@steipete/summarize-core 0.11.1 → 0.12.0

package/dist/esm/transcription/whisper/core.js

@@ -2,459 +2,385 @@ import { randomUUID } from "node:crypto";
 import { promises as fs } from "node:fs";
 import { tmpdir } from "node:os";
 import { basename, join } from "node:path";
-import { resolvePreferredOnnxModel, transcribeWithOnnxCli, transcribeWithOnnxCliFile, } from "../onnx-cli.js";
+import { transcribeWithOnnxCli, transcribeWithOnnxCliFile } from "../onnx-cli.js";
+import { transcribeChunkedFile } from "./chunking.js";
 import { DEFAULT_SEGMENT_SECONDS, MAX_OPENAI_UPLOAD_BYTES } from "./constants.js";
-import { transcribeWithFal } from "./fal.js";
-import { isFfmpegAvailable, runFfmpegSegment, transcodeBytesToMp3 } from "./ffmpeg.js";
+import { isFfmpegAvailable, transcodeBytesToMp3 } from "./ffmpeg.js";
 import { shouldRetryGroqViaFfmpeg, transcribeWithGroq } from "./groq.js";
-import { shouldRetryOpenAiViaFfmpeg, transcribeWithOpenAi } from "./openai.js";
-import { ensureWhisperFilenameExtension, formatBytes, readFirstBytes, wrapError } from "./utils.js";
+import { resolveOnnxModelPreference } from "./preferences.js";
+import { transcribeBytesWithRemoteFallbacks, transcribeFileWithRemoteFallbacks, transcribeOversizedBytesViaTempFile, } from "./remote.js";
+import { ensureWhisperFilenameExtension, formatBytes, wrapError } from "./utils.js";
 import { isWhisperCppReady, transcribeWithWhisperCppFile } from "./whisper-cpp.js";
-function resolveTranscriberPreference(env) {
-    const raw = env.SUMMARIZE_TRANSCRIBER?.trim().toLowerCase();
-    if (raw === "auto" || raw === "whisper" || raw === "parakeet" || raw === "canary")
-        return raw;
-    return "auto";
-}
-function resolveOnnxModelPreference(env) {
-    const preference = resolveTranscriberPreference(env);
-    if (preference === "parakeet" || preference === "canary")
-        return preference;
-    if (preference === "auto")
-        return resolvePreferredOnnxModel(env);
-    return null;
-}
-export async function transcribeMediaWithWhisper({ bytes, mediaType, filename, groqApiKey, skipGroq = false, openaiApiKey, falApiKey, totalDurationSeconds = null, onProgress, env = process.env, }) {
+export async function transcribeMediaWithWhisper({ bytes, mediaType, filename, groqApiKey, skipGroq = false, assemblyaiApiKey = null, geminiApiKey = null, openaiApiKey, falApiKey, totalDurationSeconds = null, onProgress, env = process.env, }) {
     const notes = [];
-    // 1. Groq (cloud, free, fastest)
     let groqError = null;
     if (groqApiKey && !skipGroq) {
-        try {
-            const text = await transcribeWithGroq(bytes, mediaType, filename, groqApiKey);
-            if (text) {
-                return { text, provider: "groq", error: null, notes };
-            }
-            groqError = new Error("Groq transcription returned empty text");
-        }
-        catch (error) {
-            groqError = wrapError("Groq transcription failed", error);
-        }
-    }
-    if (!skipGroq && groqApiKey && groqError && shouldRetryGroqViaFfmpeg(groqError)) {
-        const canTranscode = await isFfmpegAvailable();
-        if (canTranscode) {
-            try {
-                notes.push("Groq could not decode media; transcoding via ffmpeg and retrying");
-                const mp3Bytes = await transcodeBytesToMp3(bytes);
-                const retried = await transcribeWithGroq(mp3Bytes, "audio/mpeg", "audio.mp3", groqApiKey);
-                if (retried) {
-                    return { text: retried, provider: "groq", error: null, notes };
-                }
-                groqError = new Error("Groq transcription returned empty text after ffmpeg transcode");
-                bytes = mp3Bytes;
-                mediaType = "audio/mpeg";
-                filename = "audio.mp3";
-            }
-            catch (error) {
-                notes.push(`ffmpeg transcode failed; cannot retry Groq decode error: ${error instanceof Error ? error.message : String(error)}`);
-            }
-        }
-        else {
-            notes.push("Groq could not decode media; install ffmpeg to enable transcoding retry");
+        const groqResult = await transcribeWithGroqFirst({
+            bytes,
+            mediaType,
+            filename,
+            groqApiKey,
+            notes,
+        });
+        bytes = groqResult.bytes;
+        mediaType = groqResult.mediaType;
+        filename = groqResult.filename;
+        if (groqResult.text) {
+            return { text: groqResult.text, provider: "groq", error: null, notes };
         }
+        groqError = groqResult.error;
     }
     if (groqError) {
-        notes.push(`Groq transcription failed; falling back to local/OpenAI: ${groqError.message}`);
+        notes.push(`Groq transcription failed; falling back to local/AssemblyAI/Gemini/OpenAI: ${groqError.message}`);
     }
-    // 2. ONNX (local)
-    const onnxPreference = resolveOnnxModelPreference(env);
-    if (onnxPreference) {
-        const onnx = await transcribeWithOnnxCli({
-            model: onnxPreference,
+    const onnx = await transcribeWithLocalOnnx({
+        bytes,
+        mediaType,
+        filename,
+        totalDurationSeconds,
+        onProgress,
+        env,
+        notes,
+    });
+    if (onnx)
+        return onnx;
+    const local = await transcribeWithLocalWhisperBytes({
+        bytes,
+        mediaType,
+        filename,
+        totalDurationSeconds,
+        onProgress,
+        notes,
+    });
+    if (local)
+        return local;
+    return await transcribeBytesWithRemoteFallbacks({
+        bytes,
+        mediaType,
+        filename,
+        notes,
+        groqApiKey,
+        groqError,
+        assemblyaiApiKey,
+        geminiApiKey,
+        openaiApiKey,
+        falApiKey,
+        env,
+        onProgress,
+        transcribeOversizedBytesWithChunking: ({ bytes, mediaType, filename, onProgress }) => transcribeOversizedBytesViaTempFile({
             bytes,
             mediaType,
             filename,
+            onProgress,
+            transcribeFile: ({ filePath, mediaType, filename, onProgress }) => transcribeMediaFileWithWhisper({
+                filePath,
+                mediaType,
+                filename,
+                groqApiKey,
+                assemblyaiApiKey,
+                geminiApiKey,
+                openaiApiKey,
+                falApiKey,
+                segmentSeconds: DEFAULT_SEGMENT_SECONDS,
+                onProgress,
+                env,
+            }),
+        }),
+    });
+}
+export async function transcribeMediaFileWithWhisper({ filePath, mediaType, filename, groqApiKey, assemblyaiApiKey = null, geminiApiKey = null, openaiApiKey, falApiKey, segmentSeconds = DEFAULT_SEGMENT_SECONDS, totalDurationSeconds = null, onProgress = null, env = process.env, }) {
+    const notes = [];
+    let skipGroqInNestedCalls = false;
+    let groqError = null;
+    if (groqApiKey) {
+        skipGroqInNestedCalls = true;
+        const groqResult = await transcribeGroqFileFirst({
+            filePath,
+            mediaType,
+            filename,
+            groqApiKey,
+            assemblyaiApiKey,
+            geminiApiKey,
+            openaiApiKey,
+            falApiKey,
+            segmentSeconds,
             totalDurationSeconds,
             onProgress,
             env,
-        });
-        if (onnx.text) {
-            if (onnx.notes.length > 0)
-                notes.push(...onnx.notes);
-            return { ...onnx, notes };
-        }
-        if (onnx.notes.length > 0)
-            notes.push(...onnx.notes);
-        if (onnx.error) {
-            notes.push(`${onnx.provider ?? "onnx"} failed; falling back to Whisper: ${onnx.error.message}`);
-        }
-    }
-    // 3. whisper.cpp (local)
-    const localReady = await isWhisperCppReady();
-    let local = null;
-    if (localReady) {
-        const nameHint = filename?.trim() ? basename(filename.trim()) : "media";
-        const tempFile = join(tmpdir(), `summarize-whisper-local-${randomUUID()}-${ensureWhisperFilenameExtension(nameHint, mediaType)}`);
-        try {
-            await fs.writeFile(tempFile, bytes);
-            try {
-                local = await transcribeWithWhisperCppFile({
-                    filePath: tempFile,
-                    mediaType,
-                    totalDurationSeconds,
-                    onProgress,
-                });
-            }
-            catch (error) {
-                local = {
-                    text: null,
-                    provider: "whisper.cpp",
-                    error: wrapError("whisper.cpp failed", error),
-                    notes: [],
-                };
-            }
-            if (local.text) {
-                if (local.notes.length > 0)
-                    notes.push(...local.notes);
-                return { ...local, notes };
-            }
-            if (local.notes.length > 0)
-                notes.push(...local.notes);
-            if (local.error) {
-                notes.push(`whisper.cpp failed; falling back to remote Whisper: ${local.error.message}`);
-            }
-        }
-        finally {
-            await fs.unlink(tempFile).catch(() => { });
-        }
-    }
-    // 4. OpenAI / FAL (cloud fallbacks)
-    if (!groqApiKey && !openaiApiKey && !falApiKey) {
-        return {
-            text: null,
-            provider: null,
-            error: new Error("No transcription providers available (install whisper-cpp or set GROQ_API_KEY, OPENAI_API_KEY, or FAL_KEY)"),
             notes,
-        };
+        });
+        if (groqResult.text)
+            return groqResult;
+        groqError = groqResult.error;
     }
-    if (openaiApiKey && bytes.byteLength > MAX_OPENAI_UPLOAD_BYTES) {
-        const canChunk = await isFfmpegAvailable();
-        if (canChunk) {
-            const tempFile = join(tmpdir(), `summarize-whisper-${randomUUID()}`);
-            try {
-                await fs.writeFile(tempFile, bytes);
-                const chunked = await transcribeMediaFileWithWhisper({
-                    filePath: tempFile,
-                    mediaType,
-                    filename,
-                    groqApiKey,
-                    openaiApiKey,
-                    falApiKey,
-                    segmentSeconds: DEFAULT_SEGMENT_SECONDS,
-                    onProgress,
-                    env,
-                });
-                return chunked;
-            }
-            finally {
-                await fs.unlink(tempFile).catch(() => { });
-            }
-        }
-        notes.push(`Media too large for Whisper upload (${formatBytes(bytes.byteLength)}); transcribing first ${formatBytes(MAX_OPENAI_UPLOAD_BYTES)} only (install ffmpeg for full transcription)`);
-        bytes = bytes.slice(0, MAX_OPENAI_UPLOAD_BYTES);
+    const onnx = await transcribeWithLocalOnnxFile({
+        filePath,
+        mediaType,
+        totalDurationSeconds,
+        onProgress,
+        env,
+        notes,
+    });
+    if (onnx)
+        return onnx;
+    const local = await transcribeWithLocalWhisperFile({
+        filePath,
+        mediaType,
+        totalDurationSeconds,
+        onProgress,
+        notes,
+    });
+    if (local)
+        return local;
+    return await transcribeFileWithRemoteFallbacks({
+        filePath,
+        mediaType,
+        filename,
+        notes,
+        groqApiKey,
+        groqError,
+        assemblyaiApiKey,
+        geminiApiKey,
+        openaiApiKey,
+        falApiKey,
+        env,
+        totalDurationSeconds,
+        onProgress,
+        transcribeChunkedFile: ({ filePath, segmentSeconds, totalDurationSeconds, onProgress }) => transcribeChunkedFile({
+            filePath,
+            segmentSeconds,
+            totalDurationSeconds,
+            onProgress,
+            transcribeSegment: ({ bytes, filename }) => transcribeMediaWithWhisper({
+                bytes,
+                mediaType: "audio/mpeg",
+                filename,
+                groqApiKey,
+                skipGroq: skipGroqInNestedCalls,
+                assemblyaiApiKey,
+                geminiApiKey,
+                openaiApiKey,
+                falApiKey,
+                env,
+            }),
+        }),
+    });
+}
+async function transcribeWithGroqFirst({ bytes, mediaType, filename, groqApiKey, notes, }) {
+    let groqError = null;
+    try {
+        const text = await transcribeWithGroq(bytes, mediaType, filename, groqApiKey);
+        if (text)
+            return { text, error: null, bytes, mediaType, filename };
+        groqError = new Error("Groq transcription returned empty text");
     }
-    let openaiError = null;
-    if (openaiApiKey) {
-        try {
-            const text = await transcribeWithOpenAi(bytes, mediaType, filename, openaiApiKey, { env });
-            if (text) {
-                return { text, provider: "openai", error: null, notes };
-            }
-            openaiError = new Error("OpenAI transcription returned empty text");
-        }
-        catch (error) {
-            openaiError = wrapError("OpenAI transcription failed", error);
-        }
+    catch (error) {
+        groqError = wrapError("Groq transcription failed", error);
     }
-    if (openaiApiKey && openaiError && shouldRetryOpenAiViaFfmpeg(openaiError)) {
+    if (groqError && shouldRetryGroqViaFfmpeg(groqError)) {
         const canTranscode = await isFfmpegAvailable();
         if (canTranscode) {
             try {
-                // Some providers hand out containers/codecs Whisper rejects. Transcoding to a small mono MP3
-                // is the most reliable cross-format fallback (and also reduces upload size).
-                notes.push("OpenAI could not decode media; transcoding via ffmpeg and retrying");
+                notes.push("Groq could not decode media; transcoding via ffmpeg and retrying");
                 const mp3Bytes = await transcodeBytesToMp3(bytes);
-                const retried = await transcribeWithOpenAi(mp3Bytes, "audio/mpeg", "audio.mp3", openaiApiKey, { env });
+                const retried = await transcribeWithGroq(mp3Bytes, "audio/mpeg", "audio.mp3", groqApiKey);
                 if (retried) {
-                    return { text: retried, provider: "openai", error: null, notes };
+                    return {
+                        text: retried,
+                        error: null,
+                        bytes: mp3Bytes,
+                        mediaType: "audio/mpeg",
+                        filename: "audio.mp3",
+                    };
                 }
-                openaiError = new Error("OpenAI transcription returned empty text after ffmpeg transcode");
+                groqError = new Error("Groq transcription returned empty text after ffmpeg transcode");
                 bytes = mp3Bytes;
                 mediaType = "audio/mpeg";
                 filename = "audio.mp3";
             }
             catch (error) {
-                notes.push(`ffmpeg transcode failed; cannot retry OpenAI decode error: ${error instanceof Error ? error.message : String(error)}`);
+                notes.push(`ffmpeg transcode failed; cannot retry Groq decode error: ${error instanceof Error ? error.message : String(error)}`);
             }
         }
         else {
-            notes.push("OpenAI could not decode media; install ffmpeg to enable transcoding retry");
+            notes.push("Groq could not decode media; install ffmpeg to enable transcoding retry");
         }
     }
-    const canUseFal = Boolean(falApiKey) && mediaType.toLowerCase().startsWith("audio/");
-    if (openaiError && canUseFal) {
-        notes.push(`OpenAI transcription failed; falling back to FAL: ${openaiError.message}`);
-    }
-    if (falApiKey && !canUseFal) {
-        notes.push(`Skipping FAL transcription: unsupported mediaType ${mediaType}`);
-    }
-    if (falApiKey && canUseFal) {
+    return { text: null, error: groqError, bytes, mediaType, filename };
+}
+async function transcribeGroqFileFirst({ filePath, mediaType, filename, groqApiKey, assemblyaiApiKey, geminiApiKey, openaiApiKey, falApiKey, segmentSeconds, totalDurationSeconds, onProgress, env, notes, }) {
+    const stat = await fs.stat(filePath);
+    if (stat.size <= MAX_OPENAI_UPLOAD_BYTES) {
+        const fileBytes = new Uint8Array(await fs.readFile(filePath));
         try {
-            const text = await transcribeWithFal(bytes, mediaType, falApiKey);
-            if (text) {
-                return { text, provider: "fal", error: null, notes };
-            }
-            return {
-                text: null,
-                provider: "fal",
-                error: new Error("FAL transcription returned empty text"),
-                notes,
-            };
+            const text = await transcribeWithGroq(fileBytes, mediaType, filename, groqApiKey);
+            if (text)
+                return { text, provider: "groq", error: null, notes };
+            const error = new Error("Groq transcription returned empty text");
+            notes.push("Groq transcription returned empty text; falling back to local/AssemblyAI/Gemini/OpenAI");
+            return { text: null, provider: "groq", error, notes };
         }
         catch (error) {
-            return {
-                text: null,
-                provider: "fal",
-                error: wrapError("FAL transcription failed", error),
-                notes,
-            };
+            const wrapped = wrapError("Groq transcription failed", error);
+            notes.push(`Groq transcription failed; falling back to local/AssemblyAI/Gemini/OpenAI: ${error instanceof Error ? error.message : String(error)}`);
+            return { text: null, provider: "groq", error: wrapped, notes };
         }
     }
-    const terminalError = openaiError ?? groqError ?? new Error("No transcription providers available");
-    const terminalProvider = openaiError
-        ? "openai"
-        : groqError
-            ? "groq"
-            : openaiApiKey
-                ? "openai"
-                : null;
-    return { text: null, provider: terminalProvider, error: terminalError, notes };
+    const canChunk = await isFfmpegAvailable();
+    if (!canChunk) {
+        const error = new Error(`File too large for Groq upload (${formatBytes(stat.size)}); trying local providers`);
+        notes.push(error.message);
+        return { text: null, provider: "groq", error, notes };
+    }
+    const chunked = await transcribeChunkedFile({
+        filePath,
+        segmentSeconds,
+        totalDurationSeconds,
+        onProgress,
+        transcribeSegment: ({ bytes, filename }) => transcribeMediaWithWhisper({
+            bytes,
+            mediaType: "audio/mpeg",
+            filename,
+            groqApiKey,
+            assemblyaiApiKey,
+            geminiApiKey,
+            openaiApiKey,
+            falApiKey,
+            env,
+        }),
+    });
+    if (chunked.notes.length > 0)
+        notes.push(...chunked.notes);
+    if (chunked.text)
+        return { ...chunked, notes };
+    const error = chunked.error ?? new Error("Groq chunked transcription failed");
+    notes.push(`Groq chunked transcription failed; falling back to local/AssemblyAI/Gemini/OpenAI: ${error.message}`);
+    return { text: null, provider: "groq", error, notes };
 }
-export async function transcribeMediaFileWithWhisper({ filePath, mediaType, filename, groqApiKey, openaiApiKey, falApiKey, segmentSeconds = DEFAULT_SEGMENT_SECONDS, totalDurationSeconds = null, onProgress = null, env = process.env, }) {
-    const notes = [];
-    const stat = await fs.stat(filePath);
-    let skipGroqInNestedCalls = false;
-    let groqError = null;
-    // 1. Groq (cloud, free, fastest) — try first even for file-based transcription
-    if (groqApiKey) {
-        skipGroqInNestedCalls = true;
-        if (stat.size <= MAX_OPENAI_UPLOAD_BYTES) {
-            const fileBytes = new Uint8Array(await fs.readFile(filePath));
-            try {
-                const text = await transcribeWithGroq(fileBytes, mediaType, filename, groqApiKey);
-                if (text) {
-                    return { text, provider: "groq", error: null, notes };
-                }
-                groqError = new Error("Groq transcription returned empty text");
-                notes.push("Groq transcription returned empty text; falling back to local/OpenAI");
-            }
-            catch (error) {
-                groqError = wrapError("Groq transcription failed", error);
-                notes.push(`Groq transcription failed; falling back to local/OpenAI: ${error instanceof Error ? error.message : String(error)}`);
-            }
-        }
-        else {
-            groqError = new Error(`File too large for Groq upload (${formatBytes(stat.size)}); trying local providers`);
-            notes.push(groqError.message);
-        }
+async function transcribeWithLocalOnnx({ bytes, mediaType, filename, totalDurationSeconds, onProgress, env, notes, }) {
+    const onnxPreference = resolveOnnxModelPreference(env);
+    if (!onnxPreference)
+        return null;
+    const onnx = await transcribeWithOnnxCli({
+        model: onnxPreference,
+        bytes,
+        mediaType,
+        filename,
+        totalDurationSeconds,
+        onProgress,
+        env,
+    });
+    if (onnx.text) {
+        if (onnx.notes.length > 0)
+            notes.push(...onnx.notes);
+        return { ...onnx, notes };
     }
-    // 2. ONNX (local)
+    if (onnx.notes.length > 0)
+        notes.push(...onnx.notes);
+    if (onnx.error) {
+        notes.push(`${onnx.provider ?? "onnx"} failed; falling back to Whisper: ${onnx.error.message}`);
+    }
+    return null;
+}
+async function transcribeWithLocalOnnxFile({ filePath, mediaType, totalDurationSeconds, onProgress, env, notes, }) {
     const onnxPreference = resolveOnnxModelPreference(env);
-    if (onnxPreference) {
-        onProgress?.({
-            partIndex: null,
-            parts: null,
-            processedDurationSeconds: null,
-            totalDurationSeconds,
-        });
-        const onnx = await transcribeWithOnnxCliFile({
-            model: onnxPreference,
-            filePath,
-            mediaType,
-            totalDurationSeconds,
-            onProgress,
-            env,
-        });
-        if (onnx.text) {
-            if (onnx.notes.length > 0)
-                notes.push(...onnx.notes);
-            return { ...onnx, notes };
-        }
+    if (!onnxPreference)
+        return null;
+    onProgress?.({
+        partIndex: null,
+        parts: null,
+        processedDurationSeconds: null,
+        totalDurationSeconds,
+    });
+    const onnx = await transcribeWithOnnxCliFile({
+        model: onnxPreference,
+        filePath,
+        mediaType,
+        totalDurationSeconds,
+        onProgress,
+        env,
+    });
+    if (onnx.text) {
         if (onnx.notes.length > 0)
             notes.push(...onnx.notes);
-        if (onnx.error) {
-            notes.push(`${onnx.provider ?? "onnx"} failed; falling back to Whisper: ${onnx.error.message}`);
-        }
+        return { ...onnx, notes };
     }
-    // 3. whisper.cpp (local)
+    if (onnx.notes.length > 0)
+        notes.push(...onnx.notes);
+    if (onnx.error) {
+        notes.push(`${onnx.provider ?? "onnx"} failed; falling back to Whisper: ${onnx.error.message}`);
+    }
+    return null;
+}
+async function transcribeWithLocalWhisperBytes({ bytes, mediaType, filename, totalDurationSeconds, onProgress, notes, }) {
     const localReady = await isWhisperCppReady();
-    let local = null;
-    if (localReady) {
-        onProgress?.({
-            partIndex: null,
-            parts: null,
-            processedDurationSeconds: null,
+    if (!localReady)
+        return null;
+    const nameHint = filename?.trim() ? basename(filename.trim()) : "media";
+    const tempFile = join(tmpdir(), `summarize-whisper-local-${randomUUID()}-${ensureWhisperFilenameExtension(nameHint, mediaType)}`);
+    try {
+        await fs.writeFile(tempFile, bytes);
+        const result = await safeTranscribeWithWhisperCppFile({
+            filePath: tempFile,
+            mediaType,
             totalDurationSeconds,
+            onProgress,
         });
-        try {
-            local = await transcribeWithWhisperCppFile({
-                filePath,
-                mediaType,
-                totalDurationSeconds,
-                onProgress,
-            });
+        if (result.text) {
+            if (result.notes.length > 0)
+                notes.push(...result.notes);
+            return { ...result, notes };
         }
-        catch (error) {
-            local = {
-                text: null,
-                provider: "whisper.cpp",
-                error: wrapError("whisper.cpp failed", error),
-                notes: [],
-            };
-        }
-        if (local.text) {
-            if (local.notes.length > 0)
-                notes.push(...local.notes);
-            return { ...local, notes };
-        }
-        if (local.notes.length > 0)
-            notes.push(...local.notes);
-        if (local.error) {
-            notes.push(`whisper.cpp failed; falling back to remote Whisper: ${local.error.message}`);
+        if (result.notes.length > 0)
+            notes.push(...result.notes);
+        if (result.error) {
+            notes.push(`whisper.cpp failed; falling back to remote Whisper: ${result.error.message}`);
         }
+        return null;
     }
-    // 4. OpenAI / FAL (cloud fallbacks)
-    if (!openaiApiKey && !falApiKey) {
-        if (groqError) {
-            return {
-                text: null,
-                provider: "groq",
-                error: groqError,
-                notes,
-            };
-        }
-        return {
-            text: null,
-            provider: null,
-            error: new Error("No transcription providers available (install whisper-cpp or set GROQ_API_KEY, OPENAI_API_KEY, or FAL_KEY)"),
-            notes,
-        };
-    }
-    if (openaiApiKey && stat.size > MAX_OPENAI_UPLOAD_BYTES) {
-        const canChunk = await isFfmpegAvailable();
-        if (!canChunk) {
-            notes.push(`Media too large for Whisper upload (${formatBytes(stat.size)}); install ffmpeg to enable chunked transcription`);
-            const head = await readFirstBytes(filePath, MAX_OPENAI_UPLOAD_BYTES);
-            const partial = await transcribeMediaWithWhisper({
-                bytes: head,
-                mediaType,
-                filename,
-                groqApiKey,
-                skipGroq: skipGroqInNestedCalls,
-                openaiApiKey,
-                falApiKey,
-                env,
-            });
-            if (partial.notes.length > 0)
-                notes.push(...partial.notes);
-            return { ...partial, notes };
-        }
-        const dir = await fs.mkdtemp(join(tmpdir(), "summarize-whisper-segments-"));
-        try {
-            const pattern = join(dir, "part-%03d.mp3");
-            await runFfmpegSegment({
-                inputPath: filePath,
-                outputPattern: pattern,
-                segmentSeconds,
-            });
-            const files = (await fs.readdir(dir))
-                .filter((name) => name.startsWith("part-") && name.endsWith(".mp3"))
-                .sort((a, b) => a.localeCompare(b));
-            if (files.length === 0) {
-                return {
-                    text: null,
-                    provider: null,
-                    error: new Error("ffmpeg produced no audio segments"),
-                    notes,
-                };
-            }
-            notes.push(`ffmpeg chunked media into ${files.length} parts (${segmentSeconds}s each)`);
-            onProgress?.({
-                partIndex: null,
-                parts: files.length,
-                processedDurationSeconds: null,
-                totalDurationSeconds,
-            });
-            const parts = [];
-            let usedProvider = null;
-            for (const [index, name] of files.entries()) {
-                const segmentPath = join(dir, name);
-                const segmentBytes = new Uint8Array(await fs.readFile(segmentPath));
-                const result = await transcribeMediaWithWhisper({
-                    bytes: segmentBytes,
-                    mediaType: "audio/mpeg",
-                    filename: name,
-                    groqApiKey,
-                    skipGroq: skipGroqInNestedCalls,
-                    openaiApiKey,
-                    falApiKey,
-                    onProgress: null,
-                    env,
-                });
-                if (!usedProvider && result.provider)
-                    usedProvider = result.provider;
-                if (result.error && !result.text) {
-                    return { text: null, provider: usedProvider, error: result.error, notes };
-                }
-                if (result.text)
-                    parts.push(result.text);
-                // Coarse but useful: update based on part boundaries. Duration is best-effort (RSS hints or
-                // ffprobe); the per-part time is stable enough to make the spinner feel alive.
-                const processedSeconds = Math.max(0, (index + 1) * segmentSeconds);
-                onProgress?.({
-                    partIndex: index + 1,
-                    parts: files.length,
-                    processedDurationSeconds: typeof totalDurationSeconds === "number" && totalDurationSeconds > 0
-                        ? Math.min(processedSeconds, totalDurationSeconds)
-                        : null,
-                    totalDurationSeconds,
-                });
-            }
-            return { text: parts.join("\n\n"), provider: usedProvider, error: null, notes };
-        }
-        finally {
-            await fs.rm(dir, { recursive: true, force: true }).catch(() => { });
-        }
+    finally {
+        await fs.unlink(tempFile).catch(() => { });
     }
-    const bytes = new Uint8Array(await fs.readFile(filePath));
+}
+async function transcribeWithLocalWhisperFile({ filePath, mediaType, totalDurationSeconds, onProgress, notes, }) {
+    const localReady = await isWhisperCppReady();
+    if (!localReady)
+        return null;
     onProgress?.({
         partIndex: null,
         parts: null,
         processedDurationSeconds: null,
         totalDurationSeconds,
     });
-    const result = await transcribeMediaWithWhisper({
-        bytes,
+    const result = await safeTranscribeWithWhisperCppFile({
+        filePath,
         mediaType,
-        filename,
-        groqApiKey,
-        skipGroq: skipGroqInNestedCalls,
-        openaiApiKey,
-        falApiKey,
-        env,
+        totalDurationSeconds,
+        onProgress,
     });
+    if (result.text) {
+        if (result.notes.length > 0)
+            notes.push(...result.notes);
+        return { ...result, notes };
+    }
     if (result.notes.length > 0)
         notes.push(...result.notes);
-    return { ...result, notes };
+    if (result.error) {
+        notes.push(`whisper.cpp failed; falling back to remote Whisper: ${result.error.message}`);
+    }
+    return null;
+}
+async function safeTranscribeWithWhisperCppFile(args) {
+    try {
+        return await transcribeWithWhisperCppFile(args);
+    }
+    catch (error) {
+        return {
+            text: null,
+            provider: "whisper.cpp",
+            error: wrapError("whisper.cpp failed", error),
+            notes: [],
+        };
+    }
 }
 //# sourceMappingURL=core.js.map