@kaana.ai/contracts 5.0.0 → 6.1.0

package/dist/internal/catalogue.d.ts

@@ -112,50 +112,6 @@ export declare const catalogueModelSchema: z.ZodObject<{
                     g711_alaw: "g711_alaw";
                 }>>;
             }, z.core.$strict>>;
-            parameters: z.ZodOptional<z.ZodObject<{
-                turnDetection: z.ZodArray<z.ZodObject<{
-                    type: z.ZodEnum<{
-                        none: "none";
-                        server_vad: "server_vad";
-                        semantic_vad: "semantic_vad";
-                    }>;
-                    fixedResponses: z.ZodBoolean;
-                    threshold: z.ZodOptional<z.ZodObject<{
-                        min: z.ZodNumber;
-                        max: z.ZodNumber;
-                    }, z.core.$strict>>;
-                    meteredUnits: z.ZodArray<z.ZodEnum<{
-                        input_tokens: "input_tokens";
-                        cached_input_tokens: "cached_input_tokens";
-                        output_tokens: "output_tokens";
-                        reasoning_tokens: "reasoning_tokens";
-                        requests: "requests";
-                        images: "images";
-                        audio_input_milliseconds: "audio_input_milliseconds";
-                        audio_output_milliseconds: "audio_output_milliseconds";
-                        video_milliseconds: "video_milliseconds";
-                        characters: "characters";
-                        embeddings: "embeddings";
-                        audio_input_tokens: "audio_input_tokens";
-                        cached_audio_input_tokens: "cached_audio_input_tokens";
-                        audio_output_tokens: "audio_output_tokens";
-                        session_milliseconds: "session_milliseconds";
-                        input_transcription_milliseconds: "input_transcription_milliseconds";
-                    }>>;
-                }, z.core.$strict>>;
-                outputModalities: z.ZodArray<z.ZodArray<z.ZodEnum<{
-                    text: "text";
-                    audio: "audio";
-                }>>>;
-                toolChoice: z.ZodArray<z.ZodEnum<{
-                    function: "function";
-                    none: "none";
-                    auto: "auto";
-                    required: "required";
-                }>>;
-                maxOutputTokens: z.ZodOptional<z.ZodNumber>;
-                inputTranscriptionPrompt: z.ZodBoolean;
-            }, z.core.$strict>>;
         }, z.core.$strict>>;
     }, z.core.$strict>;
     license: z.ZodObject<{
@@ -509,50 +465,6 @@ export declare const modelCatalogueEntrySchema: z.ZodObject<{
                     g711_alaw: "g711_alaw";
                 }>>;
             }, z.core.$strict>>;
-            parameters: z.ZodOptional<z.ZodObject<{
-                turnDetection: z.ZodArray<z.ZodObject<{
-                    type: z.ZodEnum<{
-                        none: "none";
-                        server_vad: "server_vad";
-                        semantic_vad: "semantic_vad";
-                    }>;
-                    fixedResponses: z.ZodBoolean;
-                    threshold: z.ZodOptional<z.ZodObject<{
-                        min: z.ZodNumber;
-                        max: z.ZodNumber;
-                    }, z.core.$strict>>;
-                    meteredUnits: z.ZodArray<z.ZodEnum<{
-                        input_tokens: "input_tokens";
-                        cached_input_tokens: "cached_input_tokens";
-                        output_tokens: "output_tokens";
-                        reasoning_tokens: "reasoning_tokens";
-                        requests: "requests";
-                        images: "images";
-                        audio_input_milliseconds: "audio_input_milliseconds";
-                        audio_output_milliseconds: "audio_output_milliseconds";
-                        video_milliseconds: "video_milliseconds";
-                        characters: "characters";
-                        embeddings: "embeddings";
-                        audio_input_tokens: "audio_input_tokens";
-                        cached_audio_input_tokens: "cached_audio_input_tokens";
-                        audio_output_tokens: "audio_output_tokens";
-                        session_milliseconds: "session_milliseconds";
-                        input_transcription_milliseconds: "input_transcription_milliseconds";
-                    }>>;
-                }, z.core.$strict>>;
-                outputModalities: z.ZodArray<z.ZodArray<z.ZodEnum<{
-                    text: "text";
-                    audio: "audio";
-                }>>>;
-                toolChoice: z.ZodArray<z.ZodEnum<{
-                    function: "function";
-                    none: "none";
-                    auto: "auto";
-                    required: "required";
-                }>>;
-                maxOutputTokens: z.ZodOptional<z.ZodNumber>;
-                inputTranscriptionPrompt: z.ZodBoolean;
-            }, z.core.$strict>>;
         }, z.core.$strict>>;
     }, z.core.$strict>;
     license: z.ZodObject<{
@@ -705,8 +617,6 @@ export declare const declaredModelSchema: z.ZodObject<{
             }>>;
         }, z.core.$strict>>;
         languages: z.ZodArray<z.ZodString>;
-        detectsLanguage: z.ZodOptional<z.ZodBoolean>;
-        transcriptionPrompt: z.ZodOptional<z.ZodBoolean>;
         audioFormats: z.ZodObject<{
             input: z.ZodArray<z.ZodEnum<{
                 "audio/mpeg": "audio/mpeg";
@@ -803,50 +713,6 @@ export declare const declaredModelSchema: z.ZodObject<{
                     g711_alaw: "g711_alaw";
                 }>>;
             }, z.core.$strict>>;
-            parameters: z.ZodOptional<z.ZodObject<{
-                turnDetection: z.ZodArray<z.ZodObject<{
-                    type: z.ZodEnum<{
-                        none: "none";
-                        server_vad: "server_vad";
-                        semantic_vad: "semantic_vad";
-                    }>;
-                    fixedResponses: z.ZodBoolean;
-                    threshold: z.ZodOptional<z.ZodObject<{
-                        min: z.ZodNumber;
-                        max: z.ZodNumber;
-                    }, z.core.$strict>>;
-                    meteredUnits: z.ZodArray<z.ZodEnum<{
-                        input_tokens: "input_tokens";
-                        cached_input_tokens: "cached_input_tokens";
-                        output_tokens: "output_tokens";
-                        reasoning_tokens: "reasoning_tokens";
-                        requests: "requests";
-                        images: "images";
-                        audio_input_milliseconds: "audio_input_milliseconds";
-                        audio_output_milliseconds: "audio_output_milliseconds";
-                        video_milliseconds: "video_milliseconds";
-                        characters: "characters";
-                        embeddings: "embeddings";
-                        audio_input_tokens: "audio_input_tokens";
-                        cached_audio_input_tokens: "cached_audio_input_tokens";
-                        audio_output_tokens: "audio_output_tokens";
-                        session_milliseconds: "session_milliseconds";
-                        input_transcription_milliseconds: "input_transcription_milliseconds";
-                    }>>;
-                }, z.core.$strict>>;
-                outputModalities: z.ZodArray<z.ZodArray<z.ZodEnum<{
-                    text: "text";
-                    audio: "audio";
-                }>>>;
-                toolChoice: z.ZodArray<z.ZodEnum<{
-                    function: "function";
-                    none: "none";
-                    auto: "auto";
-                    required: "required";
-                }>>;
-                maxOutputTokens: z.ZodOptional<z.ZodNumber>;
-                inputTranscriptionPrompt: z.ZodBoolean;
-            }, z.core.$strict>>;
         }, z.core.$strict>>;
     }, z.core.$strict>;
     license: z.ZodObject<{
@@ -917,6 +783,10 @@ export declare const declaredModelSchema: z.ZodObject<{
             currency: z.ZodString;
         }, z.core.$strict>>;
     }, z.core.$strict>>;
+    pricePerMinute: z.ZodOptional<z.ZodObject<{
+        amount: z.core.$ZodBranded<z.ZodString, "ExactDecimal", "out">;
+        currency: z.ZodString;
+    }, z.core.$strict>>;
     evaluations: z.ZodDefault<z.ZodArray<z.ZodObject<{
         suite: z.ZodString;
         metric: z.ZodString;
@@ -975,8 +845,6 @@ export declare const declaredCatalogueSchema: z.ZodObject<{
                 }>>;
             }, z.core.$strict>>;
             languages: z.ZodArray<z.ZodString>;
-            detectsLanguage: z.ZodOptional<z.ZodBoolean>;
-            transcriptionPrompt: z.ZodOptional<z.ZodBoolean>;
             audioFormats: z.ZodObject<{
                 input: z.ZodArray<z.ZodEnum<{
                     "audio/mpeg": "audio/mpeg";
@@ -1073,50 +941,6 @@ export declare const declaredCatalogueSchema: z.ZodObject<{
                         g711_alaw: "g711_alaw";
                     }>>;
                 }, z.core.$strict>>;
-                parameters: z.ZodOptional<z.ZodObject<{
-                    turnDetection: z.ZodArray<z.ZodObject<{
-                        type: z.ZodEnum<{
-                            none: "none";
-                            server_vad: "server_vad";
-                            semantic_vad: "semantic_vad";
-                        }>;
-                        fixedResponses: z.ZodBoolean;
-                        threshold: z.ZodOptional<z.ZodObject<{
-                            min: z.ZodNumber;
-                            max: z.ZodNumber;
-                        }, z.core.$strict>>;
-                        meteredUnits: z.ZodArray<z.ZodEnum<{
-                            input_tokens: "input_tokens";
-                            cached_input_tokens: "cached_input_tokens";
-                            output_tokens: "output_tokens";
-                            reasoning_tokens: "reasoning_tokens";
-                            requests: "requests";
-                            images: "images";
-                            audio_input_milliseconds: "audio_input_milliseconds";
-                            audio_output_milliseconds: "audio_output_milliseconds";
-                            video_milliseconds: "video_milliseconds";
-                            characters: "characters";
-                            embeddings: "embeddings";
-                            audio_input_tokens: "audio_input_tokens";
-                            cached_audio_input_tokens: "cached_audio_input_tokens";
-                            audio_output_tokens: "audio_output_tokens";
-                            session_milliseconds: "session_milliseconds";
-                            input_transcription_milliseconds: "input_transcription_milliseconds";
-                        }>>;
-                    }, z.core.$strict>>;
-                    outputModalities: z.ZodArray<z.ZodArray<z.ZodEnum<{
-                        text: "text";
-                        audio: "audio";
-                    }>>>;
-                    toolChoice: z.ZodArray<z.ZodEnum<{
-                        function: "function";
-                        none: "none";
-                        auto: "auto";
-                        required: "required";
-                    }>>;
-                    maxOutputTokens: z.ZodOptional<z.ZodNumber>;
-                    inputTranscriptionPrompt: z.ZodBoolean;
-                }, z.core.$strict>>;
             }, z.core.$strict>>;
         }, z.core.$strict>;
         license: z.ZodObject<{
@@ -1187,6 +1011,10 @@ export declare const declaredCatalogueSchema: z.ZodObject<{
                 currency: z.ZodString;
             }, z.core.$strict>>;
         }, z.core.$strict>>;
+        pricePerMinute: z.ZodOptional<z.ZodObject<{
+            amount: z.core.$ZodBranded<z.ZodString, "ExactDecimal", "out">;
+            currency: z.ZodString;
+        }, z.core.$strict>>;
         evaluations: z.ZodDefault<z.ZodArray<z.ZodObject<{
             suite: z.ZodString;
             metric: z.ZodString;

package/dist/internal/realtime.d.ts

@@ -102,12 +102,12 @@ export declare const realtimeSessionRequestSchema: z.ZodObject<{
         tools: z.ZodOptional<z.ZodArray<z.ZodObject<{
             type: z.ZodLiteral<"function">;
             name: z.ZodString;
-            parameters: z.ZodRecord<z.ZodString, z.ZodUnknown>;
             description: z.ZodOptional<z.ZodString>;
+            parameters: z.ZodRecord<z.ZodString, z.ZodUnknown>;
         }, z.core.$strict>>>;
         toolChoice: z.ZodOptional<z.ZodUnion<readonly [z.ZodEnum<{
-            none: "none";
             auto: "auto";
+            none: "none";
             required: "required";
         }>, z.ZodObject<{
             type: z.ZodLiteral<"function">;
@@ -242,12 +242,12 @@ export declare const realtimeSessionCreatedEventSchema: z.ZodObject<{
         tools: z.ZodOptional<z.ZodArray<z.ZodObject<{
             type: z.ZodLiteral<"function">;
             name: z.ZodString;
-            parameters: z.ZodRecord<z.ZodString, z.ZodUnknown>;
             description: z.ZodOptional<z.ZodString>;
+            parameters: z.ZodRecord<z.ZodString, z.ZodUnknown>;
         }, z.core.$strict>>>;
         toolChoice: z.ZodOptional<z.ZodUnion<readonly [z.ZodEnum<{
-            none: "none";
             auto: "auto";
+            none: "none";
             required: "required";
         }>, z.ZodObject<{
             type: z.ZodLiteral<"function">;
@@ -499,12 +499,12 @@ export declare const realtimeServerEventSchema: z.ZodDiscriminatedUnion<[z.ZodOb
         tools: z.ZodOptional<z.ZodArray<z.ZodObject<{
             type: z.ZodLiteral<"function">;
             name: z.ZodString;
-            parameters: z.ZodRecord<z.ZodString, z.ZodUnknown>;
             description: z.ZodOptional<z.ZodString>;
+            parameters: z.ZodRecord<z.ZodString, z.ZodUnknown>;
         }, z.core.$strict>>>;
         toolChoice: z.ZodOptional<z.ZodUnion<readonly [z.ZodEnum<{
-            none: "none";
             auto: "auto";
+            none: "none";
             required: "required";
         }>, z.ZodObject<{
             type: z.ZodLiteral<"function">;
@@ -575,12 +575,12 @@ export declare const realtimeServerEventSchema: z.ZodDiscriminatedUnion<[z.ZodOb
         tools: z.ZodOptional<z.ZodArray<z.ZodObject<{
             type: z.ZodLiteral<"function">;
             name: z.ZodString;
-            parameters: z.ZodRecord<z.ZodString, z.ZodUnknown>;
             description: z.ZodOptional<z.ZodString>;
+            parameters: z.ZodRecord<z.ZodString, z.ZodUnknown>;
         }, z.core.$strict>>>;
         toolChoice: z.ZodOptional<z.ZodUnion<readonly [z.ZodEnum<{
-            none: "none";
             auto: "auto";
+            none: "none";
             required: "required";
         }>, z.ZodObject<{
             type: z.ZodLiteral<"function">;

package/dist/internal/request.d.ts

@@ -366,8 +366,8 @@ export declare const inferenceRequestSchema: z.ZodObject<{
         strict: z.ZodOptional<z.ZodBoolean>;
     }, z.core.$strict>>>;
     toolChoice: z.ZodOptional<z.ZodUnion<readonly [z.ZodEnum<{
-        none: "none";
         auto: "auto";
+        none: "none";
         required: "required";
     }>, z.ZodObject<{
         type: z.ZodLiteral<"function">;
@@ -645,8 +645,8 @@ export declare const scopedInferenceRequestSchema: z.ZodObject<{
         strict: z.ZodOptional<z.ZodBoolean>;
     }, z.core.$strict>>>;
     toolChoice: z.ZodOptional<z.ZodUnion<readonly [z.ZodEnum<{
-        none: "none";
         auto: "auto";
+        none: "none";
         required: "required";
     }>, z.ZodObject<{
         type: z.ZodLiteral<"function">;
@@ -876,15 +876,6 @@ export declare const privateAutoInferenceRequestSchema: z.ZodObject<{
         language: z.ZodOptional<z.ZodString>;
         prompt: z.ZodOptional<z.ZodString>;
     }, z.core.$strict>>;
-    toolChoice: z.ZodOptional<z.ZodUnion<readonly [z.ZodEnum<{
-        none: "none";
-        auto: "auto";
-        required: "required";
-    }>, z.ZodObject<{
-        type: z.ZodLiteral<"function">;
-        name: z.ZodString;
-    }, z.core.$strict>]>>;
-    maxOutputTokens: z.ZodOptional<z.ZodNumber>;
     tools: z.ZodDefault<z.ZodArray<z.ZodObject<{
         type: z.ZodLiteral<"function">;
         name: z.ZodString;
@@ -892,6 +883,7 @@ export declare const privateAutoInferenceRequestSchema: z.ZodObject<{
         parameters: z.ZodRecord<z.ZodString, z.ZodUnknown>;
         strict: z.ZodOptional<z.ZodBoolean>;
     }, z.core.$strict>>>;
+    maxOutputTokens: z.ZodOptional<z.ZodNumber>;
     stream: z.ZodBoolean;
     responseFormat: z.ZodOptional<z.ZodDiscriminatedUnion<[z.ZodObject<{
         type: z.ZodLiteral<"text">;
@@ -903,6 +895,14 @@ export declare const privateAutoInferenceRequestSchema: z.ZodObject<{
         schema: z.ZodRecord<z.ZodString, z.ZodUnknown>;
         strict: z.ZodBoolean;
     }, z.core.$strict>], "type">>;
+    toolChoice: z.ZodOptional<z.ZodUnion<readonly [z.ZodEnum<{
+        auto: "auto";
+        none: "none";
+        required: "required";
+    }>, z.ZodObject<{
+        type: z.ZodLiteral<"function">;
+        name: z.ZodString;
+    }, z.core.$strict>]>>;
     routingPolicy: z.ZodObject<{
         routingPolicyId: z.ZodString;
         policyVersion: z.ZodNumber;

package/dist/public/audio.d.ts

@@ -320,23 +320,19 @@ export declare const transcriptionSchema: z.ZodObject<{
 /** Why a transcription request cannot be served by a model, before anything is held. */
 export interface TranscriptionRequestIssue {
     /** The request's field at fault, as `invalid_request.param` names it. */
-    readonly param: 'file' | 'language' | 'prompt';
+    readonly param: 'file';
     readonly message: string;
 }
 /**
- * Check a transcription request against the model's published `audio` — the
- * transcription counterpart of {@link checkSpeechRequest}. Kaana refuses a
- * request with `invalid_request` and the first issue's `param` BEFORE asking
- * Oxy for a grant; a client may run the same check to choose a model, and
- * omit `language` where `detectsLanguage` lets the model detect it.
- *
- * `mediaType` is the upload's (the multipart `file` part's content type, or
- * `audio.mediaType` of the JSON form). Returns every issue, in field order.
+ * Check a transcription upload against the model's published `audio` — the
+ * transcription counterpart of {@link checkSpeechRequest}. Kaana refuses an
+ * upload in a format the model does not read with `invalid_request` and
+ * `param: 'file'` BEFORE asking Oxy for a grant. A language or a prompt is
+ * never refused: Kaana sends the provider its spelling of the language, or
+ * none for the model to detect, and the prompt where it is taken (Kaana ADR
+ * 0003). `mediaType` is the upload's.
  */
-export declare function checkTranscriptionRequest(request: {
-    readonly language?: string;
-    readonly prompt?: string;
-}, mediaType: AudioInputMediaType, audio: ModelAudioCapabilities): TranscriptionRequestIssue[];
+export declare function checkTranscriptionRequest(mediaType: AudioInputMediaType, audio: ModelAudioCapabilities): TranscriptionRequestIssue[];
 export type SpeechStreamFormat = z.infer<typeof speechStreamFormatSchema>;
 export type SpeechRequest = z.infer<typeof speechRequestSchema>;
 export type SpeechAudioDeltaEvent = z.infer<typeof speechAudioDeltaEventSchema>;

package/dist/public/audio.js

@@ -161,13 +161,13 @@ export const transcriptionRequestSchema = z
     .object({
     model: modelReferenceSchema,
     /**
-     * The spoken language, when known: one of the model's `audio.languages`.
-     * Absent lets a model that `detectsLanguage` detect it.
+     * The spoken language, when known, as the client's own BCP 47 tag. Kaana
+     * sends the provider its spelling of it, or none for the model to detect.
      */
     language: languageTagSchema.optional(),
     /**
-     * Vocabulary or context that helps the model; never an instruction. Only
-     * for a model whose `audio.transcriptionPrompt` is true.
+     * A hint — vocabulary or context that helps the model, never an
+     * instruction — used where the model takes one.
      */
     prompt: z.string().max(4096).optional(),
 })
@@ -221,29 +221,16 @@ export const transcriptionSchema = z
 })
     .strict();
 /**
- * Check a transcription request against the model's published `audio` — the
- * transcription counterpart of {@link checkSpeechRequest}. Kaana refuses a
- * request with `invalid_request` and the first issue's `param` BEFORE asking
- * Oxy for a grant; a client may run the same check to choose a model, and
- * omit `language` where `detectsLanguage` lets the model detect it.
- *
- * `mediaType` is the upload's (the multipart `file` part's content type, or
- * `audio.mediaType` of the JSON form). Returns every issue, in field order.
+ * Check a transcription upload against the model's published `audio` — the
+ * transcription counterpart of {@link checkSpeechRequest}. Kaana refuses an
+ * upload in a format the model does not read with `invalid_request` and
+ * `param: 'file'` BEFORE asking Oxy for a grant. A language or a prompt is
+ * never refused: Kaana sends the provider its spelling of the language, or
+ * none for the model to detect, and the prompt where it is taken (Kaana ADR
+ * 0003). `mediaType` is the upload's.
  */
-export function checkTranscriptionRequest(request, mediaType, audio) {
-    const issues = [];
-    if (!audio.audioFormats.input.includes(mediaType)) {
-        issues.push({ param: 'file', message: `this model does not take ${mediaType} audio` });
-    }
-    if (request.language !== undefined && !audio.languages.includes(request.language)) {
-        issues.push({
-            param: 'language',
-            message: `this model does not declare the language ${request.language}` +
-                (audio.detectsLanguage === true ? '; omit it to let the model detect the language' : ''),
-        });
-    }
-    if (request.prompt !== undefined && audio.transcriptionPrompt !== true) {
-        issues.push({ param: 'prompt', message: 'this model takes no prompt' });
-    }
-    return issues;
+export function checkTranscriptionRequest(mediaType, audio) {
+    return audio.audioFormats.input.includes(mediaType)
+        ? []
+        : [{ param: 'file', message: `this model does not take ${mediaType} audio` }];
 }

package/dist/public/audioFormats.d.ts

@@ -51,24 +51,13 @@ export declare const realtimeAudioFormatSchema: z.ZodEnum<{
     g711_alaw: "g711_alaw";
 }>;
 /**
- * What a session's responses produce. Declared here, beside the encodings and
- * for the same reason, since contract set 5.0.0: the catalogue publishes which
- * sets of them a model serves (`capabilities.realtime.parameters`).
+ * What a session's responses produce. Declared here, beside the encodings,
+ * since contract set 5.0.0.
  */
 export declare const realtimeOutputModalitySchema: z.ZodEnum<{
     text: "text";
     audio: "audio";
 }>;
-/**
- * How the end of a caller's turn is detected: the discriminant of
- * `realtimeTurnDetectionSchema`, published per model by the catalogue
- * (contract set 5.0.0).
- */
-export declare const realtimeTurnDetectionTypeSchema: z.ZodEnum<{
-    none: "none";
-    server_vad: "server_vad";
-    semantic_vad: "semantic_vad";
-}>;
 /**
  * The media type a speech request's `response_format` produces. A client
  * knows it before the first byte, which is why the speech stream events do

package/dist/public/audioFormats.js

@@ -43,17 +43,10 @@ export const audioInputMediaTypeSchema = z.enum([
  */
 export const realtimeAudioFormatSchema = z.enum(['pcm16_24khz', 'g711_ulaw', 'g711_alaw']);
 /**
- * What a session's responses produce. Declared here, beside the encodings and
- * for the same reason, since contract set 5.0.0: the catalogue publishes which
- * sets of them a model serves (`capabilities.realtime.parameters`).
+ * What a session's responses produce. Declared here, beside the encodings,
+ * since contract set 5.0.0.
  */
 export const realtimeOutputModalitySchema = z.enum(['text', 'audio']);
-/**
- * How the end of a caller's turn is detected: the discriminant of
- * `realtimeTurnDetectionSchema`, published per model by the catalogue
- * (contract set 5.0.0).
- */
-export const realtimeTurnDetectionTypeSchema = z.enum(['none', 'server_vad', 'semantic_vad']);
 const SPEECH_MEDIA_TYPES = {
     mp3: 'audio/mpeg',
     opus: 'audio/ogg',