llm-application-dev

1# LLM Application Development
2 
3## Prompt Engineering
4 
5### Structured Prompts
6```typescript
7const systemPrompt = `You are a helpful assistant that answers questions about our product.
8 
9RULES:
10- Only answer questions about our product
11- If you don't know, say "I don't know"
12- Keep responses concise (under 100 words)
13- Never make up information
14 
15CONTEXT:
16{context}`;
17 
18const userPrompt = `Question: {question}`;
19```
20 
21### Few-Shot Examples
22```typescript
23const prompt = `Classify the sentiment of customer feedback.
24 
25Examples:
26Input: "Love this product!"
27Output: positive
28 
29Input: "Worst purchase ever"
30Output: negative
31 
32Input: "It works fine"
33Output: neutral
34 
35Input: "${customerFeedback}"
36Output:`;
37```
38 
39### Chain of Thought
40```typescript
41const prompt = `Solve this step by step:
42 
43Question: ${question}
44 
45Let's think through this:
461. First, identify the key information
472. Then, determine the approach
483. Finally, calculate the answer
49 
50Step-by-step solution:`;
51```
52 
53## API Integration
54 
55### OpenAI Pattern
56```typescript
57import OpenAI from 'openai';
58 
59const openai = new OpenAI({ apiKey: process.env.OPENAI_API_KEY });
60 
61async function chat(messages: Message[]): Promise<string> {
62  const response = await openai.chat.completions.create({
63    model: 'gpt-4',
64    messages,
65    temperature: 0.7,
66    max_tokens: 500,
67  });
68 
69  return response.choices[0].message.content ?? '';
70}
71```
72 
73### Anthropic Pattern
74```typescript
75import Anthropic from '@anthropic-ai/sdk';
76…