fix(vscode): preserve spoken transcription wording

.changeset/verbatim-speech-transcripts.md

@@ -0,0 +1,6 @@
+---
+"kilo-code": patch
+"@kilocode/kilo-gateway": patch
+---
+
+Keep GPT-4o voice transcriptions closer to the words that were spoken.

packages/kilo-gateway/src/server/routes.ts

@@ -429,6 +429,7 @@ export function createKiloRoutes(deps: KiloRoutesDeps) {
             format: z.string(),
           }),
           language: z.string().optional(),
+          prompt: z.string().optional(),
           temperature: z.number().optional(),
         }),
       ),

packages/kilo-vscode/src/speech-to-text/models.ts

@@ -2,6 +2,7 @@ export interface SpeechToTextModelDef {
   readonly id: string
   readonly label: string
   readonly provider: string
+  readonly verbatim?: boolean
 }
 
 const models: SpeechToTextModelDef[] = [
@@ -9,11 +10,13 @@ const models: SpeechToTextModelDef[] = [
     id: "openai/gpt-4o-mini-transcribe",
     label: "GPT-4o Mini Transcribe",
     provider: "OpenAI",
+    verbatim: true,
   },
   {
     id: "openai/gpt-4o-transcribe",
     label: "GPT-4o Transcribe",
     provider: "OpenAI",
+    verbatim: true,
   },
   {
     id: "openai/whisper-1",

packages/kilo-vscode/src/speech-to-text/transcribe.ts

@@ -3,6 +3,8 @@ import { getErrorMessage } from "../kilo-provider-utils"
 import { getSpeechToTextModel } from "./models"
 
 const PATH = "/kilo/audio/transcriptions"
+const PROMPT =
+  "Transcribe exactly what is spoken. Do not paraphrase, summarize, infer intent, or rewrite for clarity. Preserve the speaker's original wording as closely as possible, including incomplete phrases and unusual wording when audible."
 
 type Req = {
   model?: string
@@ -39,6 +41,8 @@ export async function transcribeSpeech(
 
   const auth = Buffer.from(`kilo:${cfg.password}`).toString("base64")
   const url = new URL(PATH, cfg.baseUrl)
+  const model = getSpeechToTextModel(input.model)
+  const prompt = model.verbatim ? PROMPT : undefined
   if (dir) url.searchParams.set("directory", dir)
 
   try {
@@ -50,12 +54,13 @@ export async function transcribeSpeech(
         "Content-Type": "application/json",
       },
       body: JSON.stringify({
-        model: getSpeechToTextModel(input.model).id,
+        model: model.id,
         input_audio: {
           data: input.data,
           format: input.format,
         },
         ...(input.language ? { language: input.language } : {}),
+        ...(prompt ? { prompt } : {}),
       }),
     })
 

packages/sdk/js/src/v2/gen/sdk.gen.ts

@@ -5763,6 +5763,7 @@ export class Audio extends HeyApiClient {
         format: string
       }
       language?: string
+      prompt?: string
       temperature?: number
     },
     options?: Options<never, ThrowOnError>,
@@ -5777,6 +5778,7 @@ export class Audio extends HeyApiClient {
             { in: "body", key: "model" },
             { in: "body", key: "input_audio" },
             { in: "body", key: "language" },
+            { in: "body", key: "prompt" },
             { in: "body", key: "temperature" },
           ],
         },

packages/sdk/js/src/v2/gen/types.gen.ts

@@ -7255,6 +7255,7 @@ export type KiloAudioTranscriptionsData = {
       format: string
     }
     language?: string
+    prompt?: string
     temperature?: number
   }
   path?: never

packages/sdk/openapi.json

@@ -10481,6 +10481,100 @@
         ]
       }
     },
+    "/kilo/audio/transcriptions": {
+      "post": {
+        "operationId": "kilo.audio.transcriptions",
+        "parameters": [
+          {
+            "in": "query",
+            "name": "directory",
+            "schema": {
+              "type": "string"
+            }
+          },
+          {
+            "in": "query",
+            "name": "workspace",
+            "schema": {
+              "type": "string"
+            }
+          }
+        ],
+        "summary": "Speech to text transcription",
+        "description": "Proxy an audio transcription request to the Kilo Gateway",
+        "responses": {
+          "200": {
+            "description": "Transcription response",
+            "content": {
+              "application/json": {
+                "schema": {
+                  "type": "object",
+                  "properties": {
+                    "text": {
+                      "type": "string"
+                    },
+                    "usage": {}
+                  },
+                  "required": ["text"]
+                }
+              }
+            }
+          },
+          "400": {
+            "description": "Bad request",
+            "content": {
+              "application/json": {
+                "schema": {
+                  "$ref": "#/components/schemas/BadRequestError"
+                }
+              }
+            }
+          }
+        },
+        "requestBody": {
+          "content": {
+            "application/json": {
+              "schema": {
+                "type": "object",
+                "properties": {
+                  "model": {
+                    "type": "string"
+                  },
+                  "input_audio": {
+                    "type": "object",
+                    "properties": {
+                      "data": {
+                        "type": "string"
+                      },
+                      "format": {
+                        "type": "string"
+                      }
+                    },
+                    "required": ["data", "format"]
+                  },
+                  "language": {
+                    "type": "string"
+                  },
+                  "prompt": {
+                    "type": "string"
+                  },
+                  "temperature": {
+                    "type": "number"
+                  }
+                },
+                "required": ["model", "input_audio"]
+              }
+            }
+          }
+        },
+        "x-codeSamples": [
+          {
+            "lang": "js",
+            "source": "import { createKiloClient } from \"@kilocode/sdk\n\nconst client = createKiloClient()\nawait client.kilo.audio.transcriptions({\n  ...\n})"
+          }
+        ]
+      }
+    },
     "/kilo/notifications": {
       "get": {
         "operationId": "kilo.notifications",