@fonoster/autopilot 0.25.0 → 0.26.0

package/README.md

@@ -201,4 +201,31 @@ Please use the following environment variables to configure the Autopilot:
 
 The Autopilot is still under heavy development. The next steps are to add support for Retrieval-Augmented Generation(RAG), improve the language model integration, and improve the Finite State Machine (FSM) that powers the conversations. If you have any suggestions or feedback, please let us know.
 
+## Audio filters
+
+Background voices reach speech recognition and the voice activity detector, and the assistant
+answers them. An assistant can ask the Media Server to filter the caller's audio first, by adding
+`audioFilters` to its configuration — the same block whether the assistant comes from a file or
+from an Application:
+
+```yaml
+audioFilters:
+  - name: aiCoustics
+    options:
+      enhancementLevel: 0.8 # 0 to 1; higher also suppresses competing voices
+
+conversationSettings:
+  firstMessage: "Hola, ¿en qué puedo ayudarte?"
+  # ...
+```
+
+For an Application, the same block goes inside `intelligence.config`, and is applied with:
+
+```sh
+fonoster applications:create -f app.yaml
+```
 
+Assistant files may be YAML or JSON. Filtering needs the optional `@ai-coustics/aic-sdk` package
+and an `AIC_SDK_LICENSE` key on the apiserver; the key never appears in assistant configuration,
+so these files stay safe to commit. If the filter cannot be applied, the call continues with
+unfiltered audio and the reason is logged.

package/dist/Autopilot.d.ts

@@ -6,6 +6,7 @@ declare class Autopilot {
     constructor(params: AutopilotParams);
     start(): Promise<void>;
     stop(): void;
+    private setupAudioFilters;
     private setupVoiceStream;
     private handleVoicePayload;
     private setupSpeechGathering;

package/dist/Autopilot.js

@@ -41,6 +41,7 @@ class Autopilot {
         });
     }
     async start() {
+        await this.setupAudioFilters();
         const vadParams = this.params.conversationSettings.vad;
         const sileroVad = new vad_1.SileroVad({
             pathToModel: vadParams.pathToModel ||
@@ -63,6 +64,26 @@ class Autopilot {
         logger.verbose("stopping autopilot");
         this.actor.stop();
     }
+    // Asked for before the audio stream is opened, so voice activity detection
+    // and speech recognition both see filtered audio from the first frame. A
+    // refusal is logged and the call continues unfiltered.
+    async setupAudioFilters() {
+        const { voice, audioFilters } = this.params;
+        if (!audioFilters?.length) {
+            return;
+        }
+        try {
+            await voice.setAudioFilters(audioFilters);
+            logger.verbose("audio filters requested", {
+                filters: audioFilters.map((filter) => filter.name)
+            });
+        }
+        catch (err) {
+            logger.warn("could not set audio filters; continuing unfiltered", {
+                error: err?.message
+            });
+        }
+    }
     async setupVoiceStream() {
         const { voice } = this.params;
         const stream = await voice.stream();

package/dist/assistants/loadAndValidateAssistant.d.ts

@@ -1,3 +1,3 @@
 import { AssistantConfig } from "./types";
-declare function loadAndValidateAssistant(path: string): AssistantConfig;
+declare function loadAndValidateAssistant(assistantPath: string): AssistantConfig;
 export { loadAndValidateAssistant };

package/dist/assistants/loadAndValidateAssistant.js

@@ -23,22 +23,25 @@ exports.loadAndValidateAssistant = loadAndValidateAssistant;
  * limitations under the License.
  */
 const fs_1 = __importDefault(require("fs"));
+const path_1 = __importDefault(require("path"));
 const common_1 = require("@fonoster/common");
 const logger_1 = require("@fonoster/logger");
+const js_yaml_1 = require("js-yaml");
 const logger = (0, logger_1.getLogger)({ service: "autopilot", filePath: __filename });
-function loadAndValidateAssistant(path) {
-    if (!fs_1.default.existsSync(path)) {
-        logger.error("assistant file not found", { path });
+function loadAndValidateAssistant(assistantPath) {
+    if (!fs_1.default.existsSync(assistantPath)) {
+        logger.error("assistant file not found", { path: assistantPath });
         process.exit(1);
     }
     try {
-        const fileContent = fs_1.default.readFileSync(path, "utf8");
-        const assistant = JSON.parse(fileContent);
+        const fileContent = fs_1.default.readFileSync(assistantPath, "utf8");
+        const isYaml = [".yaml", ".yml"].includes(path_1.default.extname(assistantPath).toLowerCase());
+        const assistant = (isYaml ? (0, js_yaml_1.load)(fileContent) : JSON.parse(fileContent));
         return common_1.assistantSchema.parse(assistant);
     }
     catch (e) {
         logger.error("error parsing or validating assistant file", {
-            path,
+            path: assistantPath,
             error: e
         });
         process.exit(1);

package/dist/handleVoiceRequest.js

@@ -120,6 +120,7 @@ async function handleVoiceRequest(req, res) {
     const { conversationSettings } = assistantConfig;
     try {
         const autopilot = new _1.default({
+            audioFilters: assistantConfig.audioFilters,
             conversationSettings: conversationSettings,
             voice: voice,
             languageModel: languageModel

package/dist/models/evaluations/createEvalEffectiveConfig.d.ts

@@ -86,6 +86,10 @@ export declare function createEvalEffectiveConfig(config: AssistantConfig, crede
             }[];
         }[];
     };
+    audioFilters?: {
+        options?: Record<string, unknown>;
+        name?: string;
+    }[];
     conversationSettings?: {
         firstMessage?: string;
         systemPrompt?: string;

package/dist/types.d.ts

@@ -16,6 +16,7 @@
  * See the License for the specific language governing permissions and
  * limitations under the License.
  */
+import { AudioFilterConfig } from "@fonoster/common";
 import { Application } from "@fonoster/types";
 import { AssistantConfig, ConversationSettings } from "./assistants";
 import { LanguageModel } from "./models";
@@ -28,6 +29,7 @@ type AutopilotParams = {
     voice: Voice;
     conversationSettings: ConversationSettings;
     languageModel: LanguageModel;
+    audioFilters?: AudioFilterConfig[];
 };
 type AutopilotApplication = Application & {
     intelligence: {

package/dist/voice/Voice.d.ts

@@ -1,3 +1,22 @@
+/**
+ * Copyright (C) 2025 by Fonoster Inc (https://fonoster.com)
+ * http://github.com/fonoster/fonoster
+ *
+ * This file is part of Fonoster
+ *
+ * Licensed under the MIT License (the "License");
+ * you may not use this file except in compliance with
+ * the License. You may obtain a copy of the License at
+ *
+ *    https://opensource.org/licenses/MIT
+ *
+ * Unless required by applicable law or agreed to in writing, software
+ * distributed under the License is distributed on an "AS IS" BASIS,
+ * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ * See the License for the specific language governing permissions and
+ * limitations under the License.
+ */
+import { AudioFilterConfig } from "@fonoster/common";
 import { VoiceResponse } from "@fonoster/voice";
 import { Voice } from "./types";
 declare class VoiceImpl implements Voice {
@@ -17,6 +36,7 @@ declare class VoiceImpl implements Voice {
     };
     constructor(mediaSessionRef: string, voice: VoiceResponse);
     answer(): Promise<void>;
+    setAudioFilters(filters: AudioFilterConfig[]): Promise<void>;
     hangup(): Promise<void>;
     say(text: string): Promise<void>;
     playDtmf(dtmf: string): Promise<void>;

package/dist/voice/Voice.js

@@ -30,6 +30,9 @@ class VoiceImpl {
     async answer() {
         await this.voice.answer();
     }
+    async setAudioFilters(filters) {
+        await this.voice.setAudioFilters(filters);
+    }
     async hangup() {
         await this.voice.hangup();
     }

package/dist/voice/types.d.ts

@@ -16,6 +16,7 @@
  * See the License for the specific language governing permissions and
  * limitations under the License.
  */
+import { AudioFilterConfig } from "@fonoster/common";
 type GatherStream = {
     stop: () => Promise<void>;
     onData: (cb: (payload: {
@@ -34,6 +35,7 @@ type TransferOptions = {
 type Voice = {
     mediaSessionRef: string;
     answer: () => Promise<void>;
+    setAudioFilters: (filters: AudioFilterConfig[]) => Promise<void>;
     hangup: () => Promise<void>;
     say: (text: string) => Promise<void>;
     playDtmf: (dtmf: string) => Promise<void>;

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "@fonoster/autopilot",
-  "version": "0.25.0",
+  "version": "0.26.0",
   "description": "Voice AI for the Fonoster platform",
   "author": "Pedro Sanders <psanders@fonoster.com>",
   "homepage": "https://github.com/fonoster/fonoster#readme",
@@ -35,11 +35,11 @@
   "dependencies": {
     "@aws-sdk/client-s3": "^3.958.0",
     "@dmitryrechkin/json-schema-to-zod": "^1.0.1",
-    "@fonoster/common": "^0.25.0",
-    "@fonoster/logger": "^0.25.0",
-    "@fonoster/sdk": "^0.25.0",
-    "@fonoster/types": "^0.25.0",
-    "@fonoster/voice": "^0.25.0",
+    "@fonoster/common": "^0.26.0",
+    "@fonoster/logger": "^0.26.0",
+    "@fonoster/sdk": "^0.26.0",
+    "@fonoster/types": "^0.26.0",
+    "@fonoster/voice": "^0.26.0",
     "@langchain/anthropic": "^1.3.3",
     "@langchain/community": "^1.1.1",
     "@langchain/core": "^1.1.8",
@@ -59,5 +59,5 @@
     "xstate": "^5.17.3",
     "zod": "^3.25.76"
   },
-  "gitHead": "3f63a173d31c0dbf2391b1eff907c8880e60614d"
+  "gitHead": "d14f414d791a33f92e8c4225f7e462abc440a142"
 }