feat(appkit): migrate agent mlflow tracing to @mlflow/core with Unity Catalog support

knip.json

@@ -8,7 +8,7 @@
   ],
   "workspaces": {
     "packages/appkit": {
-      "ignoreDependencies": ["vitest"]
+      "ignoreDependencies": ["vitest", "@databricks/sdk-experimental"]
     },
     "packages/appkit-ui": {
       "ignoreDependencies": ["tailwindcss", "tw-animate-css"]

packages/appkit/package.json

@@ -71,6 +71,8 @@
   "dependencies": {
     "@ast-grep/napi": "0.37.0",
     "@databricks/lakebase": "workspace:*",
+    "@databricks/sdk-experimental": "0.17.0",
+    "@mlflow/core": "0.4.0",
     "@opentelemetry/api": "1.9.0",
     "@opentelemetry/api-logs": "0.219.0",
     "@opentelemetry/auto-instrumentations-node": "0.77.0",
@@ -83,8 +85,8 @@
     "@opentelemetry/resources": "2.8.0",
     "@opentelemetry/sdk-logs": "0.219.0",
     "@opentelemetry/sdk-metrics": "2.8.0",
-    "@opentelemetry/sdk-node": "0.219.0",
     "@opentelemetry/sdk-trace-base": "2.8.0",
+    "@opentelemetry/sdk-trace-node": "2.8.0",
     "@opentelemetry/semantic-conventions": "1.38.0",
     "@types/semver": "7.7.1",
     "apache-arrow": "21.1.0",
@@ -95,7 +97,6 @@
     "jiti": "2.6.1",
     "js-yaml": "4.3.1",
     "magic-string": "0.30.21",
-    "mlflow-tracing": "0.1.3",
     "obug": "2.1.1",
     "pg": "8.18.0",
     "picocolors": "1.1.1",

packages/appkit/src/core/appkit.ts

@@ -223,6 +223,12 @@ export class AppKit<TPlugins extends InputPluginMap> {
     const instance = new AppKit(mergedConfig);
 
     await Promise.all(instance.#setupPromises);
+
+    // Build the global tracer provider now that every plugin's setup() has run
+    // and contributed any span processors. Deferred to here so a single provider
+    // carries all processors (OTLP + plugin-contributed); see TelemetryManager.
+    TelemetryManager.start();
+
     await instance.#context.emitLifecycle("setup:complete");
 
     const handle = instance as unknown as PluginMap<T>;

packages/appkit/src/core/tests/appkit-as-user-exports.test.ts

@@ -45,6 +45,8 @@ vi.mock("../../telemetry", async () => {
     ...actual,
     TelemetryManager: {
       initialize: vi.fn(),
+      start: vi.fn(),
+      registerSpanProcessor: vi.fn(),
       getProvider: () => ({
         getTracer: () => ({
           startActiveSpan: vi.fn((_name: string, fn: (span: any) => any) =>

packages/appkit/src/plugins/agents/agents.ts

@@ -64,6 +64,7 @@ import {
   currentTraceId,
   initAgentTracing,
   linkTraceToRun,
+  startAgentTracing,
   traceAgent,
 } from "./mlflow";
 import { composePromptForAgent } from "./prompt";
@@ -201,6 +202,12 @@ export class AgentsPlugin extends Plugin implements ToolProvider {
 
   async setup() {
     await initAgentTracing();
+    // Seed mlflow's config right after TelemetryManager.start() (before the
+    // server serves), so the first turn's request-root span is forwarded and
+    // that turn assembles into a trace — not dropped as a cold-start artifact.
+    this.context?.onLifecycle("setup:complete", () => {
+      startAgentTracing();
+    });
     const { agents, defaultAgentName } = await this.buildAgentRegistry();
     this.agents = agents;
     this.defaultAgentName = defaultAgentName;