STACK

Run read-only analytics queries against one of the user's estimates and their company's recent estimate history — the engine behind estimate reviews. Before running a review, read the stack://workflows/review-estimate workflow resource (discover it via search_resources): it is the recipe — which checks to run, the defaults, how to present findings — and it carries the full query grammar. Pass queriesJson as a JSON array of up to 15 named queries, each {"name": <label>, "kind": "analytic"|"checks", …} composed per that grammar. When queriesJson is omitted, runs the single default unit-cost-deviation query (lookbackWeeks/thresholdPct apply only then). Obtain the estimateId from find_estimates (call find_project first if you only have a project name); never invent one. All numbers are server-computed; a flagged result is a candidate for explanation, not a declared error.

query_estimate_analytics

Set, change, or clear tag values on takeoffs in one of your company's own libraries. It puts already-existing tags onto takeoffs; it does NOT create new tags (use create_tags_in_library for that). Use find_library_takeoffs first to get the library, its takeoffs, and the tag group/value ids. For each takeoff, give a tag value to set a group or null to clear it; each group holds one value, and groups you don't name are left alone. It writes to all the listed takeoffs at once (all-or-nothing), so never call it until the user has confirmed the library, the takeoffs, and the exact changes. Returns one row per takeoff with each group's result: assigned, changed, cleared, or unchanged.

assign_tags_to_library_takeoff

Copy selected takeoffs from one project into a library you own, so they become reusable library takeoffs (the reverse of import_library_takeoffs). Pick the takeoffs with find_takeoffs and the destination library with get_company_options (the libraries section). The copy is all-or-nothing and copying again creates duplicates — NEVER call this before the user has confirmed both the exact takeoffs and the destination library. Tags are not copied. Returns each created library takeoff (id, name) plus a count.

copy_takeoffs_to_library

Create a new STACK project. Required: project name and bid date (never invent the bid date — ask if not supplied). Optional: office ids, status id, currency (ISO 4217), unit system ("metric" or "imperial"), notes, project number, privacy flag, and address fields. Office and status are given by id — get them from get_company_options (offices / statuses). Returns the new project's id and name.

create_project

Create tags in one of your company's own libraries, so the tags are ready to use on the takeoffs stored there. This creates the tag VOCABULARY only — it does not put tags on any takeoff. Name each tag as a category and a value (for example "Trade" / "Electrical"). Pick the destination library with get_company_options (the libraries section). Tags that already exist are left as they are (reported back untouched), so only what is missing is created. This is a deliberate write to your company's shared vocabulary — NEVER call this before the user has confirmed both the destination library and the exact tags to create. Returns a per-item report: each tag mapped to a status (created | valueAdded | exists) plus a count of how many were newly created.

create_tags_in_library

Create one or more NEW takeoffs on one STACK project. Pass an ARRAY of heterogeneous specs — each entry has its own name, measurementType, and optional style and tags (a single-element array is the one-off case). measurementType is given BY NAME — one of Area, Linear, Count, LinearWithDrop, PitchedArea, PitchedLinear, SurfaceArea, Volume2D, Volume3D — and is REQUIRED and IMMUTABLE — it cannot be changed after creation, so confirm it with the user before creating. Inapplicable style fields are SKIPPED and REPORTED, never silently written. The batch is ALL-OR-NOTHING: one invalid spec rejects the entire create; a single call is capped at 250 takeoffs. TO MATCH AN EXISTING TAKEOFF'S STYLE (from find_takeoffs): use the .id of its lineWidth / symbolType / symbolTypeSize refs as this spec's lineWidth / symbolTypeId / symbolTypeSizeId — never the ref object itself, and note the field names differ (symbolType -> symbolTypeId, symbolTypeSize -> symbolTypeSizeId). NEVER call this before the user has confirmed the exact set — propose the full list with name + measurement type for each and wait for explicit confirmation first. Returns a created/skipped envelope: new takeoff ids/names/measurement types (plus tag groups assigned), and which style fields were skipped on which spec.

create_takeoff

Apply a bulk visual and tag edit to the takeoffs on one STACK project. TARGET IN EXACTLY ONE OF TWO WAYS — provide ids OR selector, never both and never neither: (A) ids — an explicit set of takeoff ids; or (B) selector — a predicate that edits ALL matching takeoffs server-side. SELECTOR MODE HAS A LARGER BLAST RADIUS: NEVER use it before the user confirms BOTH the exact predicate AND the affected count (run find_takeoffs with the same filters first). UNIFORM PER CALL: applies the SAME styleChanges and tagAssignments to EVERY targeted takeoff. To give different takeoffs different values, decompose into a separate call per distinct value-set. Inapplicable style fields are skipped and reported, never silently written. tagAssignments: set tagValueId to assign/change a group's value, or null to CLEAR it. NEVER call this before the user has confirmed the exact change — propose the edit with affected takeoffs and before/after, and wait for explicit confirmation first. Returns an applied/skipped/failed envelope (plus matchedCount in selector mode): which takeoffs were mutated, which style fields were skipped on which measurement types, and which ids or tag values failed.

edit_takeoffs

List the estimates on one of the user's projects so you can pick one to review — this is how you identify an estimate. If you only have a project name, call find_project first to resolve the project; then call find_estimates with that id to see the project's estimates; then pass a returned estimate id to get_estimate_worksheet or query_estimate_analytics to run the review. Each line gives the estimate id, name, last-modified date, and which one is the project default. Read-only — it lists, it does not change anything.

find_estimates

Search the current user's STACK projects whose bid date falls within the past year (a rolling 12-month window) or is upcoming (future-dated); projects with no bid date are excluded. Results are scoped to what the user can see; the window cannot be widened. To list every project the user can see ("show me my projects" / "list all projects"), use intent=name with an empty or omitted query. Each hit already carries whether the project is private (isPrivate), its status name (status), and its assignees by name (assignedUsers) — no follow-up call is needed just to read a project's privacy or status. Before choosing intent=name with a query, check whether the user's criterion is a configurable project field (division, region, type, office, status, bid-date window): call get_company_options with the projectFields section first, and use intent=metadata with the matching field id rather than a name search, which matches only the project name and will miss field criteria.

find_project

Return one of the user's estimates as its full rendered worksheet model — line items, per-cost-type costs and markups, waste, labels/tags, and estimate-level totals, as compact JSON — so you can run an ad-hoc or novel structural review directly; for a standard review, query_estimate_analytics runs the predefined checks instead. The review is read-only: propose, never mutate. Before reviewing, read the stack://workflows/review-estimate workflow resource (discover it via search_resources) — it is the recipe for how to read the model and present findings. If no worksheet snapshot exists yet, the tool returns an actionable snapshot-needed message. Obtain the estimateId from find_estimates (call find_project first if you only have a project name); never invent one.

get_estimate_worksheet