@dbx-tools/model 0.3.28 → 0.3.30

package/README.md

@@ -172,6 +172,15 @@ rather than the SDK's typed `servingEndpoints.query`. Mint `authHeaders()` per
 request: the SDK refreshes the underlying token as it nears expiry, so you never
 track lifetimes yourself.
 
+Every serving path this repo talks to is a constant here, with a matching URL
+builder: `INVOCATIONS_SUFFIX` / `invocationsUrl()`, `RESPONSES_PATH` /
+`responsesUrl()`, `OPEN_RESPONSES_PATH` / `openResponsesUrl()`, and
+`CHAT_COMPLETIONS_PATH` / `chatCompletionsUrl()`. Import one instead of writing a
+`/serving-endpoints/...` literal in a consumer: which path a model can accept is
+a property of the model (see `isResponsesOnly()` and `responsesUpstreamUrl()`),
+so a hard-coded string in one package silently diverges when that routing
+changes.
+
 ## Use Static Fallbacks
 
 ```ts
@@ -192,8 +201,8 @@ policy decisions; fallbacks are a last resort.
   resolver functions.
 - `serving` - Databricks serving-endpoint listing, cache management, fuzzy
   search, and endpoint-id resolution.
-- `invoke` - invocations URL construction and per-request auth headers for
-  calling an endpoint over raw HTTP.
+- `invoke` - serving-path constants, URL construction, and per-request auth
+  headers for calling an endpoint over raw HTTP.
 - `classes` - model-class parsing, ordering, and class-ceiling helpers.
 - `fallback` - static fallback model ids per class.
 

package/package.json

@@ -13,16 +13,16 @@
   "dependencies": {
     "@databricks/appkit": "^0.43.0",
     "fuse.js": "^7.4.2",
-    "@dbx-tools/appkit": "0.3.28",
-    "@dbx-tools/shared-model": "0.3.28",
-    "@dbx-tools/shared-core": "0.3.28"
+    "@dbx-tools/appkit": "0.3.30",
+    "@dbx-tools/shared-core": "0.3.30",
+    "@dbx-tools/shared-model": "0.3.30"
   },
   "main": "index.ts",
   "license": "UNLICENSED",
   "publishConfig": {
     "access": "public"
   },
-  "version": "0.3.28",
+  "version": "0.3.30",
   "types": "index.ts",
   "type": "module",
   "exports": {

package/src/invoke.ts

@@ -38,6 +38,14 @@ export const RESPONSES_PATH = "serving-endpoints/responses";
  */
 export const OPEN_RESPONSES_PATH = "serving-endpoints/open-responses";
 
+/**
+ * Workspace-level OpenAI Chat Completions path. Body carries `model` (endpoint
+ * id), unlike {@link invocationsUrl} which names the endpoint in the URL. Use
+ * this when the caller has a model id rather than a per-endpoint route - e.g.
+ * attaching a provider tool spec to whichever endpoint was resolved.
+ */
+export const CHAT_COMPLETIONS_PATH = "serving-endpoints/chat/completions";
+
 /**
  * The OpenAI-compatible chat-completions invocations URL for an endpoint id.
  *
@@ -62,6 +70,11 @@ export function openResponsesUrl(host: string): string {
   return new URL(OPEN_RESPONSES_PATH, host).toString();
 }
 
+/** Workspace OpenAI Chat Completions URL (`POST`, model in the body). */
+export function chatCompletionsUrl(host: string): string {
+  return new URL(CHAT_COMPLETIONS_PATH, host).toString();
+}
+
 /**
  * True when the endpoint is known to reject Chat Completions `/invocations`
  * and require the Responses API (Codex models today). Callers should route