codethinker / app.py
nananie143's picture
Create app.py
3f46926 verified
import gradio as gr
import torch
from transformers import AutoModelForCausalLM, AutoTokenizer
from fastapi import FastAPI, HTTPException, Request
from pydantic import BaseModel
import uvicorn
from typing import List, Dict, Optional
from collections import defaultdict
from queue import PriorityQueue
import random
# Load the model and tokenizer
MODEL_NAME = "unit-mesh/autodev-coder-deepseek-6.7b-finetunes"
tokenizer = AutoTokenizer.from_pretrained(MODEL_NAME)
model = AutoModelForCausalLM.from_pretrained(MODEL_NAME, torch_dtype=torch.float16, device_map="auto")
# Custom CSS for OpenWebUI-like design
custom_css = """
#chatbot {
font-family: Arial, sans-serif;
max-width: 800px;
margin: auto;
padding: 20px;
border-radius: 10px;
box-shadow: 0 4px 6px rgba(0, 0, 0, 0.1);
}
#sidebar {
background-color: #f5f5f5;
padding: 20px;
border-radius: 10px;
}
.message.user {
background-color: #007bff;
color: white;
border-radius: 10px 10px 0 10px;
padding: 10px;
margin: 5px 0;
max-width: 70%;
margin-left: auto;
}
.message.bot {
background-color: #e9ecef;
color: black;
border-radius: 10px 10px 10px 0;
padding: 10px;
margin: 5px 0;
max-width: 70%;
margin-right: auto;
}
.dark-mode #chatbot {
background-color: #2d2d2d;
color: #ffffff;
}
.dark-mode #sidebar {
background-color: #1e1e1e;
color: #ffffff;
}
.dark-mode .message.user {
background-color: #0056b3;
}
.dark-mode .message.bot {
background-color: #3d3d3d;
color: #ffffff;
}
"""
# Enhanced Reasoning Algorithms
class DeductiveReasoner:
def __init__(self, rules: Dict[str, str]):
self.rules = rules
def infer(self, premise: str, specific_case: str) -> str:
for condition, conclusion in self.rules.items():
if condition in specific_case:
return f"Given the premise '{premise}' and the specific case '{specific_case}', the conclusion is: {conclusion}"
return f"Given the premise '{premise}', no applicable rule was found for the specific case '{specific_case}'."
class InductiveReasoner:
def __init__(self):
self.patterns = defaultdict(int)
def learn(self, examples: List[str]):
for example in examples:
words = example.split()
for i in range(len(words) - 1):
self.patterns[(words[i], words[i + 1])] += 1
def infer(self) -> str:
if not self.patterns:
return "No patterns have been learned yet."
most_common_pattern = max(self.patterns, key=self.patterns.get)
return f"From the learned examples, the most common pattern is: '{most_common_pattern[0]} {most_common_pattern[1]}'."
class AbductiveReasoner:
def __init__(self, hypotheses: Dict[str, float]):
self.hypotheses = hypotheses
def evaluate(self, observation: str, likelihoods: Dict[str, float]) -> str:
posterior = {
hypothesis: prior * likelihoods.get(hypothesis, 0.0)
for hypothesis, prior in self.hypotheses.items()
}
best_hypothesis = max(posterior, key=posterior.get)
return f"Given the observation '{observation}', the most plausible explanation is: {best_hypothesis} (posterior probability: {posterior[best_hypothesis]:.2f})."
class BayesianReasoner:
def __init__(self, prior: float):
self.prior = prior
def update(self, evidence: str, likelihood: float) -> str:
posterior = self.prior * likelihood
self.prior = posterior # Update the prior for future reasoning
return f"Given the evidence '{evidence}', the updated probability is: {posterior:.2f}."
class HeuristicSearcher:
def __init__(self, heuristic_func):
self.heuristic_func = heuristic_func
def search(self, start, goal):
frontier = PriorityQueue()
frontier.put((0, start))
came_from = {}
cost_so_far = {}
came_from[start] = None
cost_so_far[start] = 0
while not frontier.empty():
_, current = frontier.get()
if current == goal:
break
for next_state in self.get_neighbors(current):
new_cost = cost_so_far[current] + 1 # Assume uniform cost
if next_state not in cost_so_far or new_cost < cost_so_far[next_state]:
cost_so_far[next_state] = new_cost
priority = new_cost + self.heuristic_func(next_state, goal)
frontier.put((priority, next_state))
came_from[next_state] = current
return f"Best solution found from {start} to {goal}."
def get_neighbors(self, state):
# Example: For a numeric state, return neighboring states
return [state - 1, state + 1]
# Initialize reasoning algorithms
deductive_reasoner = DeductiveReasoner(
rules={
"error": "Check for syntax errors in the code.",
"loop": "Optimize the loop structure for better performance.",
"null": "Ensure proper null checks are in place.",
}
)
inductive_reasoner = InductiveReasoner()
inductive_reasoner.learn(["If it rains, the ground gets wet.", "If you study, you pass the exam."])
abductive_reasoner = AbductiveReasoner(
hypotheses={"syntax error": 0.3, "logical error": 0.5, "runtime error": 0.2}
)
bayesian_reasoner = BayesianReasoner(prior=0.5)
heuristic_searcher = HeuristicSearcher(heuristic_func=lambda state, goal: abs(state - goal))
# Chatbot function with reasoning enhancements
def chatbot_response(message, history, reasoning_algorithm, file_content=None):
history = history or []
reasoning = {
"Deductive": deductive_reasoner.infer("General rule", message),
"Inductive": inductive_reasoner.infer(),
"Abductive": abductive_reasoner.evaluate(message, {"syntax error": 0.8, "logical error": 0.5}),
"Bayesian": bayesian_reasoner.update(message, likelihood=0.7),
"Heuristic": heuristic_searcher.search(start=0, goal=10),
}.get(reasoning_algorithm, "Invalid reasoning algorithm.")
# Append file content if provided
if file_content:
reasoning += f"\n\nFile Content:\n{file_content}"
history.append((message, reasoning))
return history, history
# File upload handler
def handle_file_upload(file):
if file:
with open(file.name, "r") as f:
content = f.read()
return content
return None
# Theme toggling
def toggle_theme(theme):
if theme == "Dark":
return gr.update(css=custom_css + ".dark-mode")
else:
return gr.update(css=custom_css)
# Gradio interface
with gr.Blocks(css=custom_css) as demo:
gr.Markdown("# OpenWebUI-like Chat Interface with Reasoning Enhancements")
with gr.Row():
with gr.Column(scale=1, elem_id="sidebar"):
gr.Markdown("### Settings")
model_selector = gr.Dropdown(["Model 1", "Model 2"], label="Select Model")
reasoning_selector = gr.Dropdown(
["Deductive", "Inductive", "Abductive", "Bayesian", "Heuristic"],
label="Select Reasoning Algorithm",
value="Deductive",
)
theme_selector = gr.Radio(["Light", "Dark"], label="Theme", value="Light")
file_upload = gr.File(label="Upload File")
with gr.Column(scale=3, elem_id="chatbot"):
chatbot = gr.Chatbot(label="Chat")
message = gr.Textbox(label="Your Message", placeholder="Type your message here...")
submit = gr.Button("Send")
state = gr.State()
# Chat interaction
submit.click(
chatbot_response,
inputs=[message, state, reasoning_selector, file_upload],
outputs=[chatbot, state],
)
# File upload handling
file_upload.change(
handle_file_upload,
inputs=file_upload,
outputs=message,
)
# Theme toggling
theme_selector.change(
toggle_theme,
inputs=theme_selector,
outputs=None,
)
# OpenAI-compatible API using FastAPI
app = FastAPI()
class ChatCompletionRequest(BaseModel):
model: str
messages: List[dict]
max_tokens: Optional[int] = 500
temperature: Optional[float] = 0.7
class ChatCompletionResponse(BaseModel):
id: str
object: str = "chat.completion"
created: int
model: str
choices: List[dict]
usage: dict
@app.post("/v1/chat/completions")
async def chat_completions(request: ChatCompletionRequest):
try:
# Extract the last user message
user_message = request.messages[-1]["content"]
# Generate a response using the model
inputs = tokenizer(user_message, return_tensors="pt").to(model.device)
outputs = model.generate(**inputs, max_length=request.max_tokens, temperature=request.temperature)
response_text = tokenizer.decode(outputs[0], skip_special_tokens=True)
# Format the response in OpenAI-compatible format
response = ChatCompletionResponse(
id="chatcmpl-12345",
created=int(torch.tensor(0)), # Placeholder for timestamp
model=request.model,
choices=[
{
"message": {
"role": "assistant",
"content": response_text,
},
"finish_reason": "stop",
"index": 0,
}
],
usage={
"prompt_tokens": len(tokenizer.encode(user_message)),
"completion_tokens": len(tokenizer.encode(response_text)),
"total_tokens": len(tokenizer.encode(user_message)) + len(tokenizer.encode(response_text)),
},
)
return response
except Exception as e:
raise HTTPException(status_code=500, detail=str(e))
# Run the FastAPI server
def run_api():
uvicorn.run(app, host="0.0.0.0", port=8000)
# Run the Gradio app
def run_gradio():
demo.launch(server_name="0.0.0.0", server_port=7860)
# Entry point
if __name__ == "__main__":
import threading
# Start the FastAPI server in a separate thread
api_thread = threading.Thread(target=run_api)
api_thread.start()
# Start the Gradio app
run_gradio()