Spaces:
Sleeping
Sleeping
File size: 8,903 Bytes
1c890c3 | 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203 204 205 206 207 208 209 210 211 212 213 214 215 216 217 218 219 220 221 222 223 224 225 226 227 228 229 230 231 232 233 234 235 236 237 238 239 240 241 242 243 244 245 246 247 248 249 250 251 252 253 254 255 256 257 258 259 260 261 262 263 264 265 | import os
import time
import nltk
import numpy as np
import random
from flask import Flask, request, render_template, jsonify, Response
from flasgger import Swagger, swag_from
from transformers import AutoTokenizer, AutoModelForCausalLM, pipeline
from huggingface_hub import login
from langchain_community.tools import DuckDuckGoSearchRun
from nltk import word_tokenize, pos_tag
from diffusers import StableDiffusionPipeline, SD3Transformer2DModel, FlowMatchEulerDiscreteScheduler, EulerDiscreteScheduler
from io import BytesIO
import base64
# β
Safe import of GPU decorator
try:
from spaces import GPU
except ImportError:
def GPU(func): return func
import torch
# Download required NLTK data (ensure it runs once)
try:
nltk.data.find('tokenizers/punkt')
except LookupError:
nltk.download('punkt')
nltk.download('punkt_tab')
nltk.download("averaged_perceptron_tagger")
nltk.download('averaged_perceptron_tagger_eng')
# Flask + Swagger setup
app = Flask(__name__, static_folder="static", template_folder="templates")
swagger = Swagger(app, template={
"swagger": "2.0",
"info": {
"title": "ChatMate Real-Time API",
"description": "LangChain + DuckDuckGo enabled AI chatbot",
"version": "1.0"
}
}, config={
"headers": [],
"specs": [{"endpoint": 'apispec', "route": '/apispec.json', "rule_filter": lambda rule: True}],
"static_url_path": "/flasgger_static",
"swagger_ui": True,
"specs_route": "/apidocs/"
})
# β
Hugging Face login (if token provided)
login(token=os.environ.get("CHAT_MATE"))
def load_image_pipeline():
model_id = "CompVis/stable-diffusion-v1-4"
device = "cuda" if torch.cuda.is_available() else "cpu"
dtype = torch.float16 if device == "cuda" else torch.float32
pipe = StableDiffusionPipeline.from_pretrained(
model_id,
torch_dtype=dtype,
use_safetensors=True,
safety_checker=None # disable safety checker to avoid blocking
)
pipe.scheduler = EulerDiscreteScheduler.from_config(pipe.scheduler.config)
pipe = pipe.to(device)
pipe.enable_attention_slicing()
return pipe
image_pipe = load_image_pipeline()
def generate_image_base64(prompt):
if not prompt.strip():
prompt = "A scenic view of mountains during sunrise"
try:
print(f"π¨ Generating image for prompt: {prompt}")
generator = torch.Generator(device=image_pipe.device).manual_seed(random.randint(0, 2**31 - 1))
result = image_pipe(
prompt,
height=512,
width=512,
guidance_scale=7.5,
num_inference_steps=40,
generator=generator
)
image = result.images[0]
buffer = BytesIO()
image.save(buffer, format="PNG")
img_b64 = base64.b64encode(buffer.getvalue()).decode("utf-8")
print("β
Image generated successfully.", img_b64)
return img_b64
except Exception as e:
print(f"β οΈ Image generation failed: {e}")
return ""
# β
Prompt checker
IMAGE_PROMPT_KEYWORDS = {"generate image", "create image", "make image", "draw", "illustrate", "paint", "show me", "visualize"}
def is_image_prompt(text):
return any(kw in text.lower() for kw in IMAGE_PROMPT_KEYWORDS)
# β
Load LLaMA 3.1 Instruct model
#model_id = "meta-llama/Llama-3.1-8B-Instruct"
model_id = "mistralai/Mixtral-8x7B-Instruct-v0.1"
tokenizer = AutoTokenizer.from_pretrained(model_id)
model = AutoModelForCausalLM.from_pretrained(model_id, device_map="auto",load_in_4bit=True,
torch_dtype=torch.float16,
trust_remote_code=True)
pipe = pipeline("text-generation", model=model, tokenizer=tokenizer, max_new_tokens=512)
# β
Simple keyword-based check for real-time info
REAL_TIME_KEYWORDS = {"latest", "current", "news", "today", "price", "time", "live", "trending", "update", "happening"}
def should_search(message):
message = message.lower()
return any(kw in message for kw in REAL_TIME_KEYWORDS)
def should_search_intent_nltk(message):
question_starters = {"what", "when", "where", "who", "how", "which", "why"}
tokens = word_tokenize(message.lower())
tags = pos_tag(tokens)
starts_with_question = tokens[0] in question_starters
has_nouns_or_verbs = any(tag.startswith("NN") or tag.startswith("VB") for _, tag in tags)
return starts_with_question and has_nouns_or_verbs
# β
Search tool
search_tool = DuckDuckGoSearchRun()
# β
Chat using model with chat template and history
@GPU
def generate_full_reply(message, history):
system_prompt = "You are a helpful AI assistant."
messages = [{"role": "system", "content": system_prompt}] + history + [{"role": "user", "content": message}]
prompt = tokenizer.apply_chat_template(messages, tokenize=False, add_generation_prompt=True)
output = pipe(prompt, do_sample=True, temperature=0.7)[0]["generated_text"]
#print(output)
return output.split(prompt)[-1].strip()
# β
Flask route
@app.route("/")
def home():
return render_template("index.html")
@app.route("/chat", methods=["POST"])
@swag_from({
'tags': ['Chat'],
'consumes': ['application/json'],
'summary': 'Get assistant reply',
'description': 'Send a message and chat history, and receive a full AI-generated response.',
'parameters': [{
'name': 'body',
'in': 'body',
'required': True,
'schema': {
'type': 'object',
'properties': {
'message': {'type': 'string', 'example': 'What is Python?'},
'history': {
'type': 'array',
'items': {
'type': 'object',
'properties': {
'role': {'type': 'string', 'example': 'user'},
'content': {'type': 'string', 'example': 'Tell me about Python'}
}
}
}
},
'required': ['message']
}
}],
'responses': {
200: {
'description': 'Assistant reply',
'schema': {
'type': 'object',
'properties': {
'reply': {'type': 'string'}
}
}
}
}
})
def chat():
data = request.get_json()
message = data.get("message")
history = data.get("history", [])
# Check if real-time search is needed
if should_search(message) or should_search_intent_nltk(message):
result = f"(Live info) {search_tool.run(message)}"
else:
result = generate_full_reply(message, history)
return jsonify({"reply": result})
@app.route("/chat-stream", methods=["POST"])
@swag_from({
'tags': ['Chat'],
'consumes': ['application/json'],
'summary': 'Stream assistant reply or generate image',
'description': 'Respond with AI-generated text or image based on prompt intent.',
'parameters': [{
'name': 'body',
'in': 'body',
'required': True,
'schema': {
'type': 'object',
'properties': {
'message': {'type': 'string', 'example': 'Draw a cyberpunk city at night'},
'history': {
'type': 'array',
'items': {
'type': 'object',
'properties': {
'role': {'type': 'string', 'example': 'user'},
'content': {'type': 'string', 'example': 'What is entanglement?'}
}
}
}
},
'required': ['message']
}
}],
'responses': {
200: {
'description': 'Streamed text or image',
'content': {'text/plain': {}}
}
}
})
def chat_stream():
data = request.get_json()
message = data.get("message")
history = data.get("history", [])
def generate():
if is_image_prompt(message):
yield "[πΌοΈ Generating image...]\n"
image_b64 = generate_image_base64(message)
yield f"[BASE64_IMAGE_START]{image_b64}[BASE64_IMAGE_END]\n"
elif should_search(message) or should_search_intent_nltk(message):
reply = f"(Live info) {search_tool.run(message)}"
for token in reply.splitlines(keepends=True):
yield token
time.sleep(0.05)
else:
reply = generate_full_reply(message, history)
for token in reply.splitlines(keepends=True):
yield token
time.sleep(0.05)
return Response(generate(), mimetype='text/plain')
# β
Warm-up on startup
if __name__ == "__main__":
print("π§ Warming up...")
_ = generate_full_reply("Hello", [])
app.run(host="0.0.0.0", port=int(os.environ.get("PORT", 7860))) |