Update app.py
Browse files
app.py
CHANGED
|
@@ -1,41 +1,46 @@
|
|
| 1 |
-
|
|
|
|
|
|
|
|
|
|
| 2 |
import speech_recognition as sr
|
| 3 |
import tempfile
|
| 4 |
import scipy.io.wavfile
|
| 5 |
import os
|
| 6 |
-
import requests
|
| 7 |
import numpy as np
|
| 8 |
|
| 9 |
-
# API
|
| 10 |
-
os.environ["MISTRAL_API_KEY"] = "your-mistral-key"
|
| 11 |
-
os.environ["GROQ_API_KEY"] = "your-groq-key"
|
| 12 |
|
| 13 |
-
|
| 14 |
-
|
| 15 |
-
|
| 16 |
-
|
| 17 |
-
|
| 18 |
-
|
| 19 |
-
|
| 20 |
-
"consult_mode": False
|
| 21 |
-
}
|
| 22 |
|
| 23 |
-
|
| 24 |
-
|
| 25 |
-
if uploaded_file is None:
|
| 26 |
return ""
|
| 27 |
try:
|
| 28 |
-
|
| 29 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
| 30 |
tmp_file_path = tmp_file.name
|
| 31 |
|
| 32 |
recognizer = sr.Recognizer()
|
| 33 |
with sr.AudioFile(tmp_file_path) as source:
|
| 34 |
-
|
| 35 |
-
|
| 36 |
-
|
|
|
|
| 37 |
except Exception as e:
|
| 38 |
-
|
|
|
|
| 39 |
|
| 40 |
def query_llm(api, messages, model=None):
|
| 41 |
headers = {
|
|
@@ -46,90 +51,127 @@ def query_llm(api, messages, model=None):
|
|
| 46 |
"messages": messages,
|
| 47 |
"model": model or ("llama3-70b-8192" if api == "GROQ" else "mistral-medium")
|
| 48 |
}
|
| 49 |
-
|
| 50 |
-
|
| 51 |
-
|
| 52 |
-
|
| 53 |
-
return "API Error"
|
| 54 |
|
| 55 |
-
|
| 56 |
-
|
| 57 |
-
|
|
|
|
|
|
|
|
|
|
| 58 |
|
| 59 |
def generate_question(answers):
|
| 60 |
-
prompt = "You are playing
|
| 61 |
for i, (q, a) in enumerate(answers, 1):
|
| 62 |
prompt += f"{i}. Q: {q}\n A: {a}\n"
|
| 63 |
-
|
|
|
|
|
|
|
| 64 |
|
| 65 |
def make_guess(answers):
|
| 66 |
-
prompt = "Based on
|
| 67 |
for i, (q, a) in enumerate(answers, 1):
|
| 68 |
prompt += f"{i}. Q: {q}\n A: {a}\n"
|
| 69 |
-
|
|
|
|
| 70 |
|
| 71 |
def get_hint(question, answers):
|
| 72 |
-
prompt = f"The player is unsure about: {question}\n\nHistory:\n"
|
| 73 |
for q, a in answers:
|
| 74 |
prompt += f"- Q: {q}\n A: {a}\n"
|
| 75 |
-
|
|
|
|
| 76 |
|
| 77 |
-
|
| 78 |
-
|
| 79 |
-
|
| 80 |
-
st.title("🎮 Kasoti - 20 Questions Game")
|
| 81 |
-
st.info("Think of a famous person, place, or object. Answer with **yes/no** or **ہاں/نہیں** only.")
|
| 82 |
|
| 83 |
-
|
| 84 |
-
|
| 85 |
"active": True,
|
| 86 |
"questions_asked": 1,
|
| 87 |
"answers": [],
|
| 88 |
"current_question": "Is it a living thing?",
|
| 89 |
"consult_mode": False
|
| 90 |
})
|
|
|
|
|
|
|
| 91 |
|
| 92 |
-
|
| 93 |
-
|
|
|
|
| 94 |
|
| 95 |
-
|
| 96 |
-
|
| 97 |
-
|
| 98 |
-
manual_input = st.text_input("💬 Or type your answer:")
|
| 99 |
-
submit = st.form_submit_button("➡️ Submit")
|
| 100 |
|
| 101 |
-
if
|
| 102 |
-
|
| 103 |
-
|
| 104 |
-
|
| 105 |
-
st.warning("⚠️ Please reply with 'yes' or 'no' (or 'ہاں/نہیں').")
|
| 106 |
else:
|
| 107 |
-
|
| 108 |
-
|
| 109 |
-
|
| 110 |
-
|
| 111 |
-
|
| 112 |
-
|
| 113 |
-
|
| 114 |
-
|
| 115 |
-
|
| 116 |
-
|
| 117 |
-
|
| 118 |
-
|
| 119 |
-
|
| 120 |
-
|
| 121 |
-
|
| 122 |
-
|
| 123 |
-
|
| 124 |
-
|
| 125 |
-
|
| 126 |
-
|
| 127 |
-
|
| 128 |
-
|
| 129 |
-
|
| 130 |
-
|
| 131 |
-
|
| 132 |
-
if
|
| 133 |
-
|
| 134 |
-
|
| 135 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
# app.py
|
| 2 |
+
|
| 3 |
+
import gradio as gr
|
| 4 |
+
import requests
|
| 5 |
import speech_recognition as sr
|
| 6 |
import tempfile
|
| 7 |
import scipy.io.wavfile
|
| 8 |
import os
|
|
|
|
| 9 |
import numpy as np
|
| 10 |
|
| 11 |
+
# Set your API keys as HF secrets or environment variables
|
| 12 |
+
os.environ["MISTRAL_API_KEY"] = os.getenv("MISTRAL_API_KEY", "your-mistral-key")
|
| 13 |
+
os.environ["GROQ_API_KEY"] = os.getenv("GROQ_API_KEY", "your-groq-key")
|
| 14 |
|
| 15 |
+
game_state = {
|
| 16 |
+
"active": False,
|
| 17 |
+
"questions_asked": 0,
|
| 18 |
+
"answers": [],
|
| 19 |
+
"current_question": None,
|
| 20 |
+
"consult_mode": False
|
| 21 |
+
}
|
|
|
|
|
|
|
| 22 |
|
| 23 |
+
def transcribe_audio(audio, language):
|
| 24 |
+
if audio is None:
|
|
|
|
| 25 |
return ""
|
| 26 |
try:
|
| 27 |
+
sr_rate, audio_data = audio
|
| 28 |
+
if isinstance(audio_data, tuple):
|
| 29 |
+
audio_data = np.array(audio_data)
|
| 30 |
+
|
| 31 |
+
with tempfile.NamedTemporaryFile(suffix=".wav", delete=False) as tmp_file:
|
| 32 |
+
scipy.io.wavfile.write(tmp_file.name, sr_rate, audio_data)
|
| 33 |
tmp_file_path = tmp_file.name
|
| 34 |
|
| 35 |
recognizer = sr.Recognizer()
|
| 36 |
with sr.AudioFile(tmp_file_path) as source:
|
| 37 |
+
audio_content = recognizer.record(source)
|
| 38 |
+
language_code = "en-US" if language == "English" else "ur-PK"
|
| 39 |
+
text = recognizer.recognize_google(audio_content, language=language_code)
|
| 40 |
+
return text.lower()
|
| 41 |
except Exception as e:
|
| 42 |
+
print(f"Transcription error: {e}")
|
| 43 |
+
return ""
|
| 44 |
|
| 45 |
def query_llm(api, messages, model=None):
|
| 46 |
headers = {
|
|
|
|
| 51 |
"messages": messages,
|
| 52 |
"model": model or ("llama3-70b-8192" if api == "GROQ" else "mistral-medium")
|
| 53 |
}
|
| 54 |
+
endpoint = {
|
| 55 |
+
"MISTRAL": "https://api.mistral.ai/v1/chat/completions",
|
| 56 |
+
"GROQ": "https://api.groq.com/openai/v1/chat/completions"
|
| 57 |
+
}[api]
|
|
|
|
| 58 |
|
| 59 |
+
response = requests.post(endpoint, headers=headers, json=payload)
|
| 60 |
+
if response.status_code == 200:
|
| 61 |
+
return response.json()["choices"][0]["message"]["content"]
|
| 62 |
+
else:
|
| 63 |
+
print(f"Error from {api} API: {response.text}")
|
| 64 |
+
return None
|
| 65 |
|
| 66 |
def generate_question(answers):
|
| 67 |
+
prompt = "You are playing a game called Kasoti (20 Questions)...\n"
|
| 68 |
for i, (q, a) in enumerate(answers, 1):
|
| 69 |
prompt += f"{i}. Q: {q}\n A: {a}\n"
|
| 70 |
+
prompt += "\nAsk ONLY the next best yes/no question."
|
| 71 |
+
response = query_llm("GROQ", [{"role": "user", "content": prompt}])
|
| 72 |
+
return response.strip() if response else "Is it something you can hold?"
|
| 73 |
|
| 74 |
def make_guess(answers):
|
| 75 |
+
prompt = "Based on the following yes/no history, make a best guess.\n\nHistory:\n"
|
| 76 |
for i, (q, a) in enumerate(answers, 1):
|
| 77 |
prompt += f"{i}. Q: {q}\n A: {a}\n"
|
| 78 |
+
response = query_llm("GROQ", [{"role": "user", "content": prompt}])
|
| 79 |
+
return response.strip() if response else "I need more information."
|
| 80 |
|
| 81 |
def get_hint(question, answers):
|
| 82 |
+
prompt = f"The player is unsure about answering: '{question}'\n\nHistory:\n"
|
| 83 |
for q, a in answers:
|
| 84 |
prompt += f"- Q: {q}\n A: {a}\n"
|
| 85 |
+
prompt += "\nSuggest a helpful hint to clarify."
|
| 86 |
+
return query_llm("MISTRAL", [{"role": "user", "content": prompt}]) or "Consider the common meaning."
|
| 87 |
|
| 88 |
+
def normalize_answer(ans):
|
| 89 |
+
ans = ans.strip().lower()
|
| 90 |
+
return "yes" if ans in ["yes", "y", "ہاں", "haan"] else "no" if ans in ["no", "n", "نہیں", "nahi"] else None
|
|
|
|
|
|
|
| 91 |
|
| 92 |
+
def start_game():
|
| 93 |
+
game_state.update({
|
| 94 |
"active": True,
|
| 95 |
"questions_asked": 1,
|
| 96 |
"answers": [],
|
| 97 |
"current_question": "Is it a living thing?",
|
| 98 |
"consult_mode": False
|
| 99 |
})
|
| 100 |
+
intro = "🎯 **Kasoti Started!**\nThink of something... I'll guess in 20 questions.\n\n➡️ First Question: **Is it a living thing?**"
|
| 101 |
+
return intro, gr.update(interactive=True), gr.update(interactive=True), "🔕 Consult Mode: OFF", gr.update(visible=False)
|
| 102 |
|
| 103 |
+
def process_answer(answer_text):
|
| 104 |
+
if not game_state["active"]:
|
| 105 |
+
return "⚠️ Start the game first.", "", "", gr.update(visible=False)
|
| 106 |
|
| 107 |
+
normalized = normalize_answer(answer_text)
|
| 108 |
+
if normalized is None:
|
| 109 |
+
return "❌ Please reply with 'yes' or 'no' (or 'ہاں/نہیں').", "", answer_text, gr.update(visible=game_state["consult_mode"])
|
|
|
|
|
|
|
| 110 |
|
| 111 |
+
if "is this correct?" in (game_state["current_question"] or "").lower():
|
| 112 |
+
if normalized == "yes":
|
| 113 |
+
game_state["active"] = False
|
| 114 |
+
return "🎉 YAY! I guessed it!", "", answer_text, gr.update(visible=False)
|
|
|
|
| 115 |
else:
|
| 116 |
+
next_q = generate_question(game_state["answers"])
|
| 117 |
+
game_state["current_question"] = next_q
|
| 118 |
+
game_state["questions_asked"] += 1
|
| 119 |
+
return next_q, "", answer_text, gr.update(visible=game_state["consult_mode"])
|
| 120 |
+
|
| 121 |
+
game_state["answers"].append((game_state["current_question"], normalized))
|
| 122 |
+
|
| 123 |
+
if game_state["questions_asked"] >= 20:
|
| 124 |
+
game_state["active"] = False
|
| 125 |
+
guess = make_guess(game_state["answers"])
|
| 126 |
+
return f"🕹️ Game over! My final guess: **{guess}**", "", answer_text, gr.update(visible=False)
|
| 127 |
+
|
| 128 |
+
if game_state["questions_asked"] % 5 == 0:
|
| 129 |
+
guess = make_guess(game_state["answers"])
|
| 130 |
+
if guess.startswith("I think it's"):
|
| 131 |
+
game_state["current_question"] = guess + " Is this correct? (yes/no)"
|
| 132 |
+
return game_state["current_question"], "", answer_text, gr.update(visible=game_state["consult_mode"])
|
| 133 |
+
|
| 134 |
+
next_q = generate_question(game_state["answers"])
|
| 135 |
+
game_state["current_question"] = next_q
|
| 136 |
+
game_state["questions_asked"] += 1
|
| 137 |
+
return next_q, "", answer_text, gr.update(visible=game_state["consult_mode"])
|
| 138 |
+
|
| 139 |
+
def toggle_consult_mode():
|
| 140 |
+
game_state["consult_mode"] = not game_state["consult_mode"]
|
| 141 |
+
return ("��� Consult Mode: ON" if game_state["consult_mode"] else "🔕 Consult Mode: OFF",
|
| 142 |
+
gr.update(visible=game_state["consult_mode"]))
|
| 143 |
+
|
| 144 |
+
def get_consult_hint():
|
| 145 |
+
if not game_state["active"] or not game_state["consult_mode"]:
|
| 146 |
+
return "ℹ️ Consult mode is not active or game not started."
|
| 147 |
+
return get_hint(game_state["current_question"], game_state["answers"])
|
| 148 |
+
|
| 149 |
+
# Colorful and polished UI
|
| 150 |
+
with gr.Blocks(css=".gr-button {background-color: #4CAF50; color: white} .gr-box {background: #f7f7ff; border: 1px solid #ccc; padding: 10px; border-radius: 10px;}") as demo:
|
| 151 |
+
gr.Markdown("## 🧠 Kasoti: 20 Questions AI Game")
|
| 152 |
+
gr.Markdown("Think of a person, place, or thing. I'll try to guess it in 20 questions or less! Just answer 'yes' or 'no'.")
|
| 153 |
+
|
| 154 |
+
with gr.Row():
|
| 155 |
+
start_btn = gr.Button("🚀 Start Game")
|
| 156 |
+
consult_btn = gr.Button("💬 Toggle Consult Mode")
|
| 157 |
+
consult_status = gr.Textbox(label="Consult Mode", interactive=False)
|
| 158 |
+
|
| 159 |
+
with gr.Row():
|
| 160 |
+
with gr.Column():
|
| 161 |
+
language = gr.Dropdown(["English", "Urdu"], label="Audio Language", value="English")
|
| 162 |
+
audio_input = gr.Audio(label="🎤 Answer via Microphone", type="numpy", sources=["microphone"])
|
| 163 |
+
transcribe_btn = gr.Button("📝 Transcribe Audio")
|
| 164 |
+
transcribed_text = gr.Textbox(label="✍️ Answer Text", interactive=True)
|
| 165 |
+
submit_btn = gr.Button("✅ Submit Answer")
|
| 166 |
+
|
| 167 |
+
with gr.Column():
|
| 168 |
+
game_output = gr.Textbox(label="🎲 Game Progress", interactive=False)
|
| 169 |
+
consult_output = gr.Textbox(label="💡 Consult Hint", visible=False)
|
| 170 |
+
|
| 171 |
+
start_btn.click(start_game, outputs=[game_output, transcribed_text, submit_btn, consult_status, consult_output])
|
| 172 |
+
consult_btn.click(toggle_consult_mode, outputs=[consult_status, consult_output])
|
| 173 |
+
consult_btn.click(get_consult_hint, outputs=[consult_output])
|
| 174 |
+
transcribe_btn.click(transcribe_audio, inputs=[audio_input, language], outputs=[transcribed_text])
|
| 175 |
+
submit_btn.click(process_answer, inputs=[transcribed_text], outputs=[game_output, transcribed_text, transcribed_text, consult_output])
|
| 176 |
+
|
| 177 |
+
demo.launch()
|