@dbx-tools/model 0.9.118 → 0.9.120

package/src/model-catalog.ts

@@ -3,7 +3,7 @@
  *
  * Lists the workspace's `/serving-endpoints` and, when the caller supplies a
  * trusted credential identity, caches the result per host and identity with a
- * TTL via AppKit's `CacheManager`. Concurrent callers in the same scope share
+ * TTL via `@dbx-tools/cache`. Concurrent callers in the same scope share
  * one in-flight promise. Surfaces each endpoint as a stable
  * {@link ServingEndpointSummary} - including the Foundation Model API
  * `quality` / `speed` / `cost` profile when present, the classified
@@ -22,8 +22,8 @@
  */
 
 import { createHash } from "node:crypto";
-import { CacheManager } from "@databricks/appkit";
 import { appkit } from "@dbx-tools/appkit";
+import { deleteCachedValue, getOrSet } from "@dbx-tools/cache/cache";
 import { errorUtils, log } from "@dbx-tools/shared-core";
 import {
   model,
@@ -33,6 +33,7 @@ import {
   type ServingEndpointSummary,
 } from "@dbx-tools/shared-model";
 
+import { modelCatalogueCache } from "./_cache.ts";
 import { classifyEndpointClasses, normalizeEndpoints, rankEndpoints } from "./_ranking.ts";
 
 const { ModelClass } = model;
@@ -54,12 +55,12 @@ export const DEFAULT_FUZZY_THRESHOLD = 0.4;
 
 /** Cache key parts under which endpoint listings are stored. */
 const CACHE_KEY_NAMESPACE = "serving-endpoints";
+const catalogueCache = modelCatalogueCache;
 
 /** Options for {@link listServingEndpoints}. */
 export interface ListServingEndpointsOptions {
   /**
-   * Override the default cache TTL for this call, in milliseconds. Forwarded to
-   * `CacheManager` as seconds.
+   * Override the default cache TTL for this call, in milliseconds.
    */
   ttlMs?: number;
   /**
@@ -73,30 +74,23 @@ export interface ListServingEndpointsOptions {
   cacheIdentity?: string;
 }
 
-/** Hash an opaque credential identity before handing it to CacheManager. */
+/** Hash an opaque credential identity before handing it to the shared cache. */
 function cacheUserKey(identity: string): string {
   return `model-${createHash("sha256").update(identity).digest("hex")}`;
 }
 
 /**
  * List Model Serving endpoints for the workspace owning `client`, routed
- * through AppKit's `CacheManager`. The manager gives us everything
- * `cachetools.TTLCache` provides plus what `cachetools-async` adds on top:
- * per-entry TTL, in-flight request coalescing (concurrent callers share one
- * fetch via the manager's internal `inFlightRequests` map), bounded size,
- * telemetry spans (`cache.getOrExecute`), and optional Lakebase persistence so
- * the catalogue survives restarts when the lakebase plugin is wired up.
+ * through the shared memory-first cache, including per-entry TTL, in-flight
+ * request coalescing, bounded process memory, and portable persistence.
  *
  * Returns plain {@link ServingEndpointSummary} objects (a stable subset of the
- * SDK type) so cache hits never expose stale SDK internals. Errors from
- * `CacheManager` or the SDK fetch propagate to the caller - we don't swallow
- * them so users see the real auth / network issue.
+ * SDK type) so cache hits never expose stale SDK internals.
  *
  * @param host - Workspace host used as the cache key. Pass the value resolved
  *   from `client.config.getHost()` so multi-host apps share one entry per
  *   workspace.
- * @param options.ttlMs - Override the default TTL just for this call. Forwarded
- *   to `CacheManager` as seconds.
+ * @param options.ttlMs - Override the default TTL just for this call.
  */
 export async function listServingEndpoints(
   client: WorkspaceClientLike,
@@ -104,13 +98,17 @@ export async function listServingEndpoints(
   options: ListServingEndpointsOptions = {},
 ): Promise<ServingEndpointSummary[]> {
   if (!options.cacheIdentity) return fetchEndpoints(client);
-  const ttlSec = Math.max(1, Math.round((options.ttlMs ?? DEFAULT_MODEL_CACHE_TTL_MS) / 1000));
-  return CacheManager.getInstanceSync().getOrExecute(
-    [CACHE_KEY_NAMESPACE, host],
+  return (await getOrSet(
+    catalogueCache,
+    catalogueCacheKey(host, options.cacheIdentity),
     () => fetchEndpoints(client),
-    cacheUserKey(options.cacheIdentity),
-    { ttl: ttlSec },
-  );
+    {
+      ttl: options.ttlMs ?? DEFAULT_MODEL_CACHE_TTL_MS,
+      returnCachedOnError: true,
+      cachedOnErrorTtl: "7d",
+      throwErrors: true,
+    },
+  ))!;
 }
 
 /**
@@ -119,7 +117,7 @@ export async function listServingEndpoints(
  * none of the cache-load enrichment ({@link listServingEndpoints} adds the
  * {@link ModelClass} stamp and the embedding-dimension probe). Use this for a
  * one-shot, dependency-light listing - e.g. a CLI that only needs names/tasks
- * for fuzzy resolution and doesn't want AppKit's `CacheManager` or the
+ * for fuzzy resolution and doesn't want shared cache persistence or the
  * per-embedding ping cost. Prefer {@link listServingEndpoints} for the cached,
  * enriched view.
  */
@@ -191,30 +189,30 @@ async function pingEmbeddingDimension(
     logger.warn("embedding ping returned no vector", { name });
     return undefined;
   } catch (err) {
-    logger.warn("embedding ping failed", { name, error: errorUtils.errorMessage(err) });
+    logger.warn("embedding ping failed", { name, error: errorUtils.toError(err) });
     return undefined;
   }
 }
 
 /**
- * Force-evict cached endpoint listings via AppKit's `CacheManager`. A host and
- * identity delete only that principal's workspace entry. Because CacheManager
- * cannot enumerate a namespace, omitting either value clears the complete
- * manager and should be reserved for tests or administrative refreshes.
+ * Force-evict cached endpoint listings. A host and identity delete only that
+ * principal's workspace entry. Omitting either value clears this catalogue.
  */
 export async function clearServingEndpointsCache(
   host?: string,
   cacheIdentity?: string,
 ): Promise<void> {
-  const cache = CacheManager.getInstanceSync();
   if (host && cacheIdentity) {
-    const key = cache.generateKey([CACHE_KEY_NAMESPACE, host], cacheUserKey(cacheIdentity));
-    await cache.delete(key);
+    await deleteCachedValue(catalogueCache, catalogueCacheKey(host, cacheIdentity));
   } else {
-    await cache.clear();
+    await catalogueCache.clear();
   }
 }
 
+function catalogueCacheKey(host: string, cacheIdentity: string): string {
+  return JSON.stringify([CACHE_KEY_NAMESPACE, host, cacheUserKey(cacheIdentity)]);
+}
+
 /**
  * Result of fuzzy-resolving a user-supplied model name against the live
  * endpoint list. `score` is the ranker's distance (`0` is exact, `1` is no match);

package/src/model-client.ts

@@ -15,6 +15,7 @@ import {
 } from "@dbx-tools/auth/client";
 import { listProfiles as listAuthProfiles } from "@dbx-tools/auth/profile";
 import type { DatabricksProfileSummary } from "@dbx-tools/shared-auth";
+import * as errorUtils from "@dbx-tools/shared-core/error-utils";
 import * as log from "@dbx-tools/shared-core/log";
 import * as object from "@dbx-tools/shared-core/object";
 import {
@@ -274,7 +275,7 @@ class DefaultModelClient implements ModelClient {
         } catch (error) {
           logger.warn("embedding dimension probe failed", {
             model: endpoint.name,
-            error,
+            error: errorUtils.toError(error),
           });
           return endpoint;
         }

package/lib/package.json

@@ -1,137 +0,0 @@
-{
-    "name": "@dbx-tools/model",
-    "description": "Workspace-aware Databricks Model Serving selection",
-    "repository": {
-        "type": "git",
-        "url": "git+https://github.com/reggie-db/dbx-tools.git",
-        "directory": "packages/js/node/model"
-    },
-    "scripts": {
-        "build": "projen build",
-        "compile": "projen compile",
-        "default": "projen default",
-        "metadata": "projen metadata",
-        "package": "projen package",
-        "post-compile": "projen post-compile",
-        "pre-compile": "projen pre-compile",
-        "prepack": "projen prepack",
-        "test": "projen test",
-        "watch": "projen watch",
-        "projen": "projen"
-    },
-    "devDependencies": {
-        "@types/cacache": "^20.0.1",
-        "@types/node": "catalog:",
-        "typescript": "^5.9.3"
-    },
-    "dependencies": {
-        "@databricks/appkit": "catalog:",
-        "@dbx-tools/appkit": "workspace:^",
-        "@dbx-tools/auth": "workspace:^",
-        "@dbx-tools/shared-auth": "workspace:^",
-        "@dbx-tools/shared-core": "workspace:^",
-        "@dbx-tools/shared-model": "workspace:^",
-        "cacache": "catalog:",
-        "cheerio": "^1.2.0",
-        "env-paths": "catalog:",
-        "fuse.js": "^7.4.2"
-    },
-    "main": "index.ts",
-    "license": "Apache-2.0",
-    "publishConfig": {
-        "access": "public",
-        "main": "./lib/index.js",
-        "types": "./lib/index.d.ts",
-        "exports": {
-            ".": {
-                "types": "./lib/index.d.ts",
-                "default": "./lib/index.js"
-            },
-            "./bindings": {
-                "types": "./lib/src/bindings.d.ts",
-                "default": "./lib/src/bindings.js"
-            },
-            "./classes": {
-                "types": "./lib/src/classes.d.ts",
-                "default": "./lib/src/classes.js"
-            },
-            "./classify": {
-                "types": "./lib/src/classify.d.ts",
-                "default": "./lib/src/classify.js"
-            },
-            "./fallback": {
-                "types": "./lib/src/fallback.d.ts",
-                "default": "./lib/src/fallback.js"
-            },
-            "./invoke": {
-                "types": "./lib/src/invoke.d.ts",
-                "default": "./lib/src/invoke.js"
-            },
-            "./metadata-cache": {
-                "types": "./lib/src/metadata-cache.d.ts",
-                "default": "./lib/src/metadata-cache.js"
-            },
-            "./metadata": {
-                "types": "./lib/src/metadata.d.ts",
-                "default": "./lib/src/metadata.js"
-            },
-            "./model-catalog": {
-                "types": "./lib/src/model-catalog.d.ts",
-                "default": "./lib/src/model-catalog.js"
-            },
-            "./model-client": {
-                "types": "./lib/src/model-client.d.ts",
-                "default": "./lib/src/model-client.js"
-            },
-            "./policy": {
-                "types": "./lib/src/policy.d.ts",
-                "default": "./lib/src/policy.js"
-            },
-            "./reasoning-translation": {
-                "types": "./lib/src/reasoning-translation.d.ts",
-                "default": "./lib/src/reasoning-translation.js"
-            },
-            "./resolve": {
-                "types": "./lib/src/resolve.d.ts",
-                "default": "./lib/src/resolve.js"
-            },
-            "./*": {
-                "types": "./lib/src/*/index.d.ts",
-                "default": "./lib/src/*/index.js"
-            },
-            "./package.json": "./package.json"
-        }
-    },
-    "version": "0.9.118",
-    "types": "index.ts",
-    "dbxToolsConfig": {
-        "tags": [
-            "node"
-        ]
-    },
-    "type": "module",
-    "exports": {
-        ".": "./index.ts",
-        "./bindings": "./src/bindings.ts",
-        "./classes": "./src/classes.ts",
-        "./classify": "./src/classify.ts",
-        "./fallback": "./src/fallback.ts",
-        "./invoke": "./src/invoke.ts",
-        "./metadata-cache": "./src/metadata-cache.ts",
-        "./metadata": "./src/metadata.ts",
-        "./model-catalog": "./src/model-catalog.ts",
-        "./model-client": "./src/model-client.ts",
-        "./policy": "./src/policy.ts",
-        "./reasoning-translation": "./src/reasoning-translation.ts",
-        "./resolve": "./src/resolve.ts",
-        "./*": "./src/*/index.ts",
-        "./package.json": "./package.json"
-    },
-    "files": [
-        "index.ts",
-        "src",
-        "exports.ts",
-        "lib"
-    ],
-    "//": "~~ Generated by projen. To modify, edit .projenrc.js and run \"bunx projen\"."
-}