Spaces:
Running
Running
Update app.py
Browse files
app.py
CHANGED
|
@@ -3,69 +3,86 @@ import gradio as gr
|
|
| 3 |
import requests
|
| 4 |
import google.generativeai as genai
|
| 5 |
|
| 6 |
-
|
| 7 |
-
from langchain_classic.memory import ConversationBufferMemory
|
| 8 |
-
|
| 9 |
-
# ------------------------------------
|
| 10 |
# Load API keys from environment
|
| 11 |
-
#
|
| 12 |
GEMINI_API_KEY = os.environ.get("GEMINI_API_KEY")
|
| 13 |
ELEVENLABS_API_KEY = os.environ.get("ELEVENLABS_API_KEY")
|
| 14 |
ELEVENLABS_VOICE_ID = "21m00Tcm4TlvDq8ikWAM"
|
| 15 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 16 |
# Configure Gemini
|
| 17 |
genai.configure(api_key=GEMINI_API_KEY)
|
| 18 |
|
| 19 |
-
#
|
| 20 |
-
#
|
| 21 |
-
#
|
| 22 |
-
|
| 23 |
{chat_history}
|
| 24 |
User: {user_message}
|
| 25 |
Chatbot:"""
|
| 26 |
|
| 27 |
-
|
| 28 |
-
|
| 29 |
-
|
| 30 |
-
|
|
|
|
| 31 |
|
| 32 |
-
|
|
|
|
|
|
|
| 33 |
|
| 34 |
-
|
| 35 |
-
|
| 36 |
-
|
| 37 |
-
gemini_model = genai.GenerativeModel("gemini-2.5-flash")
|
| 38 |
|
| 39 |
-
|
| 40 |
-
|
| 41 |
-
|
| 42 |
-
self.memory_history = []
|
| 43 |
|
| 44 |
-
def
|
| 45 |
-
|
| 46 |
-
for msg in self.memory_history:
|
| 47 |
-
full_prompt += msg + "\n"
|
| 48 |
-
full_prompt += f"User: {user_message}\nChatbot:"
|
| 49 |
|
| 50 |
-
|
| 51 |
-
answer = response.text
|
| 52 |
|
| 53 |
-
|
| 54 |
-
|
| 55 |
-
|
| 56 |
-
|
| 57 |
-
|
| 58 |
|
| 59 |
-
|
|
|
|
|
|
|
|
|
|
| 60 |
|
| 61 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 62 |
|
| 63 |
-
#
|
| 64 |
-
# ElevenLabs
|
| 65 |
-
#
|
| 66 |
def generate_audio_elevenlabs(text):
|
| 67 |
-
|
| 68 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 69 |
|
| 70 |
try:
|
| 71 |
client = ElevenLabs(api_key=ELEVENLABS_API_KEY)
|
|
@@ -75,32 +92,52 @@ def generate_audio_elevenlabs(text):
|
|
| 75 |
model="eleven_monolingual_v1"
|
| 76 |
)
|
| 77 |
|
|
|
|
| 78 |
output_path = f"/tmp/audio_{abs(hash(text)) % 100000}.mp3"
|
| 79 |
save(audio, output_path)
|
| 80 |
return output_path
|
| 81 |
-
|
| 82 |
except Exception as e:
|
| 83 |
-
|
|
|
|
| 84 |
return ""
|
| 85 |
|
| 86 |
-
#
|
| 87 |
# Combined response
|
| 88 |
-
#
|
| 89 |
-
def
|
| 90 |
-
|
| 91 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 92 |
return text, audio_path
|
| 93 |
|
| 94 |
-
#
|
| 95 |
-
# Gradio
|
| 96 |
-
#
|
| 97 |
def chat_bot_response(message, history):
|
| 98 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 99 |
return text
|
| 100 |
|
| 101 |
-
#
|
| 102 |
-
# UI (
|
| 103 |
-
#
|
| 104 |
demo = gr.ChatInterface(
|
| 105 |
fn=chat_bot_response,
|
| 106 |
title="🤖 Gemini + ElevenLabs Chatbot",
|
|
@@ -112,7 +149,7 @@ demo = gr.ChatInterface(
|
|
| 112 |
"What's the weather like today?",
|
| 113 |
"Explain quantum computing in simple terms"
|
| 114 |
],
|
| 115 |
-
theme=gr.themes.Soft()
|
| 116 |
)
|
| 117 |
|
| 118 |
if __name__ == "__main__":
|
|
|
|
| 3 |
import requests
|
| 4 |
import google.generativeai as genai
|
| 5 |
|
| 6 |
+
# -----------------------------
|
|
|
|
|
|
|
|
|
|
| 7 |
# Load API keys from environment
|
| 8 |
+
# -----------------------------
|
| 9 |
GEMINI_API_KEY = os.environ.get("GEMINI_API_KEY")
|
| 10 |
ELEVENLABS_API_KEY = os.environ.get("ELEVENLABS_API_KEY")
|
| 11 |
ELEVENLABS_VOICE_ID = "21m00Tcm4TlvDq8ikWAM"
|
| 12 |
|
| 13 |
+
# Safety: fail early if no keys
|
| 14 |
+
if not GEMINI_API_KEY:
|
| 15 |
+
raise RuntimeError("Missing GEMINI_API_KEY in environment")
|
| 16 |
+
if not ELEVENLABS_API_KEY:
|
| 17 |
+
# Not strictly fatal if you only want text, but warn:
|
| 18 |
+
print("Warning: ELEVENLABS_API_KEY not set — audio generation will fail.")
|
| 19 |
+
|
| 20 |
# Configure Gemini
|
| 21 |
genai.configure(api_key=GEMINI_API_KEY)
|
| 22 |
|
| 23 |
+
# -----------------------------
|
| 24 |
+
# Simple prompt template + memory (no LangChain)
|
| 25 |
+
# -----------------------------
|
| 26 |
+
PROMPT_TEMPLATE = """You are a helpful assistant.
|
| 27 |
{chat_history}
|
| 28 |
User: {user_message}
|
| 29 |
Chatbot:"""
|
| 30 |
|
| 31 |
+
# Use a simple in-memory ring buffer for conversation (keeps last N messages)
|
| 32 |
+
class SimpleMemory:
|
| 33 |
+
def __init__(self, max_messages=20):
|
| 34 |
+
self.max_messages = max_messages
|
| 35 |
+
self.history = [] # list of strings like "User: ...", "Chatbot: ..."
|
| 36 |
|
| 37 |
+
def add_user(self, text):
|
| 38 |
+
self.history.append(f"User: {text}")
|
| 39 |
+
self._trim()
|
| 40 |
|
| 41 |
+
def add_bot(self, text):
|
| 42 |
+
self.history.append(f"Chatbot: {text}")
|
| 43 |
+
self._trim()
|
|
|
|
| 44 |
|
| 45 |
+
def _trim(self):
|
| 46 |
+
if len(self.history) > self.max_messages:
|
| 47 |
+
self.history = self.history[-self.max_messages:]
|
|
|
|
| 48 |
|
| 49 |
+
def as_text(self):
|
| 50 |
+
return "\n".join(self.history)
|
|
|
|
|
|
|
|
|
|
| 51 |
|
| 52 |
+
memory = SimpleMemory(max_messages=20)
|
|
|
|
| 53 |
|
| 54 |
+
# -----------------------------
|
| 55 |
+
# Gemini wrapper (simple)
|
| 56 |
+
# -----------------------------
|
| 57 |
+
# Keep a single model instance
|
| 58 |
+
gemini_model = genai.GenerativeModel("gemini-2.5-flash")
|
| 59 |
|
| 60 |
+
def generate_text_with_gemini(user_message):
|
| 61 |
+
# Build full prompt
|
| 62 |
+
chat_history = memory.as_text()
|
| 63 |
+
full_prompt = PROMPT_TEMPLATE.format(chat_history=chat_history, user_message=user_message)
|
| 64 |
|
| 65 |
+
# Use the SDK to generate
|
| 66 |
+
# NOTE: API may return a response object; using .generate_content as before
|
| 67 |
+
response = gemini_model.generate_content(full_prompt)
|
| 68 |
+
# response.text contains the assistant output
|
| 69 |
+
text = response.text if hasattr(response, "text") else str(response)
|
| 70 |
+
return text
|
| 71 |
|
| 72 |
+
# -----------------------------
|
| 73 |
+
# ElevenLabs audio generation (lazy import)
|
| 74 |
+
# -----------------------------
|
| 75 |
def generate_audio_elevenlabs(text):
|
| 76 |
+
"""
|
| 77 |
+
Returns output_path on success, empty string on failure.
|
| 78 |
+
Uses /tmp directory (Hugging Face friendly).
|
| 79 |
+
"""
|
| 80 |
+
try:
|
| 81 |
+
from elevenlabs.client import ElevenLabs
|
| 82 |
+
from elevenlabs import save
|
| 83 |
+
except Exception as e:
|
| 84 |
+
print("ElevenLabs import failed:", e)
|
| 85 |
+
return ""
|
| 86 |
|
| 87 |
try:
|
| 88 |
client = ElevenLabs(api_key=ELEVENLABS_API_KEY)
|
|
|
|
| 92 |
model="eleven_monolingual_v1"
|
| 93 |
)
|
| 94 |
|
| 95 |
+
# Safe temp path for HF Spaces
|
| 96 |
output_path = f"/tmp/audio_{abs(hash(text)) % 100000}.mp3"
|
| 97 |
save(audio, output_path)
|
| 98 |
return output_path
|
|
|
|
| 99 |
except Exception as e:
|
| 100 |
+
# log and return empty result so UI still shows text
|
| 101 |
+
print("ElevenLabs generation error:", e)
|
| 102 |
return ""
|
| 103 |
|
| 104 |
+
# -----------------------------
|
| 105 |
# Combined response
|
| 106 |
+
# -----------------------------
|
| 107 |
+
def get_text_and_audio(user_message):
|
| 108 |
+
# Text
|
| 109 |
+
text = generate_text_with_gemini(user_message)
|
| 110 |
+
|
| 111 |
+
# Update memory
|
| 112 |
+
memory.add_user(user_message)
|
| 113 |
+
memory.add_bot(text)
|
| 114 |
+
|
| 115 |
+
# Audio (non-blocking style: we still run it but return text even if audio fails)
|
| 116 |
+
audio_path = ""
|
| 117 |
+
if ELEVENLABS_API_KEY:
|
| 118 |
+
audio_path = generate_audio_elevenlabs(text)
|
| 119 |
+
|
| 120 |
return text, audio_path
|
| 121 |
|
| 122 |
+
# -----------------------------
|
| 123 |
+
# Chat handler used by Gradio UI (unchanged UI)
|
| 124 |
+
# -----------------------------
|
| 125 |
def chat_bot_response(message, history):
|
| 126 |
+
"""
|
| 127 |
+
Gradio ChatInterface function signature:
|
| 128 |
+
- message: new user message (string)
|
| 129 |
+
- history: prior messages (list) provided by Gradio (we don't need to rely on it)
|
| 130 |
+
Return: text response (Gradio ChatInterface will render it)
|
| 131 |
+
"""
|
| 132 |
+
text, audio_path = get_text_and_audio(message)
|
| 133 |
+
|
| 134 |
+
# Optionally: attach audio playback by returning a dict or specific component.
|
| 135 |
+
# To keep the UI identical and simple, we return text only — same behavior as before.
|
| 136 |
return text
|
| 137 |
|
| 138 |
+
# -----------------------------
|
| 139 |
+
# UI (exact ChatInterface parameters preserved, including theme)
|
| 140 |
+
# -----------------------------
|
| 141 |
demo = gr.ChatInterface(
|
| 142 |
fn=chat_bot_response,
|
| 143 |
title="🤖 Gemini + ElevenLabs Chatbot",
|
|
|
|
| 149 |
"What's the weather like today?",
|
| 150 |
"Explain quantum computing in simple terms"
|
| 151 |
],
|
| 152 |
+
theme=gr.themes.Soft() # keep theme — requirements pin ensures support
|
| 153 |
)
|
| 154 |
|
| 155 |
if __name__ == "__main__":
|