@alexkroman1/aai 1.5.1 → 1.6.0

package/.turbo/turbo-build.log

@@ -1,5 +1,5 @@
 
-> @alexkroman1/aai@1.5.1 build /home/runner/work/agent/agent/packages/aai
+> @alexkroman1/aai@1.6.0 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       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/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/_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/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/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
+ℹ dist/host/runtime-barrel.js       103.81 kB │ gzip: 29.39 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/llm-barrel.js    0.36 kB │ gzip:  0.21 kB
+ℹ dist/sdk/providers/stt-barrel.js    0.30 kB │ gzip:  0.20 kB
+ℹ dist/sdk/providers/tts-barrel.js    0.26 kB │ gzip:  0.16 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/constants-C2nirZUI.js          3.10 kB │ gzip:  1.38 kB
+ℹ dist/soniox-DCQ3GqJq.js             2.03 kB │ gzip:  0.54 kB
+ℹ dist/xai-jfQsxxPZ.js                1.18 kB │ gzip:  0.35 kB
+ℹ dist/ws-upgrade-BeOQ7fXL.js         1.14 kB │ gzip:  0.54 kB
+ℹ dist/cartesia-BfQPOQ7Y.js           1.08 kB │ gzip:  0.50 kB
+ℹ 14 files, total: 132.56 kB
+✔ Build complete in 43ms

package/CHANGELOG.md

@@ -1,5 +1,17 @@
 # @alexkroman1/aai
 
+## 1.6.0
+
+### Minor Changes
+
+- fd3a167: Pluggable Vercel AI SDK LLM providers in pipeline mode: add openai, google, mistral, xai, groq typed factories alongside the existing anthropic. Each is a { model } descriptor; the host resolver lazy-loads the corresponding @ai-sdk/\* package via createRequire. All six AI SDK packages move to optional peer dependencies, so self-hosted users only install the ones they actually use; the managed server installs all six as direct deps in aai-server.
+- c8707d6: Add ElevenLabs Scribe (scribe_v2_realtime via @elevenlabs/elevenlabs-js) and Soniox (stt-rt-v3 via direct WebSocket) STT providers alongside assemblyai and deepgram. Both follow the existing typed-descriptor pattern; agent bundles stay free of provider SDKs and the host resolver constructs the live session at createRuntime time.
+
+### Patch Changes
+
+- 149786b: Auto-resume AssemblyAI S2S sessions after transient WebSocket closes (1005, 1006, 1011, 3005) using session.resume within the 30s server window. Drops the in-flight reply via onCancelled so the session unblocks; falls back to the existing 'connection' error on fatal codes (1008/3006/3007/3008/3009) or when resume fails.
+- 877348c: Pipeline mode: insert separator between LLM text segments split by a mid-turn tool call so consecutive deltas don't fuse into '...up.Got it' in the transcript and TTS output.
+
 ## 1.5.1
 
 ### Patch Changes

package/dist/host/providers/resolve.d.ts

@@ -10,6 +10,12 @@
  * The guest sandbox never imports these functions, which is how the agent
  * bundle stays free of `@ai-sdk/anthropic` / `assemblyai` /
  * `@cartesia/cartesia-js`.
+ *
+ * `@ai-sdk/*` packages are loaded via `createRequire` lazily so self-hosted
+ * users only need to install the providers they actually reference. A
+ * missing package is reported as a friendly "package not installed" error
+ * at first session start, not as a module-load failure when the host
+ * package itself is imported.
  */
 import type { LanguageModel } from "ai";
 import type { LlmProvider, SttOpener, SttProvider, TtsOpener, TtsProvider } from "../../sdk/providers.ts";
@@ -27,7 +33,7 @@ export declare function resolveTts(descriptor: TtsProvider): TtsOpener;
  * Resolve an {@link LlmProvider} descriptor into a Vercel AI SDK
  * {@link LanguageModel}.
  *
- * The API key is pulled from the agent's env (e.g. `ANTHROPIC_API_KEY`).
+ * The API key is pulled from the agent's env (e.g. `OPENAI_API_KEY`).
  * Missing keys throw here — the pipeline session would fail on first
  * `streamText` call otherwise, and the error is clearer at construction.
  */

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

@@ -0,0 +1,16 @@
+/**
+ * ElevenLabs Scribe streaming STT opener (host-only).
+ *
+ * The user-facing descriptor factory (`elevenlabs(...)`) lives in
+ * `sdk/providers/stt/elevenlabs.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: `"scribe_v2_realtime"`. Audio is sent as base64-encoded
+ * PCM_16000; partial transcripts arrive on `transcript`, finals on
+ * `committed_transcript`.
+ */
+import type { ElevenLabsOptions } from "../../../sdk/providers/stt/elevenlabs.ts";
+import { type SttOpener } from "../../../sdk/providers.ts";
+/** Build an {@link SttOpener} from resolved ElevenLabs descriptor options. */
+export declare function openElevenLabs(opts?: ElevenLabsOptions): SttOpener;

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

@@ -0,0 +1,25 @@
+/**
+ * Soniox real-time STT opener (host-only).
+ *
+ * The user-facing descriptor factory (`soniox(...)`) lives in
+ * `sdk/providers/stt/soniox.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.
+ *
+ * Soniox's published JS client (`@soniox/speech-to-text-web`) is
+ * browser-only — it depends on `MediaRecorder` and `getUserMedia`. For
+ * server-side use we talk to the WebSocket directly:
+ *   `wss://stt-rt.soniox.com/transcribe-websocket`
+ *
+ * Wire format:
+ *   - First text frame: JSON config with api_key, model, audio_format,
+ *     sample_rate, num_channels (and optional language hints).
+ *   - Subsequent binary frames: 16-bit signed little-endian PCM audio.
+ *   - Server replies: JSON `{ tokens: [{ text, is_final }] }` messages.
+ *     Final tokens accumulate; non-final tokens are a rolling preview.
+ *   - On error: `{ error_code, error_message }`.
+ */
+import type { SonioxOptions } from "../../../sdk/providers/stt/soniox.ts";
+import { type SttOpener } from "../../../sdk/providers.ts";
+/** Build an {@link SttOpener} from resolved Soniox descriptor options. */
+export declare function openSoniox(opts?: SonioxOptions): SttOpener;