offline first: malformed syntax, arity problems and unknown table, column or measure references are refused before anything is written, whatever the Record verified edits setting. An unchanged expression is a no-op. The change is applied to the live shared session, appears immediately in the vs Code UI, and is a single undoable step.

Entitlement: Free entry point. Paid advice and selected automatic actions still need Pro where stated. Release class: supported. Risk class: high.

Permission: the current target policy applies. Exact approvals and human-only actions remain as stated in this operation’s description.

Catches unbalanced brackets (error) and unknown table/column/measure references (warning), with line:column. Offline/read-only, no live engine needed. By default valid=false only on a structural error. strict=true judges it the way a DAX write does: every warning, such as a missing column, is an error and valid is false.

Entitlement: Free entry point. Paid advice and selected automatic actions still need Pro where stated. Release class: supported. Risk class: standard.

Permission: the current target policy applies. Exact approvals and human-only actions remain as stated in this operation’s description.

The extra shapes matter: summarizecolumns leaves pin every dimension, so a wrong-scope rewrite (wrong denominator, collapsed context transition, over-broad all/allexcept) can match at every leaf yet diverge at a subtotal or the grand total. This gate catches that. Use it, over a representative group-by matrix, before replacing a measure with a hand-applied rewrite (update_measure); optimize_measure and apply_plan run the same proof themselves. Read-only.

Entitlement: Free entry point. Paid advice and selected automatic actions still need Pro where stated. Release class: supported. Risk class: standard.

Permission: the current target policy applies. Exact approvals and human-only actions remain as stated in this operation’s description.

Returns observed column and table bytes, component bytes, encoding, an observed-denominator percentage, and an explicit storageMode. Import mode establishes complete model totals, directLake is resident-only, and unknown leaves completeness unconfirmed. Requires a live connection.

Entitlement: Free entry point. Paid advice and selected automatic actions still need Pro where stated. Release class: supported. Risk class: standard.

Permission: the current target policy applies. Exact approvals and human-only actions remain as stated in this operation’s description.

Classifies fact/dimension from FK topology, maps SQL types to tom types, hides key columns, and proposes additive measures + a time-intelligence set. storageMode 'import' (M) or 'directLake'. authMode: azcli (default) | serviceprincipal | interactive | devicecode. If the endpoint rejects the sign-in with an authentication error, the workspace is usually in a different tenant than your az login. Pass tenantId. To reuse a SQL endpoint the person already saved on Connections (same sign-in, no password stored), read it with list_sql_sources and pass its server, database, authMode and tenantId. Which links are drafted inactive, each with a plain note on the relationship saying why (a measure can still use it with userelationship): a fact table's own links always stay on (to the date table and to each dimension it references); when two of a fact's links reach the same table (an invoice date and a due date), the one listed first among the fact's columns stays on and the rest are off; any other link (a snowflake hop, a dimension's own date key) that would give a table a second path is off; a table linked to itself is off and not built. A source date table (a non-fact table named like date or calendar with a date column) is set as dateTable and marked as the date table on build, keyed on its date column; only a source with none gets a calendar table. The result is a first draft. Refine it, then create_model + build_model_from_spec. (Fabric often declares no enforced keys, so relationships may be empty. Add them in the spec.) Returns the spec view.

Entitlement: Free entry point. Paid advice and selected automatic actions still need Pro where stated. Release class: supported. Risk class: critical.

Permission: the current target policy applies. Exact approvals and human-only actions remain as stated in this operation’s description.

The autogen is a first draft. Refine it (set_spec / the Spec tab), then build_model_from_spec. Returns the spec view.

Entitlement: Free entry point. Paid advice and selected automatic actions still need Pro where stated. Release class: supported. Risk class: standard.

Permission: the current target policy applies. Exact approvals and human-only actions remain as stated in this operation’s description.

No client parameters.

Composes the authoring primitives in dependency order: data source + shared M -> tables (import/Direct Lake/calculated, with columns + data types + keys) -> relationships (FK->PK) -> date table -> measures -> time-intelligence. Existing objects are skipped (re-run safe); per-row failures are reported without aborting. Build into the current session. Call create_model first for a from-scratch build. Returns a report (created/skipped/errors + before->after counts).

Entitlement: Free entry point. Paid advice and selected automatic actions still need Pro where stated. Release class: supported. Risk class: standard.

Permission: the current target policy applies. Exact approvals and human-only actions remain as stated in this operation’s description.

No client parameters.

The spec never mutated it). Returns the empty spec view.

Entitlement: Free entry point. Paid advice and selected automatic actions still need Pro where stated. Release class: supported. Risk class: standard.

Permission: the current target policy applies. Exact approvals and human-only actions remain as stated in this operation’s description.

No client parameters.

Read-only. Use it to understand the whole model at once or to see what the built docs will contain.

Entitlement: Free entry point. Paid advice and selected automatic actions still need Pro where stated. Release class: supported. Risk class: standard.

Permission: the current target policy applies. Exact approvals and human-only actions remain as stated in this operation’s description.

Use this to discover where to add business context for the exported documentation. The narrative is additional context for the docs, separate from each object's first-class Description (which you set with set_description).

Entitlement: Free entry point. Paid advice and selected automatic actions still need Pro where stated. Release class: supported. Risk class: standard.

Permission: the current target policy applies. Exact approvals and human-only actions remain as stated in this operation’s description.

No client parameters.

ObjRef is 'model' for the model-wide narrative, or 'table:Name' / 'measure:Table/Name' (narrative is supported on the model, tables and measures). Common sectionKeys: 'overview', 'businessContext', 'notes' (per object); 'overview', 'glossary', 'methodology' (model). Returns null if empty.

Entitlement: Free entry point. Paid advice and selected automatic actions still need Pro where stated. Release class: supported. Risk class: standard.

Permission: the current target policy applies. Exact approvals and human-only actions remain as stated in this operation’s description.

The spec is an authoring artifact (not the model itself); build_model_from_spec materialises it. Returns { version, source, spec } (spec is null if none is loaded). Read-only.

Entitlement: Free entry point. Paid advice and selected automatic actions still need Pro where stated. Release class: supported. Risk class: standard.

Permission: the current target policy applies. Exact approvals and human-only actions remain as stated in this operation’s description.

No client parameters.

Broadcasts live to the Spec tab. Returns the loaded spec view.

Entitlement: Free entry point. Paid advice and selected automatic actions still need Pro where stated. Release class: supported. Risk class: standard.

Permission: the current target policy applies. Exact approvals and human-only actions remain as stated in this operation’s description.

Pass an absolute path, e.g. '.../model.spec.json'. Returns the spec view.

Entitlement: Free entry point. Paid advice and selected automatic actions still need Pro where stated. Release class: supported. Risk class: standard.

Permission: the current target policy applies. Exact approvals and human-only actions remain as stated in this operation’s description.

ObjRef is 'model' or 'table:Name' / 'measure:Table/Name' (narrative is supported on the model, tables and measures; other refs are rejected). Common sectionKeys: 'overview', 'businessContext', 'notes' (per object); 'overview', 'glossary', 'methodology' (model). Set markdown to empty/null to clear the section. This is the collaborate-on-the-docs path: it's stored as a model annotation, broadcasts live to the human's Documentation tab, and is undoable.

Entitlement: Free entry point. Paid advice and selected automatic actions still need Pro where stated. Release class: supported. Risk class: high.

Permission: the current target policy applies. Exact approvals and human-only actions remain as stated in this operation’s description.

Use this to author or refine the spec; it broadcasts live to the Spec tab. Shape: { name, compatibilityLevel, storageMode: 'import'|'directLake', source: { kind:'fabric-sql', server, database, schema }, tables: [{ name, role:'fact'|'dimension'|'date'|'calculated'|'isolated', entity, schema, mExpression, calculatedExpression, sourceName, columns: [{ name, dataType, sourceColumn, isKey, hidden, summarizeBy }] }], relationships: [{ fromTable, fromColumn, toTable, toColumn, cardinality:'manyToOne'(default)|'oneToOne'|'oneToMany'|'manyToMany', crossFilter, isActive, note }], measures: [{ table, name, dax (or expression), description, formatString, displayFolder }], timeIntelligence: ['ytd','py',...], timeIntelligenceBaseMeasures: [...], dateTable: { name, dateColumn, startExpr, endExpr, markAsDate } }. A relationship's note is plain words for the person on why a link is off; the build ignores it. When dateTable.name is one of the spec's own tables, the build marks that table as the date table keyed on dateColumn (or its one date column) and adds no calendar table; a table already marked is left as it is. Otherwise a calendar table named dateTable.name is added from startExpr and endExpr. Does not touch the model. Returns the new spec view.

Entitlement: Free entry point. Paid advice and selected automatic actions still need Pro where stated. Release class: supported. Risk class: high.

Permission: the current target policy applies. Exact approvals and human-only actions remain as stated in this operation’s description.

Both details carry the Best practice letter, A to F, as bestPractice with bestPracticeWhy (ai_readiness_scan's grade is the other letter, AI readiness). The rule: Best practice is A when the model has no open errors and no open warnings. Info notes never count against the letter. A finding you accepted counts as passed. An error or warning rule that could not be checked also stops an A. Below A, each open error counts 5 points and each open warning 2 points. The points are divided by the number of objects checked, so a big model is not marked down for its size. Tables, columns, measures, relationships, roles and the like are objects; partitions are not, because the table already stands for them. B is up to 0.10 points per object, C is over 0.10 up to 0.25, D is over 0.25 up to 0.50, and F is over 0.50. It uses the bundled Power BI rule set plus any custom rules saved on the model (load_rules and reset_rules with system='bpa'). detail='summary' returns counts only: rule count, violation count, auto-fixable count, rule errors, orphaned waivers, and counts by category, rule and severity. It is the lighter first call on a large model. detail='full' (the default) also returns the violations, each with rule, category, severity, object, message and whether it can be fixed automatically (CanAutoFix). The filters narrow only the returned violations, and the counts always cover the whole model. A filter given with detail='summary' is refused, because a summary has no list to narrow. For AI readiness use ai_readiness_scan. Read-only.

Entitlement: Free entry point. Paid advice and selected automatic actions still need Pro where stated. Release class: supported. Risk class: standard.

Permission: the current target policy applies. Exact approvals and human-only actions remain as stated in this operation’s description.

The audit trail behind the scorecard's waived count. This is the raw store, so it can include waivers whose rule id is no longer in the loaded rule set: those accept nothing, and bpa_scan and ai_readiness_scan report them as OrphanedWaivers with a count. Compare the two before you trust a waiver.

Entitlement: Free entry point. Paid advice and selected automatic actions still need Pro where stated. Release class: supported. Risk class: standard.

Permission: the current target policy applies. Exact approvals and human-only actions remain as stated in this operation’s description.

No client parameters.

System: 'bpa' or 'air'. objRef set = un-waive that instance; empty/'*' = remove the rule-level waiver. Never gated (it makes the model more compliant). Undoable; also clears Tabular Editor's ignore annotation for a per-instance BPA waiver.

Entitlement: Free entry point. Paid advice and selected automatic actions still need Pro where stated. Release class: supported. Risk class: standard.

Permission: the current target policy applies. Exact approvals and human-only actions remain as stated in this operation’s description.

'we keep these unused columns'). system: 'bpa' or 'air'. objRef set = waive this instance; objRef empty/'*' = waive the entire rule, every instance model-wide. a reason is required: a waiver is an audited decision (stored with who+when on the model), never a silent suppression. The finding is still surfaced (tagged 'waived' with its reason) and the scorecard reports a waived count, so the grade is honest. Hard gates (Copilot index ceiling, >50% measures undescribed) are not lifted by waivers. Persisted on the model (undoable, travels with it); per-instance BPA waivers also write Tabular Editor's BestPracticeAnalyzer_IgnoreRules so te3 honours them. Re-scan to see the score move.

Entitlement: Free entry point. Paid advice and selected automatic actions still need Pro where stated. Release class: supported. Risk class: standard.

Permission: the current target policy applies. Exact approvals and human-only actions remain as stated in this operation’s description.

Pure file authoring, no Fabric call. Returns the contents; write=true also writes them into the repo.

Entitlement: Free entry point. Paid advice and selected automatic actions still need Pro where stated. Release class: supported. Risk class: critical.

Permission: the current target policy applies. Exact approvals and human-only actions remain as stated in this operation’s description.

Dry run by default (enumerates parts + target, writes nothing); the live publish (commit=true) is human-only from the Deploy tab. An agent's commit is refused, previewing is allowed.

Entitlement: Free entry point. Paid advice and selected automatic actions still need Pro where stated. Release class: dry-run-only. Risk class: critical.

Permission: the current target policy applies. Exact approvals and human-only actions remain as stated in this operation’s description.

With no change set to scope to (no edits in this session yet, a model-wide edit, or no saved or published copy) it judges the whole model instead: wholeModel=true and checkLine opens with why. Read-only. waived BPA findings never block (they're audited acceptances, bpaWaivedBlocking counts them); readiness hard-gates still apply when this change touches them.

Entitlement: Free entry point. Paid advice and selected automatic actions still need Pro where stated. Release class: supported. Risk class: critical.

Permission: the current target policy applies. Exact approvals and human-only actions remain as stated in this operation’s description.

No client parameters.

Metadata only (names/descriptions/visibility/categories/format strings/folders/summarize-by, measure + calc-column DAX, partition sources (M / calc-table DAX / legacy), shared expressions, and the linguistic schema) via Model.SaveChanges(), matched by LineageTag so renames land on the right object. no data refresh. Absence never deletes: an object that is only on the live model is reported in LiveOnly and left alone unless you pass it in deleteRefs (the opt-in tick). A new calculated table is created (and Calculate-recalced on commit); other structural changes (new data/import tables, new data columns, new partitions, source-type or Direct Lake rebinds, incremental-refresh policy) are reported in Unmatched/LiveOnly, never silently dropped. Empty endpoint = deploy back to the model this session opened (the round-trip); or pass endpoint + database. dry run by default: commit=true writes, and that write needs the review token the preview returned. A committed write first saves a restore point of the live model's metadata (list_restore_points) so rollback_push can undo it; if that snapshot fails, a write with deleteRefs is refused and a write without deletes goes ahead with a warning. Needs a Read-Write XMLA endpoint and a write principal (authMode=serviceprincipal is the reliable one).

Entitlement: Free entry point. Paid advice and selected automatic actions still need Pro where stated. Release class: supervised. Risk class: critical.

Permission: the current target policy applies. Exact approvals and human-only actions remain as stated in this operation’s description.

Dry run by default (the New/Update preview + the readiness gate); commit=true deploys. An agent may deploy to non-production stages only. Promoting to production needs a human confirming from the Deploy tab with the confirmToken their own dry-run shows. A failing gate blocks unless forceOverride=true. items = source item ids (empty = all supported).

Entitlement: Free entry point. Paid advice and selected automatic actions still need Pro where stated. Release class: dry-run-only. Risk class: critical.

Permission: the current target policy applies. Exact approvals and human-only actions remain as stated in this operation’s description.

Read-only.

Entitlement: Free entry point. Paid advice and selected automatic actions still need Pro where stated. Release class: supported. Risk class: critical.

Permission: the current target policy applies. Exact approvals and human-only actions remain as stated in this operation’s description.

Dry run by default (commit=false reports the pending change count; writes nothing). Pass commit=true to commit. items = item objectIds for a selective commit (empty = all).

Entitlement: Free entry point. Paid advice and selected automatic actions still need Pro where stated. Release class: dry-run-only. Risk class: critical.

Permission: the current target policy applies. Exact approvals and human-only actions remain as stated in this operation’s description.

Dry run by default (commit=false). Pass commit=true to connect. Needs workspace Admin. After connecting, initialize + fabric_git_update/commit to sync.

Entitlement: Free entry point. Paid advice and selected automatic actions still need Pro where stated. Release class: dry-run-only. Risk class: critical.

Permission: the current target policy applies. Exact approvals and human-only actions remain as stated in this operation’s description.

Read-only.

Entitlement: Free entry point. Paid advice and selected automatic actions still need Pro where stated. Release class: supported. Risk class: critical.

Permission: the current target policy applies. Exact approvals and human-only actions remain as stated in this operation’s description.

Dry run by default (commit=false). Pass commit=true to disconnect. Needs workspace Admin.

Entitlement: Free entry point. Paid advice and selected automatic actions still need Pro where stated. Release class: dry-run-only. Risk class: critical.

Permission: the current target policy applies. Exact approvals and human-only actions remain as stated in this operation’s description.

Read-only.

Entitlement: Free entry point. Paid advice and selected automatic actions still need Pro where stated. Release class: supported. Risk class: critical.

Permission: the current target policy applies. Exact approvals and human-only actions remain as stated in this operation’s description.

Dry run by default (commit=false reports the incoming change count + any conflicts). Pass commit=true to apply. conflictPolicy = PreferRemote (default) | PreferWorkspace; allowOverride=true permits overwriting items that have workspace changes.

Entitlement: Free entry point. Paid advice and selected automatic actions still need Pro where stated. Release class: dry-run-only. Risk class: critical.

Permission: the current target policy applies. Exact approvals and human-only actions remain as stated in this operation’s description.

Needs an Admin role on the pipeline. Live read; read-only.

Entitlement: Free entry point. Paid advice and selected automatic actions still need Pro where stated. Release class: supported. Risk class: standard.

Permission: the current target policy applies. Exact approvals and human-only actions remain as stated in this operation’s description.

SemanticModel/Report, with source/target item ids + last deployment time). Use it to resolve the SemanticModel itemId to deploy. Needs Contributor on the stage workspace. Live read; read-only.

Entitlement: Free entry point. Paid advice and selected automatic actions still need Pro where stated. Release class: supported. Risk class: standard.

Permission: the current target policy applies. Exact approvals and human-only actions remain as stated in this operation’s description.

Use get_pipeline_stages then get_stage_items to drill into a pipeline's Dev/Test/Prod stages. Live read; read-only.

Entitlement: Free entry point. Paid advice and selected automatic actions still need Pro where stated. Release class: supported. Risk class: critical.

Permission: the current target policy applies. Exact approvals and human-only actions remain as stated in this operation’s description.

Every committed deploy_live write and every committed apply_model_diff push to an XMLA endpoint writes one first, so the push can be undone. This is the only way back from a live delete (a deploy re-creates measures, calculated columns, calculated tables and named expressions, but a relationship, role, perspective, hierarchy, partition, culture, datasource or data table removed from a published model is otherwise gone for good). Newest first. Only the newest 10 per target are kept. Filter by endpoint + database, or omit both for every target. Snapshots live in ~/.semanticus/restore and hold the target's full model metadata (measures, M queries, role filters, connection strings, never row data) in clear text; purge_restore_points removes them. An integrityError means rollback is refused, while purge remains available through its validated storage path. Read-only and free. Next: rollback_push(id) to preview a restore.

Entitlement: Free entry point. Paid advice and selected automatic actions still need Pro where stated. Release class: supported. Risk class: standard.

Permission: the current target policy applies. Exact approvals and human-only actions remain as stated in this operation’s description.

Live read against api.fabric.microsoft.com using your Entra identity. Read-only.

Entitlement: Free entry point. Paid advice and selected automatic actions still need Pro where stated. Release class: supported. Risk class: standard.

Permission: the current target policy applies. Exact approvals and human-only actions remain as stated in this operation’s description.

Read-only, deploys nothing. ('Update' items may be identical; Different-vs-NoDifference is only known after a real deploy.)

Entitlement: Free entry point. Paid advice and selected automatic actions still need Pro where stated. Release class: supported. Risk class: critical.

Permission: the current target policy applies. Exact approvals and human-only actions remain as stated in this operation’s description.

Choose exactly one selector: id removes one restore point, or olderThanDays selects every snapshot older than that non-negative age. dry run by default: confirm=false returns the exact candidates and a token without deleting. After reviewing them, pass confirm=true with that exact confirmToken; a stale or changed candidate set is refused. Reports partial file-delete failures honestly. Use this to clear model metadata from disk only when you accept that the guarded live push can no longer be undone.

Entitlement: Free entry point. Paid advice and selected automatic actions still need Pro where stated. Release class: supported. Risk class: high.

Permission: the current target policy applies. Exact approvals and human-only actions remain as stated in this operation’s description.

Types: Full (reload + recalc, the default), DataOnly (reload, no recalc), Calculate (recalc only), ClearValues (empty it, destructive), Automatic (refresh only if stale), Add (append rows). live data write against the Analysis Services engine: runs as a dry run by default (reports what would run, connects to nothing); pass commit=true to actually execute. Leave endpoint empty to target the live model this session was opened from (open_live/open_local binds it); a file model has nothing to refresh. See list_refresh_types for full explanations.

Entitlement: Free entry point. Paid advice and selected automatic actions still need Pro where stated. Release class: experimental. Risk class: critical.

Permission: the current target policy applies. Exact approvals and human-only actions remain as stated in this operation’s description.