From 0bc2509e41af14b4481b4081760cb42d030156c3 Mon Sep 17 00:00:00 2001
From: ysslang <ysslang@qq.com>
Date: Sun, 1 Mar 2026 05:50:00 +0000
Subject: [PATCH] feat: add STT prompt injection configurable via
 STT_PROMPT_INJECTION

---
 .env.example              |  1 +
 src/bot/handlers/voice.ts | 18 +++++++++++++++++-
 src/config.ts             |  1 +
 3 files changed, 19 insertions(+), 1 deletion(-)

diff --git a/.env.example b/.env.example
index 439abdd..02bdb91 100644
--- a/.env.example
+++ b/.env.example
@@ -72,3 +72,4 @@ OPENCODE_MODEL_ID=big-pickle
 # STT_API_KEY=
 # STT_MODEL=
 # STT_LANGUAGE=
+# STT_PROMPT_INJECTION=true # Set to "true" to inject a warning about potential STT typos to the LLM. Or set a custom string.
diff --git a/src/bot/handlers/voice.ts b/src/bot/handlers/voice.ts
index 67d9f68..8a73671 100644
--- a/src/bot/handlers/voice.ts
+++ b/src/bot/handlers/voice.ts
@@ -219,8 +219,24 @@ export async function handleVoiceMessage(ctx: Context, deps: VoiceMessageDeps):
 
     logger.info(`[Voice] Transcribed audio: ${recognizedText.length} chars`);
 
+    // --- NEW LOGIC: Inject Prompt for LLM ---
+    let textForLLM = recognizedText;
+    const injectionSetting = config.stt.promptInjection;
+
+    if (injectionSetting) {
+      const normalizedSetting = injectionSetting.trim().toLowerCase();
+      if (normalizedSetting === "true" || normalizedSetting === "1") {
+        const defaultPrompt =
+          "[Note: The following text is transcribed from voice. There may be homophone or mispronunciation errors. Please interpret the intended meaning based on context.]";
+        textForLLM = `${defaultPrompt}\n${recognizedText}`;
+      } else if (normalizedSetting !== "false" && normalizedSetting !== "0") {
+        textForLLM = `${injectionSetting}\n${recognizedText}`;
+      }
+    }
+
     // Process the recognized text as a prompt
-    await processPrompt(ctx, recognizedText, deps);
+    await processPrompt(ctx, textForLLM, deps);
+
   } catch (err) {
     const errorMessage = err instanceof Error ? err.message : "unknown error";
     logger.error("[Voice] Error processing voice message:", err);
diff --git a/src/config.ts b/src/config.ts
index 87931a9..0a8474f 100644
--- a/src/config.ts
+++ b/src/config.ts
@@ -131,5 +131,6 @@ export const config = {
     apiKey: getEnvVar("STT_API_KEY", false),
     model: getEnvVar("STT_MODEL", false) || "whisper-large-v3-turbo",
     language: getEnvVar("STT_LANGUAGE", false),
+    promptInjection: getEnvVar("STT_PROMPT_INJECTION", false),
   },
 };