@juspay/neurolink 13.6.5 → 13.8.0

package/dist/server/utils/validation.d.ts

@@ -173,12 +173,70 @@ export declare const TranscribeDictionaryEntrySchema: z.ZodObject<{
     heardAs: z.ZodOptional<z.ZodArray<z.ZodString>>;
     meaning: z.ZodOptional<z.ZodString>;
 }, z.core.$strip>;
-/**
- * Engine and correction settings shared by `POST /agent/transcribe`, the
- * OpenAI-compatible route's extension fields and the transcribe WebSocket's
- * config frame. Unknown keys (including `credentials`) are stripped: a caller
- * cannot swap in its own key or point an engine at another host.
- */
+/** A BCP-47 output tag (`es`, `fr-CA`, `hi-Latn`, `zh-Hant`) or `source`. */
+export declare const TranscribeOutputLanguageSchema: z.ZodString;
+/** Output languages and how they are rendered. */
+export declare const TranscribeOutputSchema: z.ZodObject<{
+    languages: z.ZodArray<z.ZodString>;
+    provider: z.ZodOptional<z.ZodString>;
+    model: z.ZodOptional<z.ZodString>;
+    timeoutMs: z.ZodOptional<z.ZodNumber>;
+    contextUtterances: z.ZodOptional<z.ZodNumber>;
+    glossary: z.ZodOptional<z.ZodBoolean>;
+}, z.core.$strip>;
+/** A text frame on the transcribe WebSocket after the config: change the output languages. */
+export declare const TranscribeOutputControlSchema: z.ZodObject<{
+    type: z.ZodLiteral<"output">;
+    languages: z.ZodArray<z.ZodString>;
+}, z.core.$strip>;
+/** A listener's frame on the session socket: which languages and speakers to follow. */
+export declare const TranscribeSubscribeSchema: z.ZodObject<{
+    type: z.ZodLiteral<"subscribe">;
+    languages: z.ZodOptional<z.ZodArray<z.ZodString>>;
+    speakers: z.ZodOptional<z.ZodArray<z.ZodString>>;
+    backlog: z.ZodOptional<z.ZodBoolean>;
+    interim: z.ZodOptional<z.ZodBoolean>;
+}, z.core.$strip>;
+export declare const STTEvalSegmentSchema: z.ZodObject<{
+    id: z.ZodString;
+    source: z.ZodEnum<{
+        primary: "primary";
+        shadow: "shadow";
+    }>;
+    model: z.ZodString;
+    participant_identity: z.ZodString;
+    participant_name: z.ZodOptional<z.ZodString>;
+    text: z.ZodString;
+    language: z.ZodOptional<z.ZodNullable<z.ZodString>>;
+    start_time: z.ZodOptional<z.ZodNullable<z.ZodNumber>>;
+    end_time: z.ZodOptional<z.ZodNullable<z.ZodNumber>>;
+    received_at: z.ZodNumber;
+}, z.core.$strip>;
+export declare const STTEvalBatchSchema: z.ZodObject<{
+    call_id: z.ZodString;
+    room: z.ZodOptional<z.ZodString>;
+    batch: z.ZodNumber;
+    final: z.ZodOptional<z.ZodBoolean>;
+    models: z.ZodObject<{
+        primary: z.ZodString;
+        shadow: z.ZodString;
+    }, z.core.$strip>;
+    segments: z.ZodArray<z.ZodObject<{
+        id: z.ZodString;
+        source: z.ZodEnum<{
+            primary: "primary";
+            shadow: "shadow";
+        }>;
+        model: z.ZodString;
+        participant_identity: z.ZodString;
+        participant_name: z.ZodOptional<z.ZodString>;
+        text: z.ZodString;
+        language: z.ZodOptional<z.ZodNullable<z.ZodString>>;
+        start_time: z.ZodOptional<z.ZodNullable<z.ZodNumber>>;
+        end_time: z.ZodOptional<z.ZodNullable<z.ZodNumber>>;
+        received_at: z.ZodNumber;
+    }, z.core.$strip>>;
+}, z.core.$strip>;
 export declare const TranscribeConfigSchema: z.ZodObject<{
     provider: z.ZodOptional<z.ZodString>;
     model: z.ZodOptional<z.ZodString>;
@@ -196,22 +254,32 @@ export declare const TranscribeConfigSchema: z.ZodObject<{
             decide: "decide";
         }>>;
         decideTimeoutMs: z.ZodOptional<z.ZodNumber>;
-        rewrite: z.ZodOptional<z.ZodUnion<readonly [z.ZodLiteral<false>, z.ZodObject<{
+        rewrite: z.ZodOptional<z.ZodUnion<readonly [z.ZodLiteral<false>, z.ZodLiteral<"auto">, z.ZodObject<{
             provider: z.ZodOptional<z.ZodString>;
             model: z.ZodOptional<z.ZodString>;
             timeoutMs: z.ZodOptional<z.ZodNumber>;
+            mode: z.ZodOptional<z.ZodEnum<{
+                auto: "auto";
+                always: "always";
+            }>>;
         }, z.core.$strip>]>>;
+        perSentence: z.ZodOptional<z.ZodBoolean>;
         transliterate: z.ZodOptional<z.ZodEnum<{
             native: "native";
             latin: "latin";
         }>>;
         punctuate: z.ZodOptional<z.ZodBoolean>;
-        secondOpinion: z.ZodOptional<z.ZodObject<{
+        cleanup: z.ZodOptional<z.ZodEnum<{
+            verbatim: "verbatim";
+            clean: "clean";
+        }>>;
+        secondOpinion: z.ZodOptional<z.ZodUnion<readonly [z.ZodLiteral<"auto">, z.ZodObject<{
             provider: z.ZodString;
             model: z.ZodOptional<z.ZodString>;
-        }, z.core.$strip>>;
+        }, z.core.$strip>]>>;
         context: z.ZodOptional<z.ZodString>;
         maxDropRatio: z.ZodOptional<z.ZodNumber>;
+        contextUtterances: z.ZodOptional<z.ZodNumber>;
     }, z.core.$strip>>;
     fallback: z.ZodOptional<z.ZodObject<{
         provider: z.ZodString;
@@ -222,6 +290,27 @@ export declare const TranscribeConfigSchema: z.ZodObject<{
             unsure: "unsure";
         }>>>;
     }, z.core.$strip>>;
+    routing: z.ZodOptional<z.ZodObject<{
+        probe: z.ZodOptional<z.ZodBoolean>;
+        lockAfterPasses: z.ZodOptional<z.ZodNumber>;
+        prior: z.ZodOptional<z.ZodEnum<{
+            none: "none";
+            carry: "carry";
+        }>>;
+        engines: z.ZodOptional<z.ZodRecord<z.ZodString, z.ZodString>>;
+    }, z.core.$strip>>;
+    chunking: z.ZodOptional<z.ZodEnum<{
+        engine: "engine";
+        utterances: "utterances";
+    }>>;
+    output: z.ZodOptional<z.ZodObject<{
+        languages: z.ZodArray<z.ZodString>;
+        provider: z.ZodOptional<z.ZodString>;
+        model: z.ZodOptional<z.ZodString>;
+        timeoutMs: z.ZodOptional<z.ZodNumber>;
+        contextUtterances: z.ZodOptional<z.ZodNumber>;
+        glossary: z.ZodOptional<z.ZodBoolean>;
+    }, z.core.$strip>>;
     diarization: z.ZodOptional<z.ZodBoolean>;
     wordTimestamps: z.ZodOptional<z.ZodBoolean>;
     timeoutMs: z.ZodOptional<z.ZodNumber>;
@@ -246,22 +335,32 @@ export declare const TranscribeRequestSchema: z.ZodObject<{
             decide: "decide";
         }>>;
         decideTimeoutMs: z.ZodOptional<z.ZodNumber>;
-        rewrite: z.ZodOptional<z.ZodUnion<readonly [z.ZodLiteral<false>, z.ZodObject<{
+        rewrite: z.ZodOptional<z.ZodUnion<readonly [z.ZodLiteral<false>, z.ZodLiteral<"auto">, z.ZodObject<{
             provider: z.ZodOptional<z.ZodString>;
             model: z.ZodOptional<z.ZodString>;
             timeoutMs: z.ZodOptional<z.ZodNumber>;
+            mode: z.ZodOptional<z.ZodEnum<{
+                auto: "auto";
+                always: "always";
+            }>>;
         }, z.core.$strip>]>>;
+        perSentence: z.ZodOptional<z.ZodBoolean>;
         transliterate: z.ZodOptional<z.ZodEnum<{
             native: "native";
             latin: "latin";
         }>>;
         punctuate: z.ZodOptional<z.ZodBoolean>;
-        secondOpinion: z.ZodOptional<z.ZodObject<{
+        cleanup: z.ZodOptional<z.ZodEnum<{
+            verbatim: "verbatim";
+            clean: "clean";
+        }>>;
+        secondOpinion: z.ZodOptional<z.ZodUnion<readonly [z.ZodLiteral<"auto">, z.ZodObject<{
             provider: z.ZodString;
             model: z.ZodOptional<z.ZodString>;
-        }, z.core.$strip>>;
+        }, z.core.$strip>]>>;
         context: z.ZodOptional<z.ZodString>;
         maxDropRatio: z.ZodOptional<z.ZodNumber>;
+        contextUtterances: z.ZodOptional<z.ZodNumber>;
     }, z.core.$strip>>;
     fallback: z.ZodOptional<z.ZodObject<{
         provider: z.ZodString;
@@ -272,6 +371,27 @@ export declare const TranscribeRequestSchema: z.ZodObject<{
             unsure: "unsure";
         }>>>;
     }, z.core.$strip>>;
+    routing: z.ZodOptional<z.ZodObject<{
+        probe: z.ZodOptional<z.ZodBoolean>;
+        lockAfterPasses: z.ZodOptional<z.ZodNumber>;
+        prior: z.ZodOptional<z.ZodEnum<{
+            none: "none";
+            carry: "carry";
+        }>>;
+        engines: z.ZodOptional<z.ZodRecord<z.ZodString, z.ZodString>>;
+    }, z.core.$strip>>;
+    chunking: z.ZodOptional<z.ZodEnum<{
+        engine: "engine";
+        utterances: "utterances";
+    }>>;
+    output: z.ZodOptional<z.ZodObject<{
+        languages: z.ZodArray<z.ZodString>;
+        provider: z.ZodOptional<z.ZodString>;
+        model: z.ZodOptional<z.ZodString>;
+        timeoutMs: z.ZodOptional<z.ZodNumber>;
+        contextUtterances: z.ZodOptional<z.ZodNumber>;
+        glossary: z.ZodOptional<z.ZodBoolean>;
+    }, z.core.$strip>>;
     diarization: z.ZodOptional<z.ZodBoolean>;
     wordTimestamps: z.ZodOptional<z.ZodBoolean>;
     timeoutMs: z.ZodOptional<z.ZodNumber>;
@@ -312,22 +432,32 @@ export declare const TranscribeStreamConfigSchema: z.ZodObject<{
             decide: "decide";
         }>>;
         decideTimeoutMs: z.ZodOptional<z.ZodNumber>;
-        rewrite: z.ZodOptional<z.ZodUnion<readonly [z.ZodLiteral<false>, z.ZodObject<{
+        rewrite: z.ZodOptional<z.ZodUnion<readonly [z.ZodLiteral<false>, z.ZodLiteral<"auto">, z.ZodObject<{
             provider: z.ZodOptional<z.ZodString>;
             model: z.ZodOptional<z.ZodString>;
             timeoutMs: z.ZodOptional<z.ZodNumber>;
+            mode: z.ZodOptional<z.ZodEnum<{
+                auto: "auto";
+                always: "always";
+            }>>;
         }, z.core.$strip>]>>;
+        perSentence: z.ZodOptional<z.ZodBoolean>;
         transliterate: z.ZodOptional<z.ZodEnum<{
             native: "native";
             latin: "latin";
         }>>;
         punctuate: z.ZodOptional<z.ZodBoolean>;
-        secondOpinion: z.ZodOptional<z.ZodObject<{
+        cleanup: z.ZodOptional<z.ZodEnum<{
+            verbatim: "verbatim";
+            clean: "clean";
+        }>>;
+        secondOpinion: z.ZodOptional<z.ZodUnion<readonly [z.ZodLiteral<"auto">, z.ZodObject<{
             provider: z.ZodString;
             model: z.ZodOptional<z.ZodString>;
-        }, z.core.$strip>>;
+        }, z.core.$strip>]>>;
         context: z.ZodOptional<z.ZodString>;
         maxDropRatio: z.ZodOptional<z.ZodNumber>;
+        contextUtterances: z.ZodOptional<z.ZodNumber>;
     }, z.core.$strip>>;
     fallback: z.ZodOptional<z.ZodObject<{
         provider: z.ZodString;
@@ -338,9 +468,34 @@ export declare const TranscribeStreamConfigSchema: z.ZodObject<{
             unsure: "unsure";
         }>>>;
     }, z.core.$strip>>;
+    routing: z.ZodOptional<z.ZodObject<{
+        probe: z.ZodOptional<z.ZodBoolean>;
+        lockAfterPasses: z.ZodOptional<z.ZodNumber>;
+        prior: z.ZodOptional<z.ZodEnum<{
+            none: "none";
+            carry: "carry";
+        }>>;
+        engines: z.ZodOptional<z.ZodRecord<z.ZodString, z.ZodString>>;
+    }, z.core.$strip>>;
+    chunking: z.ZodOptional<z.ZodEnum<{
+        engine: "engine";
+        utterances: "utterances";
+    }>>;
+    output: z.ZodOptional<z.ZodObject<{
+        languages: z.ZodArray<z.ZodString>;
+        provider: z.ZodOptional<z.ZodString>;
+        model: z.ZodOptional<z.ZodString>;
+        timeoutMs: z.ZodOptional<z.ZodNumber>;
+        contextUtterances: z.ZodOptional<z.ZodNumber>;
+        glossary: z.ZodOptional<z.ZodBoolean>;
+    }, z.core.$strip>>;
     diarization: z.ZodOptional<z.ZodBoolean>;
     wordTimestamps: z.ZodOptional<z.ZodBoolean>;
     timeoutMs: z.ZodOptional<z.ZodNumber>;
+    mode: z.ZodOptional<z.ZodEnum<{
+        file: "file";
+        live: "live";
+    }>>;
     streaming: z.ZodOptional<z.ZodObject<{
         mode: z.ZodOptional<z.ZodEnum<{
             auto: "auto";
@@ -355,8 +510,247 @@ export declare const TranscribeStreamConfigSchema: z.ZodObject<{
         onsetRms: z.ZodOptional<z.ZodNumber>;
         prerollMs: z.ZodOptional<z.ZodNumber>;
         minSpeechMs: z.ZodOptional<z.ZodNumber>;
+        concurrency: z.ZodOptional<z.ZodNumber>;
+        finalRetries: z.ZodOptional<z.ZodNumber>;
+        vad: z.ZodOptional<z.ZodObject<{
+            engine: z.ZodOptional<z.ZodEnum<{
+                auto: "auto";
+                silero: "silero";
+                energy: "energy";
+            }>>;
+            onset: z.ZodOptional<z.ZodNumber>;
+            sustain: z.ZodOptional<z.ZodNumber>;
+            startMs: z.ZodOptional<z.ZodNumber>;
+        }, z.core.$strip>>;
+        normalize: z.ZodOptional<z.ZodBoolean>;
     }, z.core.$strip>>;
 }, z.core.$strip>;
+export declare const STTRealtimeProfileSchema: z.ZodObject<{
+    timeoutMs: z.ZodOptional<z.ZodNumber>;
+    provider: z.ZodOptional<z.ZodString>;
+    output: z.ZodOptional<z.ZodObject<{
+        languages: z.ZodArray<z.ZodString>;
+        provider: z.ZodOptional<z.ZodString>;
+        model: z.ZodOptional<z.ZodString>;
+        timeoutMs: z.ZodOptional<z.ZodNumber>;
+        contextUtterances: z.ZodOptional<z.ZodNumber>;
+        glossary: z.ZodOptional<z.ZodBoolean>;
+    }, z.core.$strip>>;
+    model: z.ZodOptional<z.ZodString>;
+    streaming: z.ZodOptional<z.ZodObject<{
+        mode: z.ZodOptional<z.ZodEnum<{
+            auto: "auto";
+            native: "native";
+            chunked: "chunked";
+        }>>;
+        sampleRate: z.ZodOptional<z.ZodNumber>;
+        endSilenceMs: z.ZodOptional<z.ZodNumber>;
+        intervalMs: z.ZodOptional<z.ZodNumber>;
+        softCutSeconds: z.ZodOptional<z.ZodNumber>;
+        maxUtteranceSeconds: z.ZodOptional<z.ZodNumber>;
+        onsetRms: z.ZodOptional<z.ZodNumber>;
+        prerollMs: z.ZodOptional<z.ZodNumber>;
+        minSpeechMs: z.ZodOptional<z.ZodNumber>;
+        concurrency: z.ZodOptional<z.ZodNumber>;
+        finalRetries: z.ZodOptional<z.ZodNumber>;
+        vad: z.ZodOptional<z.ZodObject<{
+            engine: z.ZodOptional<z.ZodEnum<{
+                auto: "auto";
+                silero: "silero";
+                energy: "energy";
+            }>>;
+            onset: z.ZodOptional<z.ZodNumber>;
+            sustain: z.ZodOptional<z.ZodNumber>;
+            startMs: z.ZodOptional<z.ZodNumber>;
+        }, z.core.$strip>>;
+        normalize: z.ZodOptional<z.ZodBoolean>;
+    }, z.core.$strip>>;
+    prompt: z.ZodOptional<z.ZodString>;
+    language: z.ZodOptional<z.ZodString>;
+    diarization: z.ZodOptional<z.ZodBoolean>;
+    wordTimestamps: z.ZodOptional<z.ZodBoolean>;
+    dictionary: z.ZodOptional<z.ZodArray<z.ZodObject<{
+        term: z.ZodString;
+        heardAs: z.ZodOptional<z.ZodArray<z.ZodString>>;
+        meaning: z.ZodOptional<z.ZodString>;
+    }, z.core.$strip>>>;
+    correction: z.ZodOptional<z.ZodObject<{
+        enabled: z.ZodOptional<z.ZodBoolean>;
+        guard: z.ZodOptional<z.ZodEnum<{
+            none: "none";
+            decide: "decide";
+        }>>;
+        decideTimeoutMs: z.ZodOptional<z.ZodNumber>;
+        rewrite: z.ZodOptional<z.ZodUnion<readonly [z.ZodLiteral<false>, z.ZodLiteral<"auto">, z.ZodObject<{
+            provider: z.ZodOptional<z.ZodString>;
+            model: z.ZodOptional<z.ZodString>;
+            timeoutMs: z.ZodOptional<z.ZodNumber>;
+            mode: z.ZodOptional<z.ZodEnum<{
+                auto: "auto";
+                always: "always";
+            }>>;
+        }, z.core.$strip>]>>;
+        perSentence: z.ZodOptional<z.ZodBoolean>;
+        transliterate: z.ZodOptional<z.ZodEnum<{
+            native: "native";
+            latin: "latin";
+        }>>;
+        punctuate: z.ZodOptional<z.ZodBoolean>;
+        cleanup: z.ZodOptional<z.ZodEnum<{
+            verbatim: "verbatim";
+            clean: "clean";
+        }>>;
+        secondOpinion: z.ZodOptional<z.ZodUnion<readonly [z.ZodLiteral<"auto">, z.ZodObject<{
+            provider: z.ZodString;
+            model: z.ZodOptional<z.ZodString>;
+        }, z.core.$strip>]>>;
+        context: z.ZodOptional<z.ZodString>;
+        maxDropRatio: z.ZodOptional<z.ZodNumber>;
+        contextUtterances: z.ZodOptional<z.ZodNumber>;
+    }, z.core.$strip>>;
+    fallback: z.ZodOptional<z.ZodObject<{
+        provider: z.ZodString;
+        model: z.ZodOptional<z.ZodString>;
+        when: z.ZodOptional<z.ZodArray<z.ZodEnum<{
+            error: "error";
+            empty: "empty";
+            unsure: "unsure";
+        }>>>;
+    }, z.core.$strip>>;
+    routing: z.ZodOptional<z.ZodObject<{
+        probe: z.ZodOptional<z.ZodBoolean>;
+        lockAfterPasses: z.ZodOptional<z.ZodNumber>;
+        prior: z.ZodOptional<z.ZodEnum<{
+            none: "none";
+            carry: "carry";
+        }>>;
+        engines: z.ZodOptional<z.ZodRecord<z.ZodString, z.ZodString>>;
+    }, z.core.$strip>>;
+    chunking: z.ZodOptional<z.ZodEnum<{
+        engine: "engine";
+        utterances: "utterances";
+    }>>;
+    completed: z.ZodOptional<z.ZodEnum<{
+        final: "final";
+        corrected: "corrected";
+    }>>;
+    correctionWaitMs: z.ZodOptional<z.ZodNumber>;
+    clientLanguage: z.ZodOptional<z.ZodBoolean>;
+    clientTurnDetection: z.ZodOptional<z.ZodBoolean>;
+}, z.core.$strip>;
+export declare const STTRealtimeProfilesSchema: z.ZodRecord<z.ZodString, z.ZodObject<{
+    timeoutMs: z.ZodOptional<z.ZodNumber>;
+    provider: z.ZodOptional<z.ZodString>;
+    output: z.ZodOptional<z.ZodObject<{
+        languages: z.ZodArray<z.ZodString>;
+        provider: z.ZodOptional<z.ZodString>;
+        model: z.ZodOptional<z.ZodString>;
+        timeoutMs: z.ZodOptional<z.ZodNumber>;
+        contextUtterances: z.ZodOptional<z.ZodNumber>;
+        glossary: z.ZodOptional<z.ZodBoolean>;
+    }, z.core.$strip>>;
+    model: z.ZodOptional<z.ZodString>;
+    streaming: z.ZodOptional<z.ZodObject<{
+        mode: z.ZodOptional<z.ZodEnum<{
+            auto: "auto";
+            native: "native";
+            chunked: "chunked";
+        }>>;
+        sampleRate: z.ZodOptional<z.ZodNumber>;
+        endSilenceMs: z.ZodOptional<z.ZodNumber>;
+        intervalMs: z.ZodOptional<z.ZodNumber>;
+        softCutSeconds: z.ZodOptional<z.ZodNumber>;
+        maxUtteranceSeconds: z.ZodOptional<z.ZodNumber>;
+        onsetRms: z.ZodOptional<z.ZodNumber>;
+        prerollMs: z.ZodOptional<z.ZodNumber>;
+        minSpeechMs: z.ZodOptional<z.ZodNumber>;
+        concurrency: z.ZodOptional<z.ZodNumber>;
+        finalRetries: z.ZodOptional<z.ZodNumber>;
+        vad: z.ZodOptional<z.ZodObject<{
+            engine: z.ZodOptional<z.ZodEnum<{
+                auto: "auto";
+                silero: "silero";
+                energy: "energy";
+            }>>;
+            onset: z.ZodOptional<z.ZodNumber>;
+            sustain: z.ZodOptional<z.ZodNumber>;
+            startMs: z.ZodOptional<z.ZodNumber>;
+        }, z.core.$strip>>;
+        normalize: z.ZodOptional<z.ZodBoolean>;
+    }, z.core.$strip>>;
+    prompt: z.ZodOptional<z.ZodString>;
+    language: z.ZodOptional<z.ZodString>;
+    diarization: z.ZodOptional<z.ZodBoolean>;
+    wordTimestamps: z.ZodOptional<z.ZodBoolean>;
+    dictionary: z.ZodOptional<z.ZodArray<z.ZodObject<{
+        term: z.ZodString;
+        heardAs: z.ZodOptional<z.ZodArray<z.ZodString>>;
+        meaning: z.ZodOptional<z.ZodString>;
+    }, z.core.$strip>>>;
+    correction: z.ZodOptional<z.ZodObject<{
+        enabled: z.ZodOptional<z.ZodBoolean>;
+        guard: z.ZodOptional<z.ZodEnum<{
+            none: "none";
+            decide: "decide";
+        }>>;
+        decideTimeoutMs: z.ZodOptional<z.ZodNumber>;
+        rewrite: z.ZodOptional<z.ZodUnion<readonly [z.ZodLiteral<false>, z.ZodLiteral<"auto">, z.ZodObject<{
+            provider: z.ZodOptional<z.ZodString>;
+            model: z.ZodOptional<z.ZodString>;
+            timeoutMs: z.ZodOptional<z.ZodNumber>;
+            mode: z.ZodOptional<z.ZodEnum<{
+                auto: "auto";
+                always: "always";
+            }>>;
+        }, z.core.$strip>]>>;
+        perSentence: z.ZodOptional<z.ZodBoolean>;
+        transliterate: z.ZodOptional<z.ZodEnum<{
+            native: "native";
+            latin: "latin";
+        }>>;
+        punctuate: z.ZodOptional<z.ZodBoolean>;
+        cleanup: z.ZodOptional<z.ZodEnum<{
+            verbatim: "verbatim";
+            clean: "clean";
+        }>>;
+        secondOpinion: z.ZodOptional<z.ZodUnion<readonly [z.ZodLiteral<"auto">, z.ZodObject<{
+            provider: z.ZodString;
+            model: z.ZodOptional<z.ZodString>;
+        }, z.core.$strip>]>>;
+        context: z.ZodOptional<z.ZodString>;
+        maxDropRatio: z.ZodOptional<z.ZodNumber>;
+        contextUtterances: z.ZodOptional<z.ZodNumber>;
+    }, z.core.$strip>>;
+    fallback: z.ZodOptional<z.ZodObject<{
+        provider: z.ZodString;
+        model: z.ZodOptional<z.ZodString>;
+        when: z.ZodOptional<z.ZodArray<z.ZodEnum<{
+            error: "error";
+            empty: "empty";
+            unsure: "unsure";
+        }>>>;
+    }, z.core.$strip>>;
+    routing: z.ZodOptional<z.ZodObject<{
+        probe: z.ZodOptional<z.ZodBoolean>;
+        lockAfterPasses: z.ZodOptional<z.ZodNumber>;
+        prior: z.ZodOptional<z.ZodEnum<{
+            none: "none";
+            carry: "carry";
+        }>>;
+        engines: z.ZodOptional<z.ZodRecord<z.ZodString, z.ZodString>>;
+    }, z.core.$strip>>;
+    chunking: z.ZodOptional<z.ZodEnum<{
+        engine: "engine";
+        utterances: "utterances";
+    }>>;
+    completed: z.ZodOptional<z.ZodEnum<{
+        final: "final";
+        corrected: "corrected";
+    }>>;
+    correctionWaitMs: z.ZodOptional<z.ZodNumber>;
+    clientLanguage: z.ZodOptional<z.ZodBoolean>;
+    clientTurnDetection: z.ZodOptional<z.ZodBoolean>;
+}, z.core.$strip>>;
 /**
  * Type guard to check if a value is an ErrorResponse
  */

package/dist/server/utils/validation.js

@@ -180,12 +180,63 @@ export const TranscribeDictionaryEntrySchema = z.object({
     meaning: z.string().max(2000).optional(),
 });
 const sttProviderName = z.string().min(1).max(100);
+/** A BCP-47 output tag (`es`, `fr-CA`, `hi-Latn`, `zh-Hant`) or `source`. */
+export const TranscribeOutputLanguageSchema = z
+    .string()
+    .trim()
+    .regex(/^(?:source|[A-Za-z]{2,3}(?:-[A-Za-z]{4})?(?:-(?:[A-Za-z]{2}|\d{3}))?)$/u, "must be a BCP-47 tag such as es, fr-CA or hi-Latn, or source");
+/** Output languages and how they are rendered. */
+export const TranscribeOutputSchema = z.object({
+    languages: z.array(TranscribeOutputLanguageSchema).max(8),
+    provider: z.string().min(1).max(100).optional(),
+    model: z.string().min(1).max(200).optional(),
+    timeoutMs: z.number().int().positive().max(120_000).optional(),
+    contextUtterances: z.number().int().min(0).max(10).optional(),
+    glossary: z.boolean().optional(),
+});
+/** A text frame on the transcribe WebSocket after the config: change the output languages. */
+export const TranscribeOutputControlSchema = z.object({
+    type: z.literal("output"),
+    languages: z.array(TranscribeOutputLanguageSchema).max(8),
+});
+/** A listener's frame on the session socket: which languages and speakers to follow. */
+export const TranscribeSubscribeSchema = z.object({
+    type: z.literal("subscribe"),
+    languages: z.array(TranscribeOutputLanguageSchema).max(8).optional(),
+    speakers: z.array(z.string().min(1).max(100)).max(64).optional(),
+    backlog: z.boolean().optional(),
+    interim: z.boolean().optional(),
+});
 /**
  * Engine and correction settings shared by `POST /agent/transcribe`, the
  * OpenAI-compatible route's extension fields and the transcribe WebSocket's
  * config frame. Unknown keys (including `credentials`) are stripped: a caller
  * cannot swap in its own key or point an engine at another host.
  */
+const EvalTimeSchema = z.number().finite().nonnegative().nullish();
+export const STTEvalSegmentSchema = z.object({
+    id: z.string().min(1).max(200),
+    source: z.enum(["primary", "shadow"]),
+    model: z.string().max(200),
+    participant_identity: z.string().max(500),
+    participant_name: z.string().max(500).optional(),
+    text: z.string().max(20_000),
+    language: z.string().max(35).nullish(),
+    start_time: EvalTimeSchema,
+    end_time: EvalTimeSchema,
+    received_at: z.number().finite().nonnegative(),
+});
+export const STTEvalBatchSchema = z.object({
+    call_id: z.string().min(1).max(500),
+    room: z.string().max(500).optional(),
+    batch: z.number().int().nonnegative(),
+    final: z.boolean().optional(),
+    models: z.object({
+        primary: z.string().max(200),
+        shadow: z.string().max(200),
+    }),
+    segments: z.array(STTEvalSegmentSchema).max(5_000),
+});
 export const TranscribeConfigSchema = z.object({
     provider: sttProviderName.optional(),
     model: z.string().min(1).max(200).optional(),
@@ -200,23 +251,31 @@ export const TranscribeConfigSchema = z.object({
         rewrite: z
             .union([
             z.literal(false),
+            z.literal("auto"),
             z.object({
                 provider: z.string().min(1).max(100).optional(),
                 model: z.string().min(1).max(200).optional(),
                 timeoutMs: z.number().int().positive().max(120_000).optional(),
+                mode: z.enum(["auto", "always"]).optional(),
             }),
         ])
             .optional(),
+        perSentence: z.boolean().optional(),
         transliterate: z.enum(["latin", "native"]).optional(),
         punctuate: z.boolean().optional(),
+        cleanup: z.enum(["verbatim", "clean"]).optional(),
         secondOpinion: z
-            .object({
-            provider: sttProviderName,
-            model: z.string().min(1).max(200).optional(),
-        })
+            .union([
+            z.literal("auto"),
+            z.object({
+                provider: sttProviderName,
+                model: z.string().min(1).max(200).optional(),
+            }),
+        ])
             .optional(),
         context: z.string().max(1000).optional(),
         maxDropRatio: z.number().min(0).max(1).optional(),
+        contextUtterances: z.number().int().min(0).max(10).optional(),
     })
         .optional(),
     fallback: z
@@ -226,6 +285,21 @@ export const TranscribeConfigSchema = z.object({
         when: z.array(z.enum(["unsure", "empty", "error"])).optional(),
     })
         .optional(),
+    routing: z
+        .object({
+        probe: z.boolean().optional(),
+        lockAfterPasses: z.number().int().min(1).max(5).optional(),
+        prior: z.enum(["carry", "none"]).optional(),
+        engines: z
+            .record(z.string().min(1).max(35), sttProviderName)
+            .refine((engines) => Object.keys(engines).length <= 64, {
+            message: "at most 64 routes",
+        })
+            .optional(),
+    })
+        .optional(),
+    chunking: z.enum(["engine", "utterances"]).optional(),
+    output: TranscribeOutputSchema.optional(),
     diarization: z.boolean().optional(),
     wordTimestamps: z.boolean().optional(),
     timeoutMs: z.number().int().positive().max(600_000).optional(),
@@ -257,6 +331,13 @@ export const TranscribeRequestSchema = TranscribeConfigSchema.extend({
  * The transcribe WebSocket's first text frame: the config plus streaming tuning.
  */
 export const TranscribeStreamConfigSchema = TranscribeConfigSchema.extend({
+    /**
+     * `"file"`: the socket carries a whole recording as fast as the client can
+     * send it, not a microphone. No interim passes, utterances transcribed
+     * several at a time, a failed final retried, file-sized engine timeouts and
+     * shorter utterance cuts, with results per utterance as they finish.
+     */
+    mode: z.enum(["live", "file"]).optional(),
     streaming: z
         .object({
         mode: z.enum(["auto", "native", "chunked"]).optional(),
@@ -268,9 +349,30 @@ export const TranscribeStreamConfigSchema = TranscribeConfigSchema.extend({
         onsetRms: z.number().positive().optional(),
         prerollMs: z.number().nonnegative().optional(),
         minSpeechMs: z.number().nonnegative().optional(),
+        concurrency: z.number().int().min(1).max(8).optional(),
+        finalRetries: z.number().int().min(0).max(3).optional(),
+        // The detector and its thresholds; where its model lives is the server's business.
+        vad: z
+            .object({
+            engine: z.enum(["auto", "silero", "energy"]).optional(),
+            onset: z.number().min(0).max(1).optional(),
+            sustain: z.number().min(0).max(1).optional(),
+            startMs: z.number().nonnegative().max(2000).optional(),
+        })
+            .optional(),
+        normalize: z.boolean().optional(),
     })
         .optional(),
 });
+export const STTRealtimeProfileSchema = TranscribeStreamConfigSchema.omit({
+    mode: true,
+}).extend({
+    completed: z.enum(["corrected", "final"]).optional(),
+    correctionWaitMs: z.number().int().min(0).max(30_000).optional(),
+    clientLanguage: z.boolean().optional(),
+    clientTurnDetection: z.boolean().optional(),
+});
+export const STTRealtimeProfilesSchema = z.record(z.string().min(1).max(100), STTRealtimeProfileSchema);
 // ============================================
 // Error Response Type Guards / Helpers
 // ============================================