[ T-04 // TRACK ]
ONTOLOGY
The semantic layer: object types, action types, functions.
- [ FM-FDE-02 ]
Think Like an FDE II: The Ontology Is the Business
The ontology is not a database schema with nicer names. It is the business rendered as things and the verbs that change them: nouns become object types, verbs become action types, judgment becomes functions. This module teaches the interview method that produces that model: how to sit with ops people, extract the nouns and verbs they actually use, find the decision inside the conversation, and know when a noun is really a report or a verb is one nobody is allowed to run. Every modeling claim is anchored in the Avinyr labs: FM-ON-01 (rows into Work Order objects, identity is RID and API name), FM-ON-02 (the action renamed from mechanism to operator verb, submission criteria as authorization, the dry run as proof), and FM-ON-03 (functions as encoded judgment, with the platform enforcing the contract).
- [ FM-ON-01 ]
Rows Into Things: Object Types, One Link, and the Branch Wall
We turned two datasets into a working slice of the ontology: a Work Order object type (66 objects, indexed), an Asset object type (66 objects, indexed, fed by an aggregate pipeline we built for the job), and a many-to-one foreign-key link between them. Every number was verified against ground truth from FM-PB-01: the aggregate kept 66 rows because the input holds 66 distinct assets, and we prove it. Four real field repairs: API names are enrollment-wide unique and case-insensitive (hit twice), the 'first' aggregate function is flagged nondeterministic, the group picker creates a duplicate group if you press Enter, and ontology work lives on a branch that Object Explorer cannot see until a proposal merges. Ours cannot merge yet, blocked by unrelated stale demo resources, and we show the exact screen that tells you so.
- [ FM-ON-02 ]
Actions and Writeback: the Edits Gate, the Dry Run, and the Broken Demo Twin
We made the FM-ON-01 Work Order type writable and built its first real action: Reassign technician, a Modify action with one parameter mapped to the Technician property, submission criteria set to everyone in the org, saved as ri.actions.main.action-type.2c9ab28c-2937-4bcb-b376-11d6a9dc5c4a. Then we tried to actually run it and mapped every wall: object search returns zero results enrollment-wide (even for demo types on main), AIP Assist flatly refuses to execute actions and says so verbatim, and Object Explorer cannot see branch types at all. The one surface that works is the action editor's own Test run: a full dry run that prefilled WO-1001's real technician (S. Kowalski, matching the FM-PB-01 ground truth CSV), computed the edit to T. Nguyen, and returned a before/after diff plus a timed execution log (0.386s total) while writing nothing. Six field repairs, including a second demo Work Order type on main that is broken at the schema-mapping level and cannot index. Plain scope statement: this module demonstrates a proposed edit, not an applied one. Real writeback needs an app surface and landed in FM-WS-02.
- [ FM-ON-03 ]
Ontology Functions in TypeScript: the Branch Wall, the Long Fix, and the Slow First Run
We created the repo FM-ON-03 ontology functions from the TypeScript Functions V2 template, hit the branch wall in Resource Imports (the Object Types picker lists only main-ontology demo types; searching Work returns zero), and built against the main-ontology [Example] Aircraft demo type instead. The function topAircraftByCarrier filters by carrierName, orders by flightCount descending, and returns tailNum, model, carrierName, flightCount. The Palantir linter rejected a plain number field with the verbatim error '"number" is not a supported type.'; the fix was Long from @osdk/functions, passed straight through. Three live preview runs verified behavior: empty carrier returns List 1000, an exact-match miss on United Airlines returns List 0, and Hawaiian Airlines Inc. returns List 41. We committed, tagged 0.0.2 (our own automation double-fired the click; the duplicate tag failed cleanly), waited roughly 3 to 4 minutes for the Published tab, and verified the published 0.0.2 run returned the same 41 rows after a roughly 2.5 to 3 minute cold start. Seven field repairs plus one honest automation artifact.