Create app.py

app.py

@@ -0,0 +1,51 @@
+# app.py
+import os
+import numpy as np
+from fastapi import FastAPI, Form
+from fastapi.responses import PlainTextResponse
+from langchain_google_genai import ChatGoogleGenerativeAI
+from pydub import AudioSegment
+
+# ---------- CONFIG ----------
+os.environ["GOOGLE_API_KEY"] = "AIzaSyD2DMFgcL0kWTQYhii8wseSHY3BRGWSebk"
+llm = ChatGoogleGenerativeAI(model="gemini-2.0-flash")
+
+app = FastAPI()
+
+# ---------- HELPERS ----------
+def text_to_pcm_array(text: str) -> str:
+    """
+    Generate speech audio for `text`, convert to PCM unsigned 8-bit 8kHz mono,
+    and return as C-style array { 0,255, ... } for ESP32.
+    """
+
+    # 1. Generate response (you can replace with direct TTS if available)
+    response = llm.invoke(text).content
+    if not response:
+        response = "No response generated."
+
+    # 2. Save response as MP3 (dummy TTS workaround – here we just create a fake audio file)
+    # In practice, replace this with real TTS saving to 'response.mp3'
+    from gtts import gTTS
+    tts = gTTS(response)
+    tts.save("response.mp3")
+
+    # 3. Load audio with pydub
+    audio = AudioSegment.from_file("response.mp3")
+
+    # Convert to 8kHz, mono, 8-bit unsigned PCM
+    audio = audio.set_frame_rate(8000).set_channels(1).set_sample_width(1)
+
+    raw_data = audio.raw_data
+    samples = np.frombuffer(raw_data, dtype=np.uint8)
+
+    # 4. Convert to C-style array
+    c_array = "{ " + ",".join(map(str, samples)) + " }"
+    return c_array
+
+
+# ---------- ENDPOINT ----------
+@app.post("/send_message_recive_pcm", response_class=PlainTextResponse)
+async def send_message_recive_pcm(message: str = Form(...)):
+    pcm_array = text_to_pcm_array(message)
+    return pcm_array