@alexkroman1/aai 1.4.5 → 1.5.1

package/.turbo/turbo-build.log

@@ -1,5 +1,5 @@
 
-> @alexkroman1/aai@1.4.5 build /home/runner/work/agent/agent/packages/aai
+> @alexkroman1/aai@1.5.1 build /home/runner/work/agent/agent/packages/aai
 > tsdown && tsc -p tsconfig.build.json
 
 ℹ tsdown v0.21.7 powered by rolldown v1.0.0-rc.12
@@ -8,19 +8,19 @@
 ℹ target: node22
 ℹ tsconfig: tsconfig.json
 ℹ Build start
-ℹ dist/host/runtime-barrel.js       77.48 kB │ gzip: 23.08 kB
+ℹ dist/host/runtime-barrel.js       87.68 kB │ gzip: 25.35 kB
 ℹ dist/sdk/protocol.js               4.75 kB │ gzip:  1.76 kB
 ℹ dist/index.js                      2.88 kB │ gzip:  1.24 kB
 ℹ dist/sdk/manifest-barrel.js        0.36 kB │ gzip:  0.20 kB
-ℹ dist/sdk/providers/stt-barrel.js   0.13 kB │ gzip:  0.11 kB
+ℹ dist/sdk/providers/tts-barrel.js   0.26 kB │ gzip:  0.16 kB
+ℹ dist/sdk/providers/stt-barrel.js   0.19 kB │ gzip:  0.14 kB
 ℹ dist/sdk/providers/llm-barrel.js   0.12 kB │ gzip:  0.11 kB
-ℹ dist/sdk/providers/tts-barrel.js   0.12 kB │ gzip:  0.10 kB
+ℹ dist/_internal-types-DFL07G3f.js   5.66 kB │ gzip:  2.10 kB
 ℹ dist/types-KUgezM6u.js             5.64 kB │ gzip:  2.39 kB
-ℹ dist/_internal-types-3p3OJZPb.js   5.61 kB │ gzip:  2.07 kB
 ℹ dist/constants-C2nirZUI.js         3.10 kB │ gzip:  1.38 kB
 ℹ dist/ws-upgrade-BeOQ7fXL.js        1.14 kB │ gzip:  0.54 kB
-ℹ dist/assemblyai-Cxg9eobY.js        0.53 kB │ gzip:  0.35 kB
-ℹ dist/anthropic-BrUCPKUc.js         0.23 kB │ gzip:  0.18 kB
-ℹ dist/cartesia-DwDk2tEu.js          0.22 kB │ gzip:  0.17 kB
-ℹ 14 files, total: 102.30 kB
-✔ Build complete in 49ms
+ℹ dist/cartesia-BfQPOQ7Y.js          1.08 kB │ gzip:  0.50 kB
+ℹ dist/assemblyai-C969QGi4.js        1.03 kB │ gzip:  0.42 kB
+ℹ dist/anthropic-CcLZygAr.js         0.23 kB │ gzip:  0.18 kB
+ℹ 14 files, total: 114.11 kB
+✔ Build complete in 45ms

package/CHANGELOG.md

@@ -1,5 +1,24 @@
 # @alexkroman1/aai
 
+## 1.5.1
+
+### Patch Changes
+
+- fbb3816: Add type: "function" to tool schemas in S2S session.update payload — AssemblyAI's S2S API rejects tool objects without it.
+
+## 1.5.0
+
+### Minor Changes
+
+- 58c5c75: Consolidate session.ts + pipeline-session.ts into a unified SessionCore with two transport strategies (S2S, pipeline). Switch connectS2s to typed callbacks (removing the nanoevents-backed S2sHandle emitter) and flatten client→server→provider dispatch from four layers to two. Wire format is JSON text events + raw PCM16 binary audio frames — the existing public protocol is unchanged. Adds Deepgram as a pipeline-mode STT option and Rime as a pipeline-mode TTS option.
+- 868b85e: Plumb agent maxSteps and toolChoice config into pipeline mode streamText
+- 58c5c75: Add Deepgram as a pipeline-mode STT provider option
+- 58c5c75: feat(aai): add Rime as a pipeline-mode TTS provider option
+
+### Patch Changes
+
+- a361363: Fix Rime TTS provider: correct WebSocket host (users-ws.rime.ai), JSON message protocol on /ws2, longer first-audio timeout so the greeting plays. Default voice for cartesia() and rime() so they can be called with no args.
+
 ## 1.4.5
 
 ### Patch Changes

package/dist/{_internal-types-3p3OJZPb.js → _internal-types-DFL07G3f.js}

@@ -122,6 +122,7 @@ function toAgentConfig(src) {
 * etc.) — the Vercel AI SDK wraps it via `jsonSchema()`.
 */
 const ToolSchemaSchema = z.object({
+	type: z.literal("function"),
 	name: z.string().min(1),
 	description: z.string().min(1),
 	parameters: z.record(z.string(), z.unknown())
@@ -136,6 +137,7 @@ const EMPTY_PARAMS = z.object({});
 */
 function agentToolsToSchemas(tools) {
 	return Object.entries(tools).map(([name, def]) => ({
+		type: "function",
 		name,
 		description: def.description,
 		parameters: z.toJSONSchema(def.parameters ?? EMPTY_PARAMS)

package/dist/assemblyai-C969QGi4.js

@@ -0,0 +1,35 @@
+//#region sdk/providers/stt/deepgram.ts
+/** Kind tag recognised by the host-side resolver. */
+const DEEPGRAM_KIND = "deepgram";
+/**
+* Build a Deepgram STT descriptor.
+*
+* The API key is resolved host-side from the agent's env
+* (`DEEPGRAM_API_KEY`); there is no factory-time key parameter, so the
+* descriptor stays free of secrets and safe to serialize.
+*/
+function deepgram(opts = {}) {
+	return {
+		kind: DEEPGRAM_KIND,
+		options: { ...opts }
+	};
+}
+//#endregion
+//#region sdk/providers/stt/assemblyai.ts
+/** Kind tag recognised by the host-side resolver. */
+const ASSEMBLYAI_KIND = "assemblyai";
+/**
+* Build an AssemblyAI STT descriptor.
+*
+* The API key is resolved host-side from the agent's env
+* (`ASSEMBLYAI_API_KEY`); there is no factory-time key parameter, so the
+* descriptor stays free of secrets and safe to serialize.
+*/
+function assemblyAI(opts = {}) {
+	return {
+		kind: ASSEMBLYAI_KIND,
+		options: { ...opts }
+	};
+}
+//#endregion
+export { deepgram as i, assemblyAI as n, DEEPGRAM_KIND as r, ASSEMBLYAI_KIND as t };

package/dist/cartesia-BfQPOQ7Y.js

@@ -0,0 +1,37 @@
+//#region sdk/providers/tts/rime.ts
+const RIME_KIND = "rime";
+/**
+* Default Rime speaker used when callers invoke `rime()` with no `voice`.
+* `cove` is a `mistv2` speaker, matching the default model below — so a
+* bare `rime()` works out of the box for new agents.
+*/
+const RIME_DEFAULT_VOICE = "cove";
+function rime(opts = {}) {
+	return {
+		kind: RIME_KIND,
+		options: {
+			...opts,
+			voice: opts.voice ?? "cove"
+		}
+	};
+}
+//#endregion
+//#region sdk/providers/tts/cartesia.ts
+const CARTESIA_KIND = "cartesia";
+/**
+* Default voice used when callers invoke `cartesia()` with no `voice`. This
+* is the same voice the example templates ship with, so a bare `cartesia()`
+* works out of the box for new agents.
+*/
+const CARTESIA_DEFAULT_VOICE = "f786b574-daa5-4673-aa0c-cbe3e8534c02";
+function cartesia(opts = {}) {
+	return {
+		kind: CARTESIA_KIND,
+		options: {
+			...opts,
+			voice: opts.voice ?? "f786b574-daa5-4673-aa0c-cbe3e8534c02"
+		}
+	};
+}
+//#endregion
+export { RIME_KIND as a, RIME_DEFAULT_VOICE as i, CARTESIA_KIND as n, rime as o, cartesia as r, CARTESIA_DEFAULT_VOICE as t };

package/dist/host/_pipeline-test-fakes.d.ts

@@ -114,4 +114,6 @@ export declare function createFakeLanguageModel(options: {
 } | {
     steps: ScriptedPart[][];
     delayMs?: number;
-}): LanguageModel;
+}): LanguageModel & {
+    readonly calls: readonly Record<string, unknown>[];
+};

package/dist/host/providers/stt/deepgram.d.ts

@@ -0,0 +1,28 @@
+/**
+ * Deepgram Nova streaming STT opener (host-only).
+ *
+ * The user-facing descriptor factory (`deepgram(...)`) lives in
+ * `sdk/providers/stt/deepgram.ts`. This module is the host-side
+ * counterpart: it takes the descriptor options + an API key and
+ * returns an {@link SttOpener} that the pipeline session drives.
+ *
+ * Default model: `"nova-3"`. Any string is forwarded verbatim to the SDK.
+ *
+ * This adapter targets the Deepgram SDK v5 (`@deepgram/sdk@^5`). The v5
+ * streaming API is:
+ *   `client.listen.v1.connect(args)` → `Promise<V1Socket>`
+ * followed by:
+ *   `socket.connect()` + `socket.waitForOpen()` to establish the connection.
+ */
+import { DeepgramClient } from "@deepgram/sdk";
+import type { DeepgramOptions } from "../../../sdk/providers/stt/deepgram.ts";
+import { type SttOpener, type SttSession } from "../../../sdk/providers.ts";
+type V1Socket = Awaited<ReturnType<InstanceType<typeof DeepgramClient>["listen"]["v1"]["connect"]>>;
+/** Internal: SttSession with a test-only handle to the raw SDK socket. */
+export interface DeepgramSession extends SttSession {
+    /** @internal Test-only: exposes the underlying SDK socket for fixture replay. */
+    readonly _connection: V1Socket;
+}
+/** Build an {@link SttOpener} from resolved Deepgram descriptor options. */
+export declare function openDeepgram(opts?: DeepgramOptions): SttOpener;
+export {};

package/dist/host/providers/tts/cartesia.d.ts

@@ -19,7 +19,7 @@
  * conversion.
  */
 import type { TTSWS } from "@cartesia/cartesia-js/resources/tts";
-import type { CartesiaOptions } from "../../../sdk/providers/tts/cartesia.ts";
+import { type CartesiaOptions } from "../../../sdk/providers/tts/cartesia.ts";
 import { type TtsOpener, type TtsSession } from "../../../sdk/providers.ts";
 /** Internal: TtsSession with a test-only handle to the raw SDK socket. */
 export interface CartesiaSession extends TtsSession {

package/dist/host/providers/tts/rime.d.ts

@@ -0,0 +1,44 @@
+/**
+ * Rime TTS opener (host-only).
+ *
+ * The user-facing descriptor factory (`rime(...)`) lives in
+ * `sdk/providers/tts/rime.ts`. This module is the host-side
+ * counterpart: it takes the descriptor options + an API key and
+ * returns a {@link TtsOpener} that the pipeline session drives.
+ *
+ * **Protocol.** Connects to Rime's `ws2` JSON WebSocket endpoint
+ * (`wss://users-ws.rime.ai/ws2`). Client-to-server messages are JSON:
+ *   - `{ "text": "..." }` — append text to the synthesis buffer
+ *   - `{ "operation": "clear" }` — drop buffered text (barge-in)
+ *   - `{ "operation": "eos" }` — drain buffer, close connection (NOT used
+ *     during a session: it would tear down the WS, forcing reconnect per
+ *     turn). We force end-of-turn synthesis with a trailing `"."` instead.
+ * The server responds with JSON frames:
+ *   - `{ type: "chunk", data: <base64 PCM16 LE>, contextId: string | null }`
+ *   - `{ type: "timestamps", ... }` (ignored)
+ *   - `{ type: "error", message: string }` (surfaced as `tts_stream_error`)
+ *
+ * **Single long-lived connection per session.** Rime buffers text until it
+ * sees terminal punctuation (`.`, `?`, `!`), so we use one WebSocket per
+ * `open()` call and reuse it across turns. `clear` resets the buffer
+ * between cancellations.
+ *
+ * **Done detection.** After `flush()` sends a trailing `"."` to force the
+ * server to synthesize any half-buffered text, we arm a quiescence timer
+ * that fires 500 ms after the last received audio chunk. When it fires,
+ * `done` is emitted.
+ *
+ * **Audio format.** The URL requests `audioFormat=pcm` at the negotiated
+ * `sampleRate`, which returns raw PCM16 little-endian. We decode the base64
+ * payload and construct a zero-copy `Int16Array` view over the decoded bytes.
+ */
+import WebSocket from "ws";
+import { type RimeOptions } from "../../../sdk/providers/tts/rime.ts";
+import { type TtsOpener, type TtsSession } from "../../../sdk/providers.ts";
+/** Internal: TtsSession with a test-only handle to the raw WebSocket. */
+export interface RimeSession extends TtsSession {
+    /** @internal Test-only: exposes the underlying raw WebSocket. */
+    readonly _ws: WebSocket;
+}
+/** Build a {@link TtsOpener} from resolved Rime descriptor options. */
+export declare function openRime(opts: RimeOptions): TtsOpener;

package/dist/host/runtime-barrel.d.ts

@@ -7,8 +7,10 @@ export * from "./builtin-tools.ts";
 export * from "./runtime.ts";
 export * from "./runtime-config.ts";
 export * from "./server.ts";
-export * from "./session.ts";
-export * from "./session-ctx.ts";
+export * from "./session-core.ts";
 export * from "./tool-executor.ts";
+export * from "./transports/pipeline-transport.ts";
+export * from "./transports/s2s-transport.ts";
+export * from "./transports/types.ts";
 export * from "./unstorage-kv.ts";
 export * from "./ws-handler.ts";