Working with LLM APIs

1# Load environment variables
2import os
3from dotenv import load_dotenv
4
5load_dotenv()
6openai_key = os.getenv("OPENAI_API_KEY")

1from openai import OpenAI
2
3client = OpenAI()  # Reads OPENAI_API_KEY from env
4
5response = client.chat.completions.create(
6    model="gpt-4-turbo",
7    messages=[
8        {"role": "system", "content": "Bạn là AI assistant hữu ích."},
9        {"role": "user", "content": "Giải thích Machine Learning trong 3 câu."}
10    ],
11    temperature=0.7,
12    max_tokens=500
13)
14
15print(response.choices[0].message.content)
16print(f"Tokens used: {response.usage.total_tokens}")

1stream = client.chat.completions.create(
2    model="gpt-4-turbo",
3    messages=[{"role": "user", "content": "Viết bài thơ về Hà Nội"}],
4    stream=True
5)
6
7for chunk in stream:
8    content = chunk.choices[0].delta.content
9    if content:
10        print(content, end="", flush=True)

1messages = [
2    {"role": "system", "content": "Bạn là Python tutor."}
3]
4
5def chat(user_message):
6    messages.append({"role": "user", "content": user_message})
7    
8    response = client.chat.completions.create(
9        model="gpt-4-turbo",
10        messages=messages,
11        temperature=0.7
12    )
13    
14    assistant_msg = response.choices[0].message.content
15    messages.append({"role": "assistant", "content": assistant_msg})
16    return assistant_msg
17
18# Conversation
19print(chat("List là gì trong Python?"))
20print(chat("Cho ví dụ với list comprehension"))
21print(chat("So sánh list và tuple"))

1import anthropic
2
3client = anthropic.Anthropic()  # Reads ANTHROPIC_API_KEY
4
5message = client.messages.create(
6    model="claude-3-5-sonnet-20241022",
7    max_tokens=1024,
8    system="Bạn là data analyst expert.",
9    messages=[
10        {"role": "user", "content": "Phân tích xu hướng e-commerce Việt Nam 2026"}
11    ]
12)
13
14print(message.content[0].text)
15print(f"Input tokens: {message.usage.input_tokens}")
16print(f"Output tokens: {message.usage.output_tokens}")

1with client.messages.stream(
2    model="claude-3-5-sonnet-20241022",
3    max_tokens=1024,
4    messages=[{"role": "user", "content": "Viết code Python sort algorithms"}]
5) as stream:
6    for text in stream.text_stream:
7        print(text, end="", flush=True)

1import google.generativeai as genai
2
3genai.configure(api_key=os.getenv("GOOGLE_API_KEY"))
4model = genai.GenerativeModel("gemini-1.5-pro")
5
6response = model.generate_content("Giải thích Docker trong 5 câu")
7print(response.text)

1import PIL.Image
2
3model = genai.GenerativeModel("gemini-1.5-pro")
4image = PIL.Image.open("chart.png")
5
6response = model.generate_content([
7    "Phân tích biểu đồ này và nêu 3 insights chính:",
8    image
9])
10print(response.text)

1from openai import OpenAI, RateLimitError, APIError
2
3client = OpenAI()
4
5def safe_call(prompt, retries=3):
6    for attempt in range(retries):
7        try:
8            response = client.chat.completions.create(
9                model="gpt-4-turbo",
10                messages=[{"role": "user", "content": prompt}],
11                timeout=30
12            )
13            return response.choices[0].message.content
14        
15        except RateLimitError:
16            print(f"Rate limited. Waiting {2**attempt}s...")
17            import time
18            time.sleep(2 ** attempt)
19        
20        except APIError as e:
21            print(f"API error: {e}")
22            if attempt == retries - 1:
23                raise
24    
25    return None

1import tiktoken
2
3def count_tokens(text, model="gpt-4"):
4    enc = tiktoken.encoding_for_model(model)
5    return len(enc.encode(text))
6
7# Example
8prompt = "Hãy phân tích dữ liệu sales Q4 2026..."
9tokens = count_tokens(prompt)
10cost = tokens * 0.03 / 1000
11print(f"Tokens: {tokens}, Est cost: ${cost:.4f}")

1# Gửi cùng prompt cho cả 3 providers, so sánh:
2prompt = "Viết hàm Python tìm số Fibonacci thứ n. Giải thích time complexity."
3
4# 1. Call OpenAI
5# 2. Call Anthropic
6# 3. Call Gemini
7# 4. Compare: quality, speed, tokens used

1class LLMClient:
2    def __init__(self, provider="openai"):
3        self.provider = provider
4        # Initialize correct client
5    
6    def chat(self, prompt, system=None, **kwargs):
7        # Route to correct API
8        # Return standardized response
9        pass
10    
11    def stream(self, prompt, system=None, **kwargs):
12        # Streaming for all providers
13        pass
14
15# Usage
16llm = LLMClient("anthropic")
17response = llm.chat("Hello!")