LLMs & Generative AI

1import anthropic
2
3client = anthropic.Anthropic(api_key="your-key")
4
5# Basic message
6message = client.messages.create(
7    model="claude-3-5-sonnet-20241022",
8    max_tokens=1024,
9    system="You are a DoD financial analyst AI.",
10    messages=[
11        {"role": "user", "content": "Summarize FY2025 defense priorities"}
12    ]
13)
14print(message.content[0].text)
15
16# Tool use / Function calling
17tools = [{
18    "name": "search_budget_data",
19    "description": "Search DoD budget database",
20    "input_schema": {
21        "type": "object",
22        "properties": {
23            "year": {"type": "integer", "description": "Fiscal year"},
24            "service": {"type": "string", "description": "Military service branch"}
25        },
26        "required": ["year"]
27    }
28}]
29
30response = client.messages.create(
31    model="claude-3-5-sonnet-20241022",
32    max_tokens=1024,
33    tools=tools,
34    messages=[{"role": "user", "content": "What was Army FY2024 budget?"}]
35)

1import { GoogleGenerativeAI } from '@google/genai';
2
3const genai = new GoogleGenerativeAI(process.env.GOOGLE_GEMINI_API_KEY!);
4
5// Function calling setup
6const tools = [{
7  functionDeclarations: [{
8    name: "get_budget_data",
9    description: "Retrieve DoD budget data",
10    parameters: {
11      type: "object",
12      properties: {
13        fiscal_year: { type: "string" },
14        component: { type: "string" }
15      }
16    }
17  }]
18}];
19
20async function runAgent(query: string) {
21  const model = genai.getGenerativeModel({
22    model: "gemini-2.0-flash",
23    tools,
24    systemInstruction: "You are a federal budget analyst AI."
25  });
26
27  const chat = model.startChat();
28  const result = await chat.sendMessage(query);
29  
30  // Handle function calls
31  const response = result.response;
32  const parts = response.candidates?.[0]?.content?.parts || [];
33  
34  for (const part of parts) {
35    if (part.functionCall) {
36      console.log("Tool called:", part.functionCall.name);
37      // Execute the function...
38    }
39  }
40  
41  return response.text();
42}

1# Multi-Agent System (MyThing Platform Pattern)
2from anthropic import Anthropic
3
4client = Anthropic()
5
6def route_query(query: str) -> str:
7    """Determine which specialized agent to use"""
8    routing = client.messages.create(
9        model="claude-3-5-haiku-20241022",
10        max_tokens=50,
11        system="Classify the query. Respond with ONLY one word: portfolio, tech, dod, or notes",
12        messages=[{"role": "user", "content": query}]
13    )
14    return routing.content[0].text.strip().lower()
15
16AGENT_CONFIGS = {
17    "portfolio": "You are a portfolio analyst for Peter Shang's projects...",
18    "tech": "You are a tech trends analyst with expertise in AI/ML...",
19    "dod": "You are a DoD policy and federal finance expert...",
20    "notes": "You help retrieve and summarize personal notes...",
21}
22
23def run_agent(query: str) -> str:
24    agent_type = route_query(query)
25    system_prompt = AGENT_CONFIGS.get(agent_type, AGENT_CONFIGS["tech"])
26    
27    response = client.messages.create(
28        model="claude-3-5-sonnet-20241022",
29        max_tokens=1024,
30        system=system_prompt,
31        messages=[{"role": "user", "content": query}]
32    )
33    
34    return f"[{agent_type.upper()} Agent] {response.content[0].text}"
35
36# Usage
37print(run_agent("What AI projects has Peter built?"))  # → portfolio
38print(run_agent("Latest trends in agentic AI?"))        # → tech
39print(run_agent("Explain FIAR audit requirements"))     # → dod