@automagik/omni 2.260530.2 → 2.260530.3

package/dist/index.js

@@ -124950,7 +124950,7 @@ import { fileURLToPath } from "url";
 // package.json
 var package_default = {
   name: "@automagik/omni",
-  version: "2.260530.2",
+  version: "2.260530.3",
   description: "LLM-optimized CLI for Omni",
   type: "module",
   bin: {

package/dist/server/index.js

@@ -230516,7 +230516,7 @@ var init_sentry_scrub = __esm(() => {
 var require_package8 = __commonJS((exports, module) => {
   module.exports = {
     name: "@omni/api",
-    version: "2.260530.2",
+    version: "2.260530.3",
     type: "module",
     exports: {
       ".": {
@@ -292374,7 +292374,7 @@ var require_bignumber = __commonJS((exports, module) => {
           return arr.reverse();
         }
         return function(str, baseIn, baseOut, sign2, callerIsToString) {
-          var alphabet, d2, e, k2, r, x, xc, y2, i = str.indexOf("."), dp = DECIMAL_PLACES, rm = ROUNDING_MODE;
+          var alphabet, d2, e, k2, r, x, xc, y2, i = str.indexOf("."), dp = DECIMAL_PLACES, rm2 = ROUNDING_MODE;
           if (i >= 0) {
             k2 = POW_PRECISION;
             POW_PRECISION = 0;
@@ -292397,7 +292397,7 @@ var require_bignumber = __commonJS((exports, module) => {
             x.c = xc;
             x.e = e;
             x.s = sign2;
-            x = div(x, y2, dp, rm, baseOut);
+            x = div(x, y2, dp, rm2, baseOut);
             xc = x.c;
             r = x.r;
             e = x.e;
@@ -292406,7 +292406,7 @@ var require_bignumber = __commonJS((exports, module) => {
           i = xc[d2];
           k2 = baseOut / 2;
           r = r || d2 < 0 || xc[d2 + 1] != null;
-          r = rm < 4 ? (i != null || r) && (rm == 0 || rm == (x.s < 0 ? 3 : 2)) : i > k2 || i == k2 && (rm == 4 || r || rm == 6 && xc[d2 - 1] & 1 || rm == (x.s < 0 ? 8 : 7));
+          r = rm2 < 4 ? (i != null || r) && (rm2 == 0 || rm2 == (x.s < 0 ? 3 : 2)) : i > k2 || i == k2 && (rm2 == 4 || r || rm2 == 6 && xc[d2 - 1] & 1 || rm2 == (x.s < 0 ? 8 : 7));
           if (d2 < 1 || !xc[0]) {
             str = r ? toFixedPoint(alphabet.charAt(1), -dp, alphabet.charAt(0)) : alphabet.charAt(0);
           } else {
@@ -292468,7 +292468,7 @@ var require_bignumber = __commonJS((exports, module) => {
           for (;!a[0] && a.length > 1; a.splice(0, 1))
             ;
         }
-        return function(x, y2, dp, rm, base2) {
+        return function(x, y2, dp, rm2, base2) {
           var cmp, e, i, more, n2, prod, prodL, q, qc, rem, remL, rem0, xi, xL, yc0, yL, yz, s = x.s == y2.s ? 1 : -1, xc = x.c, yc = y2.c;
           if (!xc || !xc[0] || !yc || !yc[0]) {
             return new BigNumber2(!x.s || !y2.s || (xc ? yc && xc[0] == yc[0] : !yc) ? NaN : xc && xc[0] == 0 || !yc ? s * 0 : s / 0);
@@ -292568,7 +292568,7 @@ var require_bignumber = __commonJS((exports, module) => {
           if (base2 == BASE) {
             for (i = 1, s = qc[0];s >= 10; s /= 10, i++)
               ;
-            round(q, dp + (q.e = i + e * LOG_BASE - 1) + 1, rm, more);
+            round(q, dp + (q.e = i + e * LOG_BASE - 1) + 1, rm2, more);
           } else {
             q.e = e;
             q.r = +more;
@@ -292576,12 +292576,12 @@ var require_bignumber = __commonJS((exports, module) => {
           return q;
         };
       }();
-      function format(n2, i, rm, id) {
+      function format(n2, i, rm2, id) {
         var c0, e, ne2, len, str;
-        if (rm == null)
-          rm = ROUNDING_MODE;
+        if (rm2 == null)
+          rm2 = ROUNDING_MODE;
         else
-          intCheck(rm, 0, 8);
+          intCheck(rm2, 0, 8);
         if (!n2.c)
           return n2.toString();
         c0 = n2.c[0];
@@ -292590,7 +292590,7 @@ var require_bignumber = __commonJS((exports, module) => {
           str = coeffToString(n2.c);
           str = id == 1 || id == 2 && (ne2 <= TO_EXP_NEG || ne2 >= TO_EXP_POS) ? toExponential(str, ne2) : toFixedPoint(str, ne2, "0");
         } else {
-          n2 = round(new BigNumber2(n2), i, rm);
+          n2 = round(new BigNumber2(n2), i, rm2);
           e = n2.e;
           str = coeffToString(n2.c);
           len = str.length;
@@ -292671,7 +292671,7 @@ var require_bignumber = __commonJS((exports, module) => {
           x.c = x.e = null;
         };
       }();
-      function round(x, sd, rm, r) {
+      function round(x, sd, rm2, r) {
         var d2, i, j2, k2, n2, ni, rd, xc = x.c, pows10 = POWS_TEN;
         if (xc) {
           out: {
@@ -292706,7 +292706,7 @@ var require_bignumber = __commonJS((exports, module) => {
               }
             }
             r = r || sd < 0 || xc[ni + 1] != null || (j2 < 0 ? n2 : n2 % pows10[d2 - j2 - 1]);
-            r = rm < 4 ? (rd || r) && (rm == 0 || rm == (x.s < 0 ? 3 : 2)) : rd > 5 || rd == 5 && (rm == 4 || r || rm == 6 && (i > 0 ? j2 > 0 ? n2 / pows10[d2 - j2] : 0 : xc[ni - 1]) % 10 & 1 || rm == (x.s < 0 ? 8 : 7));
+            r = rm2 < 4 ? (rd || r) && (rm2 == 0 || rm2 == (x.s < 0 ? 3 : 2)) : rd > 5 || rd == 5 && (rm2 == 4 || r || rm2 == 6 && (i > 0 ? j2 > 0 ? n2 / pows10[d2 - j2] : 0 : xc[ni - 1]) % 10 & 1 || rm2 == (x.s < 0 ? 8 : 7));
             if (sd < 1 || !xc[0]) {
               xc.length = 0;
               if (r) {
@@ -292778,15 +292778,15 @@ var require_bignumber = __commonJS((exports, module) => {
       P4.comparedTo = function(y2, b3) {
         return compare2(this, new BigNumber2(y2, b3));
       };
-      P4.decimalPlaces = P4.dp = function(dp, rm) {
+      P4.decimalPlaces = P4.dp = function(dp, rm2) {
         var c, n2, v2, x = this;
         if (dp != null) {
           intCheck(dp, 0, MAX);
-          if (rm == null)
-            rm = ROUNDING_MODE;
+          if (rm2 == null)
+            rm2 = ROUNDING_MODE;
           else
-            intCheck(rm, 0, 8);
-          return round(new BigNumber2(x), dp + x.e + 1, rm);
+            intCheck(rm2, 0, 8);
+          return round(new BigNumber2(x), dp + x.e + 1, rm2);
         }
         if (!(c = x.c))
           return null;
@@ -292885,13 +292885,13 @@ var require_bignumber = __commonJS((exports, module) => {
           y2 = ONE2.div(y2);
         return m2 ? y2.mod(m2) : k2 ? round(y2, POW_PRECISION, ROUNDING_MODE, more) : y2;
       };
-      P4.integerValue = function(rm) {
+      P4.integerValue = function(rm2) {
         var n2 = new BigNumber2(this);
-        if (rm == null)
-          rm = ROUNDING_MODE;
+        if (rm2 == null)
+          rm2 = ROUNDING_MODE;
         else
-          intCheck(rm, 0, 8);
-        return round(n2, n2.e + 1, rm);
+          intCheck(rm2, 0, 8);
+        return round(n2, n2.e + 1, rm2);
       };
       P4.isEqualTo = P4.eq = function(y2, b3) {
         return compare2(this, new BigNumber2(y2, b3)) === 0;
@@ -293128,15 +293128,15 @@ var require_bignumber = __commonJS((exports, module) => {
         }
         return normalise(y2, xc, ye);
       };
-      P4.precision = P4.sd = function(sd, rm) {
+      P4.precision = P4.sd = function(sd, rm2) {
         var c, n2, v2, x = this;
         if (sd != null && sd !== !!sd) {
           intCheck(sd, 1, MAX);
-          if (rm == null)
-            rm = ROUNDING_MODE;
+          if (rm2 == null)
+            rm2 = ROUNDING_MODE;
           else
-            intCheck(rm, 0, 8);
-          return round(new BigNumber2(x), sd, rm);
+            intCheck(rm2, 0, 8);
+          return round(new BigNumber2(x), sd, rm2);
         }
         if (!(c = x.c))
           return null;
@@ -293213,36 +293213,36 @@ var require_bignumber = __commonJS((exports, module) => {
         }
         return round(r, r.e + DECIMAL_PLACES + 1, ROUNDING_MODE, m2);
       };
-      P4.toExponential = function(dp, rm) {
+      P4.toExponential = function(dp, rm2) {
         if (dp != null) {
           intCheck(dp, 0, MAX);
           dp++;
         }
-        return format(this, dp, rm, 1);
+        return format(this, dp, rm2, 1);
       };
-      P4.toFixed = function(dp, rm) {
+      P4.toFixed = function(dp, rm2) {
         if (dp != null) {
           intCheck(dp, 0, MAX);
           dp = dp + this.e + 1;
         }
-        return format(this, dp, rm);
+        return format(this, dp, rm2);
       };
-      P4.toFormat = function(dp, rm, format2) {
+      P4.toFormat = function(dp, rm2, format2) {
         var str, x = this;
         if (format2 == null) {
-          if (dp != null && rm && typeof rm == "object") {
-            format2 = rm;
-            rm = null;
+          if (dp != null && rm2 && typeof rm2 == "object") {
+            format2 = rm2;
+            rm2 = null;
           } else if (dp && typeof dp == "object") {
             format2 = dp;
-            dp = rm = null;
+            dp = rm2 = null;
           } else {
             format2 = FORMAT;
           }
         } else if (typeof format2 != "object") {
           throw Error(bignumberError + "Argument not an object: " + format2);
         }
-        str = x.toFixed(dp, rm);
+        str = x.toFixed(dp, rm2);
         if (x.c) {
           var i, arr = str.split("."), g1 = +format2.groupSize, g2 = +format2.secondaryGroupSize, groupSeparator = format2.groupSeparator || "", intPart = arr[0], fractionPart = arr[1], isNeg = x.s < 0, intDigits = isNeg ? intPart.slice(1) : intPart, len = intDigits.length;
           if (g2) {
@@ -293310,10 +293310,10 @@ var require_bignumber = __commonJS((exports, module) => {
       P4.toNumber = function() {
         return +valueOf(this);
       };
-      P4.toPrecision = function(sd, rm) {
+      P4.toPrecision = function(sd, rm2) {
         if (sd != null)
           intCheck(sd, 1, MAX);
-        return format(this, sd, rm, 2);
+        return format(this, sd, rm2, 2);
       };
       P4.toString = function(b3) {
         var str, n2 = this, s = n2.s, e = n2.e;
@@ -320605,6 +320605,18 @@ import { statSync } from "fs";
 import { tmpdir as tmpdir8 } from "os";
 import { basename as basename4, join as join18 } from "path";
 import { promisify as promisify6 } from "util";
+function shouldStopAudioFallback(result) {
+  return result.success || Boolean(result.errorMessage?.includes("Normalized audio still exceeds provider upload limit"));
+}
+async function cleanupNormalizedAudio(normalizedAudioPromise) {
+  if (!normalizedAudioPromise)
+    return;
+  try {
+    const normalized = await normalizedAudioPromise;
+    if (normalized.cleanupPath)
+      await fs12.rm(normalized.cleanupPath, { recursive: true, force: true });
+  } catch {}
+}
 function buildPrompt2(language, options, defaultPrompt, defaultGlossary) {
   const parts = [];
   if (options?.prompt || defaultPrompt)
@@ -320638,15 +320650,6 @@ function extractText2(content) {
     return raw2;
   }
 }
-async function normalizeFileForOpenAiAudioChat(filePath, mimeType) {
-  const normalized = await normalizeAudioFileForProvider(filePath, mimeType);
-  try {
-    return { audio: await fs12.readFile(normalized.filePath), format: normalized.format === "wav" ? "wav" : "mp3" };
-  } finally {
-    if (normalized.cleanupPath)
-      await fs12.rm(normalized.cleanupPath, { recursive: true, force: true });
-  }
-}
 async function normalizeAudioFileForProvider(filePath, mimeType) {
   const format = toOpenAiAudioFormat2(mimeType);
   const stats = statSync(filePath);
@@ -320676,6 +320679,10 @@ async function normalizeAudioFileForProvider(filePath, mimeType) {
     ], {
       timeout: 5 * 60000
     });
+    const normalizedStats = await fs12.stat(output);
+    if (normalizedStats.size > PROVIDER_AUDIO_TARGET_BYTES) {
+      throw new Error(`Normalized audio still exceeds provider upload limit: ${normalizedStats.size} bytes > ${PROVIDER_AUDIO_TARGET_BYTES} bytes`);
+    }
     return { filePath: output, mimeType: "audio/mpeg", format: "mp3", cleanupPath: dir };
   } catch (error3) {
     await fs12.rm(dir, { recursive: true, force: true });
@@ -320746,51 +320753,81 @@ var init_audio2 = __esm(() => {
       const durationSeconds = options?.durationSeconds ?? this.estimateDuration(filePath);
       const provider = options?.provider ?? this.config.audioProvider ?? "openai";
       const preferredModel = options?.model ?? this.config.audioModel;
-      const attempts = [];
-      if (provider === "openai") {
-        attempts.push(() => this.transcribeWithOpenAiAudioChat(filePath, language, mimeType, options, preferredModel ?? OPENAI_AUDIO_CHAT_MODEL), () => this.transcribeWithOpenAiTranscriptions(filePath, language, mimeType, options, OPENAI_TRANSCRIBE_MODEL), () => this.transcribeWithGemini(filePath, language, mimeType, options, GEMINI_AUDIO_MODEL), () => this.transcribeWithGroq(filePath, language, mimeType));
-      } else if (provider === "gemini") {
-        attempts.push(() => this.transcribeWithGemini(filePath, language, mimeType, options, preferredModel ?? GEMINI_AUDIO_MODEL), () => this.transcribeWithOpenAiAudioChat(filePath, language, mimeType, options, OPENAI_AUDIO_CHAT_MODEL), () => this.transcribeWithGroq(filePath, language, mimeType));
-      } else {
-        attempts.push(() => this.transcribeWithGroq(filePath, language, mimeType), () => this.transcribeWithOpenAiTranscriptions(filePath, language, mimeType, options, preferredModel ?? OPENAI_TRANSCRIBE_MODEL));
-      }
+      let normalizedAudioPromise;
+      const getNormalizedAudio = () => {
+        normalizedAudioPromise ??= normalizeAudioFileForProvider(filePath, mimeType);
+        return normalizedAudioPromise;
+      };
+      const attempts = this.buildTranscriptionAttempts(provider, language, options, preferredModel, getNormalizedAudio);
       let result = this.createFailedResult("No transcription attempts configured", provider, preferredModel ?? "unknown");
-      for (const attempt of attempts) {
-        result = await attempt();
-        if (result.success)
-          break;
-        this.log.warn("Audio transcription attempt failed; trying next fallback", {
-          provider: result.provider,
-          model: result.model,
-          error: result.errorMessage
-        });
+      try {
+        for (const attempt of attempts) {
+          result = await attempt();
+          if (shouldStopAudioFallback(result))
+            break;
+          this.log.warn("Audio transcription attempt failed; trying next fallback", {
+            provider: result.provider,
+            model: result.model,
+            error: result.errorMessage
+          });
+        }
+      } finally {
+        await cleanupNormalizedAudio(normalizedAudioPromise);
       }
       result.processingTimeMs = Math.round(performance.now() - startTime2);
-      if (result.success && durationSeconds) {
-        result.duration = durationSeconds;
-        const pricingKey = result.provider === "groq" ? "groq_whisper" : "openai_whisper";
-        result.costCents = calculateCost(pricingKey, result.model, { durationSeconds });
-      }
-      if (result.success) {
-        this.log.info("Audio transcription successful", {
-          provider: result.provider,
-          model: result.model,
-          processingTimeMs: result.processingTimeMs,
-          costCents: result.costCents
-        });
-      } else {
+      this.applyDurationAndCost(result, durationSeconds);
+      this.logProcessingResult(result);
+      return result;
+    }
+    applyDurationAndCost(result, durationSeconds) {
+      if (!result.success || !durationSeconds)
+        return;
+      result.duration = durationSeconds;
+      const pricingKey = result.provider === "groq" ? "groq_whisper" : "openai_whisper";
+      result.costCents = calculateCost(pricingKey, result.model, { durationSeconds });
+    }
+    logProcessingResult(result) {
+      if (!result.success) {
         this.log.error("Audio transcription failed", { error: result.errorMessage });
+        return;
       }
-      return result;
+      this.log.info("Audio transcription successful", {
+        provider: result.provider,
+        model: result.model,
+        processingTimeMs: result.processingTimeMs,
+        costCents: result.costCents
+      });
+    }
+    buildTranscriptionAttempts(provider, language, options, preferredModel, getNormalizedAudio) {
+      if (provider === "openai") {
+        return [
+          () => this.transcribeWithOpenAiAudioChat(language, options, preferredModel ?? OPENAI_AUDIO_CHAT_MODEL, getNormalizedAudio),
+          () => this.transcribeWithOpenAiTranscriptions(language, options, OPENAI_TRANSCRIBE_MODEL, getNormalizedAudio),
+          () => this.transcribeWithGemini(language, options, GEMINI_AUDIO_MODEL, getNormalizedAudio),
+          () => this.transcribeWithGroq(language, getNormalizedAudio)
+        ];
+      }
+      if (provider === "gemini") {
+        return [
+          () => this.transcribeWithGemini(language, options, preferredModel ?? GEMINI_AUDIO_MODEL, getNormalizedAudio),
+          () => this.transcribeWithOpenAiAudioChat(language, options, OPENAI_AUDIO_CHAT_MODEL, getNormalizedAudio),
+          () => this.transcribeWithGroq(language, getNormalizedAudio)
+        ];
+      }
+      return [
+        () => this.transcribeWithGroq(language, getNormalizedAudio),
+        () => this.transcribeWithOpenAiTranscriptions(language, options, preferredModel ?? OPENAI_TRANSCRIBE_MODEL, getNormalizedAudio)
+      ];
     }
-    async transcribeWithOpenAiAudioChat(filePath, language, mimeType, options, model) {
+    async transcribeWithOpenAiAudioChat(language, options, model, getNormalizedAudio) {
       if (!this.config.openaiApiKey) {
         return this.createFailedResult("OpenAI client not configured (missing API key)", "openai", model);
       }
       const timeouts = getMediaTimeouts();
       try {
         const text3 = await this.executeWithResilience("openai", async () => {
-          const normalized = await normalizeFileForOpenAiAudioChat(filePath, mimeType);
+          const normalized = await getNormalizedAudio();
+          const audio = await fs12.readFile(normalized.filePath);
           const response = await fetch(OPENAI_CHAT_COMPLETIONS_URL, {
             method: "POST",
             headers: {
@@ -320810,8 +320847,8 @@ var init_audio2 = __esm(() => {
                     {
                       type: "input_audio",
                       input_audio: {
-                        data: normalized.audio.toString("base64"),
-                        format: normalized.format
+                        data: audio.toString("base64"),
+                        format: normalized.format === "wav" ? "wav" : "mp3"
                       }
                     }
                   ]
@@ -320831,36 +320868,31 @@ var init_audio2 = __esm(() => {
         return this.createFailedResult(errorMsg, "openai", model);
       }
     }
-    async transcribeWithOpenAiTranscriptions(filePath, language, mimeType, options, model) {
+    async transcribeWithOpenAiTranscriptions(language, options, model, getNormalizedAudio) {
       if (!this.config.openaiApiKey) {
         return this.createFailedResult("OpenAI client not configured (missing API key)", "openai", model);
       }
       const timeouts = getMediaTimeouts();
       try {
         const text3 = await this.executeWithResilience("openai", async () => {
-          const normalized = await normalizeAudioFileForProvider(filePath, mimeType);
-          try {
-            const fileBuffer = await Bun.file(normalized.filePath).arrayBuffer();
-            const form = new FormData;
-            form.append("file", new File([fileBuffer], `audio.${normalized.format}`, { type: normalized.mimeType }));
-            form.append("model", model);
-            form.append("response_format", "json");
-            if (language)
-              form.append("language", language.toLowerCase() === "pt-br" ? "pt" : language);
-            form.append("prompt", buildPrompt2(language, options, this.config.audioPrompt, this.config.audioGlossary));
-            const response = await fetch(OPENAI_TRANSCRIPTIONS_URL, {
-              method: "POST",
-              headers: { Authorization: `Bearer ${this.config.openaiApiKey}` },
-              body: form
-            });
-            if (!response.ok)
-              throw new Error(`OpenAI transcription error (${response.status}): ${await response.text()}`);
-            const data = await response.json();
-            return data.text ?? "";
-          } finally {
-            if (normalized.cleanupPath)
-              await fs12.rm(normalized.cleanupPath, { recursive: true, force: true });
-          }
+          const normalized = await getNormalizedAudio();
+          const fileBuffer = await Bun.file(normalized.filePath).arrayBuffer();
+          const form = new FormData;
+          form.append("file", new File([fileBuffer], `audio.${normalized.format}`, { type: normalized.mimeType }));
+          form.append("model", model);
+          form.append("response_format", "json");
+          if (language)
+            form.append("language", language.toLowerCase() === "pt-br" ? "pt" : language);
+          form.append("prompt", buildPrompt2(language, options, this.config.audioPrompt, this.config.audioGlossary));
+          const response = await fetch(OPENAI_TRANSCRIPTIONS_URL, {
+            method: "POST",
+            headers: { Authorization: `Bearer ${this.config.openaiApiKey}` },
+            body: form
+          });
+          if (!response.ok)
+            throw new Error(`OpenAI transcription error (${response.status}): ${await response.text()}`);
+          const data = await response.json();
+          return data.text ?? "";
         }, { timeoutMs: timeouts.audioTimeoutMs });
         return this.createSuccessResult(text3, "openai", model, language);
       } catch (error3) {
@@ -320868,46 +320900,41 @@ var init_audio2 = __esm(() => {
         return this.createFailedResult(errorMsg, "openai", model);
       }
     }
-    async transcribeWithGemini(filePath, language, mimeType, options, model) {
+    async transcribeWithGemini(language, options, model, getNormalizedAudio) {
       if (!this.config.geminiApiKey) {
         return this.createFailedResult("Gemini client not configured (missing API key)", "gemini", model);
       }
       const timeouts = getMediaTimeouts();
       try {
         const text3 = await this.executeWithResilience("gemini", async () => {
-          const normalized = await normalizeAudioFileForProvider(filePath, mimeType);
-          try {
-            const fileBuffer = await Bun.file(normalized.filePath).arrayBuffer();
-            const url = `${GEMINI_GENERATE_URL}/${encodeURIComponent(model)}:generateContent?key=${encodeURIComponent(this.config.geminiApiKey ?? "")}`;
-            const response = await fetch(url, {
-              method: "POST",
-              headers: { "Content-Type": "application/json" },
-              body: JSON.stringify({
-                contents: [
-                  {
-                    role: "user",
-                    parts: [
-                      { text: buildPrompt2(language, options, this.config.audioPrompt, this.config.audioGlossary) },
-                      {
-                        inlineData: {
-                          data: Buffer.from(fileBuffer).toString("base64"),
-                          mimeType: normalizeGeminiAudioMimeType(normalized.mimeType)
-                        }
+          const normalized = await getNormalizedAudio();
+          const fileBuffer = await Bun.file(normalized.filePath).arrayBuffer();
+          const url = `${GEMINI_GENERATE_URL}/${encodeURIComponent(model)}:generateContent?key=${encodeURIComponent(this.config.geminiApiKey ?? "")}`;
+          const response = await fetch(url, {
+            method: "POST",
+            headers: { "Content-Type": "application/json" },
+            body: JSON.stringify({
+              contents: [
+                {
+                  role: "user",
+                  parts: [
+                    { text: buildPrompt2(language, options, this.config.audioPrompt, this.config.audioGlossary) },
+                    {
+                      inlineData: {
+                        data: Buffer.from(fileBuffer).toString("base64"),
+                        mimeType: normalizeGeminiAudioMimeType(normalized.mimeType)
                       }
-                    ]
-                  }
-                ]
-              })
-            });
-            if (!response.ok)
-              throw new Error(`Gemini transcription error (${response.status}): ${await response.text()}`);
-            const data = await response.json();
-            return data.candidates?.[0]?.content?.parts?.map((p2) => p2.text ?? "").join(`
+                    }
+                  ]
+                }
+              ]
+            })
+          });
+          if (!response.ok)
+            throw new Error(`Gemini transcription error (${response.status}): ${await response.text()}`);
+          const data = await response.json();
+          return data.candidates?.[0]?.content?.parts?.map((p2) => p2.text ?? "").join(`
 `) ?? "";
-          } finally {
-            if (normalized.cleanupPath)
-              await fs12.rm(normalized.cleanupPath, { recursive: true, force: true });
-          }
         }, { timeoutMs: timeouts.audioTimeoutMs });
         return this.createSuccessResult(text3, "gemini", model, language);
       } catch (error3) {
@@ -320915,29 +320942,24 @@ var init_audio2 = __esm(() => {
         return this.createFailedResult(errorMsg, "gemini", model);
       }
     }
-    async transcribeWithGroq(filePath, language, mimeType) {
+    async transcribeWithGroq(language, getNormalizedAudio) {
       const client = this.getGroqClient();
       if (!client)
         return this.createFailedResult("Groq client not configured (missing API key)", "groq", GROQ_WHISPER_MODEL);
       const timeouts = getMediaTimeouts();
       try {
         const text3 = await this.executeWithResilience("groq", async () => {
-          const normalized = await normalizeAudioFileForProvider(filePath, mimeType ?? "");
-          try {
-            const filename = `audio.${normalized.format}`;
-            const fileBuffer = await Bun.file(normalized.filePath).arrayBuffer();
-            const file = new File([fileBuffer], filename, { type: normalized.mimeType });
-            const transcription = await client.audio.transcriptions.create({
-              file,
-              model: GROQ_WHISPER_MODEL,
-              language: language.toLowerCase() === "pt-br" ? "pt" : language,
-              response_format: "text"
-            });
-            return typeof transcription === "string" ? transcription : transcription.text ?? "";
-          } finally {
-            if (normalized.cleanupPath)
-              await fs12.rm(normalized.cleanupPath, { recursive: true, force: true });
-          }
+          const normalized = await getNormalizedAudio();
+          const filename = `audio.${normalized.format}`;
+          const fileBuffer = await Bun.file(normalized.filePath).arrayBuffer();
+          const file = new File([fileBuffer], filename, { type: normalized.mimeType });
+          const transcription = await client.audio.transcriptions.create({
+            file,
+            model: GROQ_WHISPER_MODEL,
+            language: language.toLowerCase() === "pt-br" ? "pt" : language,
+            response_format: "text"
+          });
+          return typeof transcription === "string" ? transcription : transcription.text ?? "";
         }, { timeoutMs: timeouts.audioTimeoutMs });
         return this.createSuccessResult(text3, "groq", GROQ_WHISPER_MODEL, language);
       } catch (error3) {
@@ -487099,7 +487121,7 @@ import { basename, join as join9 } from "path";
 // ../channel-whatsapp/src/utils/download.ts
 init_src2();
 import { createWriteStream as createWriteStream3 } from "fs";
-import { mkdir as mkdir2, writeFile as writeFile3 } from "fs/promises";
+import { mkdir as mkdir2, rm, writeFile as writeFile3 } from "fs/promises";
 import { dirname as dirname2, join as join8 } from "path";
 import { Transform as Transform2 } from "stream";
 import { pipeline as pipeline2 } from "stream/promises";
@@ -487208,6 +487230,13 @@ async function downloadMediaToFile(msg, outputPath, maxSizeBytes = getWhatsAppMe
   const mediaInfo = detectMediaType(msg);
   if (!mediaInfo)
     return null;
+  const stream = await downloadMediaMessage(msg, "stream", {});
+  const size = await writeMediaStreamToFile(stream, outputPath, maxSizeBytes);
+  if (size === 0)
+    return null;
+  return { mimeType: mediaInfo.mimeType, size };
+}
+async function writeMediaStreamToFile(stream, outputPath, maxSizeBytes = getWhatsAppMediaDownloadMaxBytes()) {
   let size = 0;
   const sizeGuard = new Transform2({
     transform(chunk2, _encoding, callback) {
@@ -487219,12 +487248,17 @@ async function downloadMediaToFile(msg, outputPath, maxSizeBytes = getWhatsAppMe
       callback(null, chunk2);
     }
   });
-  const stream = await downloadMediaMessage(msg, "stream", {});
   await mkdir2(dirname2(outputPath), { recursive: true });
-  await pipeline2(stream, sizeGuard, createWriteStream3(outputPath));
-  if (size === 0)
-    return null;
-  return { mimeType: mediaInfo.mimeType, size };
+  try {
+    await pipeline2(stream, sizeGuard, createWriteStream3(outputPath));
+  } catch (error) {
+    await rm(outputPath, { force: true });
+    throw error;
+  }
+  if (size === 0) {
+    await rm(outputPath, { force: true });
+  }
+  return size;
 }
 
 // ../channel-whatsapp/src/handlers/media.ts

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "@automagik/omni",
-  "version": "2.260530.2",
+  "version": "2.260530.3",
   "description": "LLM-optimized CLI for Omni",
   "type": "module",
   "bin": {