From 8e97f926ce13c5a17444ebe0909e87495251bfdd Mon Sep 17 00:00:00 2001
From: draxxris <24984408+draxxris@users.noreply.github.com>
Date: Tue, 30 Jun 2026 07:38:22 -0400
Subject: [PATCH] change: support custom tmpDir path

---
 README.md  |  1 +
 lib/stt.js | 38 +++++++++++++++++++++++++-------------
 2 files changed, 26 insertions(+), 13 deletions(-)

diff --git a/README.md b/README.md
index 3830f3b..be2a2fa 100644
--- a/README.md
+++ b/README.md
@@ -143,6 +143,7 @@ For unauthenticated local endpoints (e.g. Ollama):
 - `reasoningEffort` _(optional)_ - reasoning level for models that support it
 - `chatTemplateKwargs` _(optional)_ - extra keyword arguments passed to the model's chat template (e.g. `{"enable_thinking": false}` for Qwen models to disable chain-of-thought)
 - `retries` _(optional)_ - number of retry attempts for transient LLM failures
+- `tmpDir` _(optional)_ - directory used for the temporary STT recording file (default `/tmp`)
 
 ### Logging
 
diff --git a/lib/stt.js b/lib/stt.js
index 51eb539..809703b 100644
--- a/lib/stt.js
+++ b/lib/stt.js
@@ -10,7 +10,8 @@ let sttApiEndpoint = null;
 let sttApiModel = null;
 let sttApiKeyEnv = null;
 
-const WAV_FILE = "/tmp/opencode-stt.wav";
+const WAV_FILENAME = "opencode-stt.wav";
+let tmpDir = "/tmp";
 
 const MODELS_DIRS = [
   path.join(os.homedir(), ".local", "share", "whisper-cpp"),
@@ -120,9 +121,10 @@ function startRecording(kv, toast, logger) {
     return;
   }
 
+  const wavFile = path.join(tmpDir, WAV_FILENAME);
   forceKillSox(logger);
   try {
-    fs.unlinkSync(WAV_FILE);
+    fs.unlinkSync(wavFile);
   } catch {}
 
   soxStderr = "";
@@ -132,7 +134,7 @@ function startRecording(kv, toast, logger) {
 
   soxProc = spawn(
     "sox",
-    [...inputArgs, "-r", "16000", "-c", "1", "-b", "16", WAV_FILE, "silence", "1", "0.1", "1%"],
+    [...inputArgs, "-r", "16000", "-c", "1", "-b", "16", wavFile, "silence", "1", "0.1", "1%"],
     {
       stdio: ["ignore", "ignore", "pipe"],
       detached: false,
@@ -195,6 +197,7 @@ function getModelPath(kv) {
 }
 
 function transcribe(kv, logger) {
+  const wavFile = path.join(tmpDir, WAV_FILENAME);
   const mp = getModelPath(kv);
   logger?.log("STT", `Local transcription requested model=${mp}`, "debug");
   if (!fs.existsSync(mp)) {
@@ -203,19 +206,19 @@ function transcribe(kv, logger) {
       error: `Model not found: ${getModelName(kv)}. Download from huggingface.co/ggerganov/whisper.cpp`,
     });
   }
-  if (!fs.existsSync(WAV_FILE)) {
-    logger?.log("STT", `Recording file missing: ${WAV_FILE}`, "error");
+  if (!fs.existsSync(wavFile)) {
+    logger?.log("STT", `Recording file missing: ${wavFile}`, "error");
     return Promise.resolve({ error: "No recording file - sox may have failed to capture audio" });
   }
-  if (fs.statSync(WAV_FILE).size <= 44) {
-    logger?.log("STT", `Recording file empty: ${WAV_FILE}`, "warn");
+  if (fs.statSync(wavFile).size <= 44) {
+    logger?.log("STT", `Recording file empty: ${wavFile}`, "warn");
     return Promise.resolve({ error: "Recording is empty - no audio captured" });
   }
 
   return new Promise((resolve) => {
     let stdout = "";
     let stderr = "";
-    const proc = spawn("whisper-cli", ["-m", mp, "-f", WAV_FILE, "-np", "-nt"], {
+    const proc = spawn("whisper-cli", ["-m", mp, "-f", wavFile, "-np", "-nt"], {
       stdio: ["ignore", "pipe", "pipe"],
     });
     logger?.log("STT", `Started whisper-cli pid=${proc.pid}`, "debug");
@@ -330,20 +333,21 @@ async function transcribeApi(kv, logger) {
     logger?.log("STT", "STT API transcription skipped: API not configured", "warn");
     return { error: "STT API not configured" };
   }
+  const wavFile = path.join(tmpDir, WAV_FILENAME);
   const model = kv.get("stt.api.model") || sttApiModel;
   logger?.log("STT", `STT API transcription requested model=${model}`, "debug");
 
-  if (!fs.existsSync(WAV_FILE)) {
-    logger?.log("STT", `Recording file missing: ${WAV_FILE}`, "error");
+  if (!fs.existsSync(wavFile)) {
+    logger?.log("STT", `Recording file missing: ${wavFile}`, "error");
     return { error: "No recording file - sox may have failed to capture audio" };
   }
-  if (fs.statSync(WAV_FILE).size <= 44) {
-    logger?.log("STT", `Recording file empty: ${WAV_FILE}`, "warn");
+  if (fs.statSync(wavFile).size <= 44) {
+    logger?.log("STT", `Recording file empty: ${wavFile}`, "warn");
     return { error: "Recording is empty - no audio captured" };
   }
 
   try {
-    const audioBuffer = await fs.promises.readFile(WAV_FILE);
+    const audioBuffer = await fs.promises.readFile(wavFile);
     const apiKey = sttApiKeyEnv ? process.env[sttApiKeyEnv] : null;
     const useOpenRouterFormat = isOpenRouterEndpoint(sttApiEndpoint);
 
@@ -487,6 +491,14 @@ export function registerSTT(api, kv, complete, prompts, opts, logger) {
     );
   }
 
+  tmpDir = opts?.tmpDir || "/tmp";
+  try {
+    fs.mkdirSync(tmpDir, { recursive: true });
+  } catch (err) {
+    logger?.log("STT", `Failed to create tmpDir ${tmpDir}: ${err.message}`, "warn");
+  }
+  logger?.log("STT", `STT temp dir=${tmpDir}`, "debug");
+
   return [
     {
       title: sttApiEndpoint ? "STT: record/transcribe (API)" : "STT: record/transcribe",