Arka Plan Bilgileri Sağlama

Arka Plan Bilgisi Neden Önemlidir

Alan Bilgisi Ne Zaman Eklenmeli

import anthropic

client = anthropic.Anthropic(api_key='sk-ant-your-key-here')

domain_knowledge = (
    'Domain: algorithmic trading at a regulated broker-dealer in the US. '
    'Relevant rules: SEC Rule 15c3-5 (Market Access Rule) requires pre-trade risk checks. '
    'Latency budget: all risk checks must complete in under 100 microseconds. '
    'Stack: C++ for the matching engine, Python for strategy logic.'
)

response = client.messages.create(
    model='claude-opus-4-5',
    max_tokens=300,
    messages=[{
        'role': 'user',
        'content': (
            f'{domain_knowledge}\n\n'
            f'Question: How should we implement position limit checks for our order router?'
        )
    }]
)
print(response.content[0].text)

Proje Ayrıntılarını Dahil Etme

import openai

client = openai.OpenAI(api_key='sk-your-key-here')

project_context = (
    'Project: a mobile recipe app (React Native, Node.js backend, MongoDB). '
    'Stage: pre-launch, 2 weeks to go live. '
    'Team: 2 frontend devs, 1 backend dev, no dedicated QA. '
    'Current issue: app search is slow (3-5 seconds) with 50,000 recipes in the DB. '
    'Constraint: no time for a full re-architecture before launch.'
)

response = client.chat.completions.create(
    model='gpt-4o',
    messages=[{
        'role': 'user',
        'content': (
            f'{project_context}\n\n'
            f'Suggest 3 quick wins to improve search speed before launch. '
            f'Each suggestion: what to do (1 sentence), estimated effort (hours), '
            f'expected improvement.'
        )
    }]
)
print(response.choices[0].message.content)

Mevcut Çalışmaları Paylaşma

import anthropic

client = anthropic.Anthropic(api_key='sk-ant-your-key-here')

existing_intro = (
    'At Northlight Labs, we build AI tools that work the way humans think — '
    'not the other way around. Founded in 2022 by two ex-Google researchers, '
    'we have helped 500 companies ship smarter products without hiring ML teams.'
)

response = client.messages.create(
    model='claude-opus-4-5',
    max_tokens=200,
    messages=[{
        'role': 'user',
        'content': (
            'Here is our existing company intro paragraph:\n\n'
            f'{existing_intro}\n\n'
            'Write a second paragraph (matching this exact tone and style) '
            'that describes our flagship product: an API that lets developers '
            'add natural language search to any database with 3 lines of code. '
            'Keep it under 60 words.'
        )
    }]
)
print(response.content[0].text)

Kısıtları Baştan Belirtme

import openai

client = openai.OpenAI(api_key='sk-your-key-here')

response = client.chat.completions.create(
    model='gpt-4o',
    messages=[{
        'role': 'user',
        'content': (
            'We need to add rate limiting to our REST API.\n\n'
            'Constraints:\n'
            '- Python 3.11 FastAPI application\n'
            '- No external dependencies beyond what is already installed (requests, pydantic, uvicorn)\n'
            '- Must work as a single process (no Redis, no shared state between workers)\n'
            '- Rate limit: 100 requests per minute per IP address\n'
            '- If limit exceeded, return HTTP 429 with Retry-After header\n\n'
            'Show me a complete implementation.'
        )
    }]
)
print(response.choices[0].message.content)

Paydaş Bilgileri

import anthropic

client = anthropic.Anthropic(api_key='sk-ant-your-key-here')

stakeholder_context = (
    'This proposal will be reviewed by:\n'
    '1. CTO: cares about technical risk and maintainability, skeptical of microservices\n'
    '2. CFO: cares about cost, wants to see clear ROI numbers\n'
    '3. VP Engineering: cares about developer experience and team morale\n'
    'The CTO has veto power.'
)

response = client.messages.create(
    model='claude-opus-4-5',
    max_tokens=400,
    messages=[{
        'role': 'user',
        'content': (
            f'{stakeholder_context}\n\n'
            f'Write a 3-paragraph executive summary for a proposal to migrate '
            f'our monolith to a modular monolith (not microservices). '
            f'Address each stakeholder\'s top concern in a separate paragraph.'
        )
    }]
)
print(response.content[0].text)

Kronolojik Yöntem

import openai

client = openai.OpenAI(api_key='sk-your-key-here')

chronological_background = (
    'Timeline:\n'
    '- 6 months ago: We launched v1 of our API with a flat pricing model ($99/month).\n'
    '- 3 months ago: Power users (top 10%) started consuming 80% of compute.\n'
    '- 1 month ago: We introduced usage-based pricing — power users complained loudly.\n'
    '- Current: 12% of power users have churned; new signups are up 15%.\n'
    '- Goal: Keep power users without losing the new-signup growth.\n'
)

response = client.chat.completions.create(
    model='gpt-4o',
    messages=[{
        'role': 'user',
        'content': (
            f'{chronological_background}\n\n'
            f'Suggest 3 pricing model adjustments that could address both groups. '
            f'For each: describe the change, who it benefits, and the revenue risk.'
        )
    }]
)
print(response.choices[0].message.content)

Kendi Verilerinizi Arka Plan Olarak Kullanma

import anthropic

client = anthropic.Anthropic(api_key='sk-ant-your-key-here')

metrics_data = (
    'Weekly active users (WAU) last 8 weeks:\n'
    'Week 1: 1,200  |  Week 2: 1,350  |  Week 3: 1,290  |  Week 4: 1,400\n'
    'Week 5: 1,380  |  Week 6: 1,220  |  Week 7: 1,100  |  Week 8: 980\n'
    'Note: We launched a pricing change in Week 5.'
)

response = client.messages.create(
    model='claude-opus-4-5',
    max_tokens=300,
    messages=[{
        'role': 'user',
        'content': (
            f'{metrics_data}\n\n'
            f'Analyze the trend, identify where decline began, '
            f'and suggest 2 hypotheses for the cause. '
            f'Be specific — reference the exact weeks in your analysis.'
        )
    }]
)
print(response.content[0].text)

Arka Plan Olarak Dahil Edilmeyecekler (NOT)

# Contrast: bloated vs lean background
bloated_background = (
    'Our company was founded in 2018 by Alice and Bob. '
    'We have 45 employees across 3 offices. We use Jira for project management. '
    'We had a product refresh in 2020. Our CTO came from Amazon. '
    'We use Slack, Notion, and Linear. '
    'We once tried Kubernetes but moved back to Docker Compose. '
    'Question: How should we handle database connection pooling in FastAPI?'
)

lean_background = (
    'Stack: FastAPI, PostgreSQL, deployed on a single 8-core server. '
    'Load: ~200 concurrent users peak. '
    'Current issue: seeing "too many connections" errors under load. '
    'Question: How should we handle database connection pooling?'
)

print('Bloated background:', len(bloated_background.split()), 'words')
print('Lean background:', len(lean_background.split()), 'words')
print('Reduction:', round((1 - len(lean_background.split())/len(bloated_background.split()))*100), '%')

Arka Planı Okunabilirlik İçin Biçimlendirme

import openai

client = openai.OpenAI(api_key='sk-your-key-here')

# Well-formatted background
formatted_background = (
    '=== BACKGROUND ===\n'
    'Project: real-time chat API\n'
    'Stack: Python FastAPI, WebSockets, Redis pub/sub, PostgreSQL\n'
    'Scale: 10,000 concurrent connections target\n'
    'Constraints: single server for now, no Kubernetes\n'
    'Current problem: messages are occasionally delivered out of order\n'
    '=== TASK ===\n'
    'Explain the root cause of out-of-order message delivery in WebSocket pub/sub '
    'and suggest the simplest fix for our stack.'
)

response = client.chat.completions.create(
    model='gpt-4o',
    messages=[{'role': 'user', 'content': formatted_background}]
)
print(response.choices[0].message.content)

Sistem Mesajında Yeniden Kullanılabilir Arka Plan

import anthropic

client = anthropic.Anthropic(api_key='sk-ant-your-key-here')

# System message carries reusable background for every user turn
system_background = (
    'You are an AI assistant embedded in Northlight Labs\' internal developer portal. '
    'Background: our stack is Python 3.11, FastAPI, PostgreSQL 15, Redis 7, Docker. '
    'We follow Google Python style guide. All code examples must include type hints. '
    'We do not use ORM — raw SQL with psycopg3. '
    'Our API follows REST conventions with snake_case JSON fields.'
)

# Now every user turn gets this background automatically
response = client.messages.create(
    model='claude-opus-4-5',
    max_tokens=300,
    system=system_background,
    messages=[
        {'role': 'user', 'content': 'Show me how to insert a row into the users table.'}
    ]
)
print(response.content[0].text)

Bilgi Kontrolü

Arka Plan Bilgisi Sağlama — Özet