commit 887030da5a18858733fd6a66aab6ad23dc912282
parent f00022d43d95e8945a1daa4d8cdad69de2965e97
Author: Amit Dutta <amitdutta4255@gmail.com>
Date: Thu, 17 Jul 2025 14:40:02 +0530
Update Chatbot.py
Diffstat:
1 file changed, 31 insertions(+), 17 deletions(-)
diff --git a/Backend/Chatbot.py b/Backend/Chatbot.py
@@ -116,6 +116,10 @@ def generate_image_response(prompt: str):
return f"An unexpected error occurred while processing your image request: {e}"
def ChatBot(Query, sender_number=None):
+ """
+ Main chatbot function that processes queries and generates responses.
+ Now yields chunks of the response for streaming.
+ """
try:
messages = json.load(open("Data/ChatLog.json"))
except FileNotFoundError:
@@ -129,44 +133,54 @@ def ChatBot(Query, sender_number=None):
# 🧠 Identity
if not is_owner and any(kw in lowered for kw in ["who am i", "amar naam", "ami ke", "আমার নাম"]):
- return "🤔 I don’t know your name yet!\nReply like this:\n`name: YourName`"
+ yield "� I don’t know your name yet!\nReply like this:\n`name: YourName`"
+ return # Exit generator after yielding
name = extract_name(Query)
if name and not is_owner:
guest_names[sender_number] = name
save_guest_names(guest_names)
- return f"✅ Got it! I’ll remember you as {name}."
+ yield f"✅ Got it! I’ll remember you as {name}."
+ return # Exit generator after yielding
if guest_names.get(sender_number) and lowered in ["hi", "hello", "hey"]:
- return f"👋 Welcome back, {guest_names[sender_number]}!"
+ yield f"👋 Welcome back, {guest_names[sender_number]}!"
+ return # Exit generator after yielding
# 🖼️ Image Generation Logic (NEW FEATURE)
if lowered.startswith("image "):
prompt = Query[len("image "):].strip()
- return generate_image_response(prompt)
+ # Image generation is a single, non-streamed response
+ # Yield the full response from generate_image_response
+ yield generate_image_response(prompt)
+ return
# 🔎 Google / YouTube Logic
if lowered.startswith("search ") or "search" in lowered:
- return search_google(Query.replace("search", "").strip())
+ # For search, we still return the full result at once, not stream it word by word
+ yield search_google(Query.replace("search", "").strip())
+ return
elif "youtube" in lowered:
- return search_youtube(Query.replace("youtube", "").strip())
+ # For YouTube, also return the full result at once
+ yield search_youtube(Query.replace("youtube", "").strip())
+ return
- # 🤖 LLM fallback
+ # 🤖 LLM fallback - now streams
messages.append({"role": "user", "content": Query})
- # Pass RealtimeInformation as a separate system message for dynamic updates
full_messages = SystemChatBot + [{"role": "system", "content": RealtimeInformation()}] + messages
try:
completion = client.chat.completions.create(
- model="llama3-70b-8192", messages=full_messages,
- max_tokens=1024, temperature=0.7
+ model="llama3-70b-8192",
+ messages=full_messages,
+ stream=True # <--- IMPORTANT: Enable streaming
)
- answer = completion.choices[0].message.content.strip()
- # Removed the platform-specific owner formatting
- messages.append({"role": "assistant", "content": answer})
- json.dump(messages, open("Data/ChatLog.json", "w"), indent=4)
- return answer
+ # Yield chunks as they arrive
+ for chunk in completion:
+ if chunk.choices[0].delta.content:
+ yield chunk.choices[0].delta.content # Yield each content chunk
except Exception as e:
- print(f"Error with LLM fallback: {e}")
- return "I'm sorry, I couldn't process that request with the AI. Please try again."
+ print(f"Error during LLM streaming: {e}")
+ yield f"I'm sorry, I encountered an error while processing your request: {e}"
+
