@alexkroman1/aai 0.3.0 → 0.5.0

package/README.md

@@ -0,0 +1,34 @@
+# aai
+
+Voice agent development kit. Define agents in TypeScript, deploy anywhere.
+
+## Install
+
+```sh
+npm install -g @alexkroman1/aai
+```
+
+## Quick start
+
+```sh
+# Create a new agent
+aai init my-agent
+cd my-agent
+
+# Start local dev server
+aai dev
+
+# Deploy to production
+aai deploy
+```
+
+## Commands
+
+| Command | Description |
+|---------|-------------|
+| `aai init [dir]` | Scaffold a new agent project |
+| `aai dev` | Start a local development server |
+| `aai deploy` | Bundle and deploy to production |
+| `aai start` | Start production server from build |
+| `aai secret <cmd>` | Manage secrets |
+| `aai rag <url>` | Ingest a site into the vector store |

package/dist/aai.js

@@ -0,0 +1,3 @@
+#!/usr/bin/env node
+if(!process.env.FORCE_COLOR&&!process.env.NO_COLOR&&process.stdout.isTTY)process.env.FORCE_COLOR='1';
+await import("./cli.js");

package/dist/cli.js

@@ -966,7 +966,8 @@ function createS2sSession(opts) {
       }
       handle.updateSession({
         system_prompt: systemPrompt,
-        tools: s2sTools
+        tools: s2sTools,
+        ...agentConfig.greeting ? { greeting: agentConfig.greeting } : {}
       });
       handle.addEventListener("ready", ((e) => {
         s2sSessionId = e.detail.session_id;
@@ -1087,13 +1088,6 @@ function createS2sSession(opts) {
     onAudioReady() {
       if (audioReady) return;
       audioReady = true;
-      if (agentConfig.greeting && s2s) {
-        s2s.updateSession({
-          system_prompt: systemPrompt,
-          tools: s2sTools,
-          greeting: agentConfig.greeting
-        });
-      }
     },
     onCancel() {
       client.event({ type: "cancelled" });
@@ -1344,7 +1338,7 @@ function wireSessionSocket(ws, opts) {
   const sid = sessionId.slice(0, 8);
   const ctx = opts.logContext ?? {};
   let session = null;
-  ws.addEventListener("open", () => {
+  function onOpen() {
     opts.onOpen?.();
     log.info("Session connected", { ...ctx, sid });
     const client = createClientSink(ws);
@@ -1353,13 +1347,18 @@ function wireSessionSocket(ws, opts) {
     ws.send(JSON.stringify({ type: "config", ...opts.readyConfig }));
     void session.start();
     log.info("Session ready", { ...ctx, sid });
-  });
+  }
+  if (ws.readyState === 1) {
+    onOpen();
+  } else {
+    ws.addEventListener("open", onOpen);
+  }
   ws.addEventListener("message", (event) => {
     if (!session) return;
     const msgEvent = event;
     const { data } = msgEvent;
-    if (data instanceof ArrayBuffer) {
-      const chunk = new Uint8Array(data);
+    if (Buffer.isBuffer(data)) {
+      const chunk = new Uint8Array(data.buffer, data.byteOffset, data.byteLength);
       if (!isValidAudioChunk(chunk)) {
         log.warn("Invalid audio chunk, dropping", {
           ...ctx,
@@ -1794,6 +1793,7 @@ async function askSelect(message, choices) {
           Select,
           {
             options: choices,
+            visibleOptionCount: choices.length,
             onChange: (value) => {
               resolve(value);
               app.unmount();
@@ -1807,7 +1807,8 @@ async function askSelect(message, choices) {
 
 // cli/_discover.ts
 function isDevMode() {
-  return path.basename(process.argv[0] ?? "") !== "aai";
+  const script = process.argv[1] ?? "";
+  return script.endsWith(".ts") || script.endsWith(".tsx");
 }
 function generateSlug() {
   return humanId({ separator: "-", capitalize: false });
@@ -2072,7 +2073,10 @@ function CommandRunner({
   const { items, log } = useStepLog();
   const [spinning, setSpinning] = useState(true);
   const [err, setErr] = useState(null);
+  const started = useRef(false);
   React.useEffect(() => {
+    if (started.current) return;
+    started.current = true;
     (async () => {
       try {
         await run(log);
@@ -2084,7 +2088,7 @@ function CommandRunner({
       setSpinning(false);
       exit();
     })();
-  }, [exit, log, onError, run]);
+  });
   return /* @__PURE__ */ jsxs2(Fragment, { children: [
     /* @__PURE__ */ jsx2(StepLog, { items }),
     err && /* @__PURE__ */ jsx2(ErrorLine, { msg: err }),
@@ -2485,6 +2489,7 @@ var DEFAULT_STYLES_CSS = `@import url("https://fonts.googleapis.com/css2?family=
 @import "tailwindcss";
 @source "./";
 @source "./components/";
+@source "./node_modules/@alexkroman1/aai/ui/";
 
 @theme {
   --color-aai-bg: #101010;
@@ -2562,7 +2567,7 @@ var INDEX_HTML = `<!DOCTYPE html>
       content="width=device-width, initial-scale=1.0, viewport-fit=cover"
     />
     <title>aai</title>
-    <link rel="icon" href="/favicon.svg" type="image/svg+xml" />
+    <link rel="icon" href="data:," />
     <link rel="stylesheet" href="../styles.css" />
   </head>
   <body>
@@ -2863,15 +2868,25 @@ init_output();
 
 // cli/_server_common.ts
 import path7 from "node:path";
+import { createServer as createViteServer } from "vite";
 init_output();
 async function loadAgentDef(cwd) {
   const agentPath = path7.resolve(cwd, "agent.ts");
-  const agentModule = await import(agentPath);
-  const agentDef = agentModule.default;
-  if (!agentDef || typeof agentDef !== "object" || !agentDef.name) {
-    throw new Error("agent.ts must export a default agent definition (from defineAgent())");
+  const vite = await createViteServer({
+    root: cwd,
+    logLevel: "silent",
+    server: { middlewareMode: true }
+  });
+  try {
+    const agentModule = await vite.ssrLoadModule(agentPath);
+    const agentDef = agentModule.default;
+    if (!agentDef || typeof agentDef !== "object" || !agentDef.name) {
+      throw new Error("agent.ts must export a default agent definition (from defineAgent())");
+    }
+    return agentDef;
+  } finally {
+    await vite.close();
   }
-  return agentDef;
 }
 async function resolveServerEnv() {
   const env = { ...process.env };
@@ -2885,13 +2900,21 @@ async function resolveServerEnv() {
   }
   return env;
 }
-async function bootServer(agentDef, html, env, port) {
+async function loadWsFromProject(cwd) {
+  const wsPath = path7.resolve(cwd, "node_modules", "ws", "index.js");
+  const mod = await import(wsPath);
+  const WS = mod.default ?? mod;
+  return (url, opts) => new WS(url, { headers: opts.headers });
+}
+async function bootServer(agentDef, html, env, port, cwd) {
   step("Start", `http://localhost:${port}`);
+  const createWebSocket = await loadWsFromProject(cwd);
   const { createServer: createServer2 } = await Promise.resolve().then(() => (init_server(), server_exports));
   const server = createServer2({
     agent: agentDef,
     clientHtml: html,
-    env
+    env,
+    createWebSocket
   });
   await server.listen(port);
   step("Ready", `http://localhost:${port}`);
@@ -2901,7 +2924,7 @@ async function bootServer(agentDef, html, env, port) {
 async function _startDevServer(cwd, port) {
   const agent = await loadAgent(cwd);
   if (!agent) {
-    throw new Error("No agent found \u2014 run `aai new` first");
+    throw new Error("No agent found \u2014 run `aai init` first");
   }
   step("Bundle", agent.slug);
   let html;
@@ -2918,7 +2941,7 @@ async function _startDevServer(cwd, port) {
   step("Load", "agent.ts");
   const agentDef = await loadAgentDef(cwd);
   const env = await resolveServerEnv();
-  await bootServer(agentDef, html, env, port);
+  await bootServer(agentDef, html, env, port, cwd);
 }
 
 // cli/dev.tsx
@@ -3332,7 +3355,7 @@ async function _startProductionServer(cwd, port) {
   step("Load", "agent");
   const agentDef = await loadAgentDef(cwd);
   const env = await resolveServerEnv();
-  await bootServer(agentDef, html, env, port);
+  await bootServer(agentDef, html, env, port, cwd);
 }
 
 // cli/start.tsx
@@ -3422,14 +3445,11 @@ async function main(args) {
       process.exit(1);
   }
 }
-var isMain = process.argv[1] === fileURLToPath3(import.meta.url);
-if (isMain) {
-  try {
-    await main(process.argv.slice(2));
-  } catch (err) {
-    error2(err instanceof Error ? err.message : String(err));
-    process.exit(1);
-  }
+try {
+  await main(process.argv.slice(2));
+} catch (err) {
+  error2(err instanceof Error ? err.message : String(err));
+  process.exit(1);
 }
 export {
   main

package/package.json

@@ -1,9 +1,9 @@
 {
   "name": "@alexkroman1/aai",
-  "version": "0.3.0",
+  "version": "0.5.0",
   "type": "module",
   "bin": {
-    "aai": "dist/cli.js"
+    "aai": "dist/aai.js"
   },
   "exports": {
     ".": "./sdk/mod.ts",
@@ -69,7 +69,6 @@
   },
   "scripts": {
     "setup": "git config core.hooksPath .githooks",
-    "build:cli": "node scripts/build-cli.mjs",
     "test": "vitest run",
     "lint": "biome check sdk/ ui/ cli/",
     "lint:fix": "biome check --write sdk/ ui/ cli/",

package/sdk/session.ts

@@ -250,12 +250,12 @@ export function createS2sSession(opts: SessionOptions): Session {
         });
         handle.resumeSession(s2sSessionId);
       }
-      // Send config without greeting first — greeting is deferred until
-      // the client's audio is ready (onAudioReady) to avoid a race where
-      // greeting audio arrives before the browser can play it.
+      // Send config with greeting on initial connect.
+      // Audio may arrive before the client is ready — the client buffers it.
       handle.updateSession({
         system_prompt: systemPrompt,
         tools: s2sTools,
+        ...(agentConfig.greeting ? { greeting: agentConfig.greeting } : {}),
       });
 
       handle.addEventListener("ready", ((e: CustomEvent<{ session_id: string }>) => {
@@ -400,14 +400,7 @@ export function createS2sSession(opts: SessionOptions): Session {
     onAudioReady(): void {
       if (audioReady) return;
       audioReady = true;
-      // Now that the client can play audio, send greeting via session.update.
-      if (agentConfig.greeting && s2s) {
-        s2s.updateSession({
-          system_prompt: systemPrompt,
-          tools: s2sTools,
-          greeting: agentConfig.greeting,
-        });
-      }
+      // Greeting audio + transcript come from S2S automatically.
     },
 
     onCancel(): void {

package/sdk/ws_handler.ts

@@ -100,7 +100,7 @@ export function wireSessionSocket(ws: SessionWebSocket, opts: WsSessionOptions):
 
   let session: Session | null = null;
 
-  ws.addEventListener("open", () => {
+  function onOpen(): void {
     opts.onOpen?.();
     log.info("Session connected", { ...ctx, sid });
 
@@ -113,16 +113,23 @@ export function wireSessionSocket(ws: SessionWebSocket, opts: WsSessionOptions):
 
     void session.start();
     log.info("Session ready", { ...ctx, sid });
-  });
+  }
+
+  // readyState 1 = OPEN — socket already open (e.g. from ws handleUpgrade)
+  if (ws.readyState === 1) {
+    onOpen();
+  } else {
+    ws.addEventListener("open", onOpen);
+  }
 
   ws.addEventListener("message", (event: Event) => {
     if (!session) return;
     const msgEvent = event as MessageEvent;
     const { data } = msgEvent;
 
-    // Binary frame → raw PCM16 audio
-    if (data instanceof ArrayBuffer) {
-      const chunk = new Uint8Array(data);
+    // Binary frame → raw PCM16 audio (ws package delivers Buffer)
+    if (Buffer.isBuffer(data)) {
+      const chunk = new Uint8Array(data.buffer, data.byteOffset, data.byteLength);
       if (!isValidAudioChunk(chunk)) {
         log.warn("Invalid audio chunk, dropping", {
           ...ctx,