refactor(topics): 멀티 에이전트용 지식 재편 — _Common(공통 기본기) + Domain_* 구조
에이전트 8종(대화형/프로그래머 C·S/디자이너/설계자/기획자/QA/PD/PM)에게 [공통 기본 능력 + 롤별 Specialty] 2층으로 지식을 주입하기 위한 재분류. 문서 내용·포맷은 무수정, 폴더 이동만 (6,372개 문서 수 보존 확인). - Topic_Programming → Domain_Programming (내부 구조 보존) - Topic_Graphic → Domain_Design - Topic_Business → Domain_Product - Topic_General → Domain_General - _Common 신설: Math(구 Topic_Math_Specialty), Reasoning(구 General/From_Thinking & Reasoning), Reasoning_Creativity(구 General/From_창의성), Communication(Poetic_Blog_Writing + From_writing) - 타 도메인의 From_* 폴더는 유지 (출처 표기일 뿐, 이미 도메인에 맞게 분류된 문서) - 빈 폴더 정리 (memory/procedures) - 에이전트→폴더 매핑은 workspace의 .astra/agent-knowledge-map.json (9개 에이전트) Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
This commit is contained in:
Antigravity Agent
@@ -0,0 +1,208 @@
|
||||
---
|
||||
id: wiki-2026-0508-saas
|
||||
title: SaaS (Software as a Service)
|
||||
category: 10_Wiki/Topics
|
||||
status: verified
|
||||
canonical_id: self
|
||||
aliases: [Software as a Service, Cloud Software, Multi-tenant SaaS]
|
||||
duplicate_of: none
|
||||
source_trust_level: A
|
||||
confidence_score: 0.9
|
||||
verification_status: applied
|
||||
tags: [saas, multi-tenancy, subscription, ai-native, vertical-saas]
|
||||
raw_sources: []
|
||||
last_reinforced: 2026-05-10
|
||||
github_commit: pending
|
||||
tech_stack:
|
||||
language: TypeScript
|
||||
framework: Next.js / Stripe / Postgres RLS
|
||||
---
|
||||
|
||||
# SaaS (Software as a Service)
|
||||
|
||||
## 매 한 줄
|
||||
> **"매 multi-tenant, subscription-based, browser-delivered software — 매 install 의 X, 매 always-latest"**. 매 1999 Salesforce ("End of Software") 로 출발 → 매 2010s SaaS 1.0 (horizontal CRM/HR) → 매 2020s vertical SaaS (Toast, Procore, Veeva) → 매 2026 AI-native SaaS (Glean, Harvey, Cursor) 가 매 outcome-based pricing 으로 매 seat-based 모델 을 흔드는 시기.
|
||||
|
||||
## 매 핵심
|
||||
|
||||
### 매 Pillars
|
||||
- **Multi-tenancy**: tenant-isolated data on shared infra (RLS, schema-per-tenant, DB-per-tenant).
|
||||
- **Subscription**: MRR/ARR, Stripe billing, dunning, proration.
|
||||
- **Self-serve onboarding**: PLG funnel, time-to-value 분 단위.
|
||||
- **Continuous delivery**: weekly/daily ship, no version skew.
|
||||
- **Observability**: per-tenant SLO, usage analytics, churn signals.
|
||||
|
||||
### 매 SaaS metrics
|
||||
- **ARR/MRR**, **NRR** (Net Revenue Retention) — best signal of product fit.
|
||||
- **CAC payback** — months to recoup acquisition.
|
||||
- **Magic Number** = (ΔARR × 4) / S&M spend.
|
||||
- **Gross margin** — typical SaaS 70-85%, AI-SaaS 50-70% (inference cost).
|
||||
- **LTV/CAC** ≥ 3, payback ≤ 12 months.
|
||||
|
||||
### 매 Pricing models
|
||||
- **Per-seat**: classic Salesforce/Slack — saturating in AI era.
|
||||
- **Usage-based**: Snowflake, Twilio, OpenAI — aligns to value.
|
||||
- **Outcome/agent-based**: 2026 AI-native — pay per resolved ticket, per qualified lead.
|
||||
- **Hybrid**: platform fee + usage overage.
|
||||
|
||||
### 매 응용
|
||||
1. Horizontal SaaS (CRM, HRIS, ITSM, comms).
|
||||
2. Vertical SaaS (legal, dental, construction, hospitality).
|
||||
3. Developer SaaS (GitHub, Vercel, Sentry).
|
||||
4. AI-native SaaS (Cursor, Glean, Harvey, Decagon).
|
||||
5. Embedded SaaS (in-app commerce, fintech).
|
||||
|
||||
## 💻 패턴
|
||||
|
||||
### Multi-tenant Postgres RLS
|
||||
```sql
|
||||
ALTER TABLE documents ENABLE ROW LEVEL SECURITY;
|
||||
CREATE POLICY tenant_isolation ON documents
|
||||
USING (tenant_id = current_setting('app.tenant_id', true)::uuid);
|
||||
|
||||
-- App connection middleware
|
||||
SET LOCAL app.tenant_id = '7f3a...';
|
||||
```
|
||||
|
||||
### Stripe subscription with proration
|
||||
```typescript
|
||||
import Stripe from 'stripe'
|
||||
const stripe = new Stripe(process.env.STRIPE_SECRET!)
|
||||
|
||||
await stripe.subscriptions.update(subId, {
|
||||
items: [{ id: itemId, price: 'price_pro_monthly', quantity: 25 }],
|
||||
proration_behavior: 'create_prorations',
|
||||
billing_cycle_anchor: 'unchanged',
|
||||
})
|
||||
```
|
||||
|
||||
### Usage-based metering (Stripe meters, 2026)
|
||||
```typescript
|
||||
await stripe.billing.meterEvents.create({
|
||||
event_name: 'api_calls',
|
||||
payload: {
|
||||
stripe_customer_id: customer.id,
|
||||
value: '1',
|
||||
},
|
||||
})
|
||||
// Subscribe customer to metered price; Stripe aggregates and bills monthly
|
||||
```
|
||||
|
||||
### Tenant context middleware (Next.js)
|
||||
```typescript
|
||||
// app/middleware.ts
|
||||
import { NextResponse } from 'next/server'
|
||||
|
||||
export async function middleware(req: Request) {
|
||||
const session = await getSession(req)
|
||||
const tenantId = session?.tenantId
|
||||
if (!tenantId) return NextResponse.redirect('/login')
|
||||
const res = NextResponse.next()
|
||||
res.headers.set('x-tenant-id', tenantId)
|
||||
return res
|
||||
}
|
||||
```
|
||||
|
||||
### Per-tenant rate limit (Upstash Redis)
|
||||
```typescript
|
||||
import { Ratelimit } from '@upstash/ratelimit'
|
||||
import { Redis } from '@upstash/redis'
|
||||
|
||||
const limiters = new Map<string, Ratelimit>()
|
||||
function tenantLimiter(plan: 'free' | 'pro' | 'enterprise') {
|
||||
const limits = { free: 100, pro: 1000, enterprise: 10000 }
|
||||
return new Ratelimit({
|
||||
redis: Redis.fromEnv(),
|
||||
limiter: Ratelimit.slidingWindow(limits[plan], '1 m'),
|
||||
})
|
||||
}
|
||||
```
|
||||
|
||||
### PLG signup with magic-link (no password)
|
||||
```typescript
|
||||
// Send magic link, no password friction
|
||||
const token = jwt.sign({ email, tenantId: nanoid() }, SECRET, { expiresIn: '15m' })
|
||||
await sendEmail(email, `https://app.example.com/auth?t=${token}`)
|
||||
// On click: provision trial tenant, redirect to onboarding wizard
|
||||
```
|
||||
|
||||
### AI-native SaaS — usage-aware inference cost
|
||||
```typescript
|
||||
async function chat(tenantId: string, msgs: Message[]) {
|
||||
const tier = await getTier(tenantId)
|
||||
const model = tier === 'enterprise' ? 'claude-opus-4-7' : 'claude-haiku-4-5'
|
||||
const resp = await anthropic.messages.create({
|
||||
model, max_tokens: 1024, messages: msgs,
|
||||
metadata: { user_id: tenantId },
|
||||
})
|
||||
await meterUsage(tenantId, {
|
||||
input_tokens: resp.usage.input_tokens,
|
||||
output_tokens: resp.usage.output_tokens,
|
||||
model,
|
||||
})
|
||||
return resp
|
||||
}
|
||||
```
|
||||
|
||||
### Outcome pricing instrumentation
|
||||
```typescript
|
||||
// Charge only when AI agent successfully resolved
|
||||
async function recordResolution(tenantId: string, ticketId: string, resolved: boolean) {
|
||||
if (resolved) {
|
||||
await stripe.billing.meterEvents.create({
|
||||
event_name: 'resolved_ticket',
|
||||
payload: { stripe_customer_id: tenantId, value: '1' },
|
||||
})
|
||||
}
|
||||
}
|
||||
```
|
||||
|
||||
### Tenant-isolated S3 (per-prefix IAM)
|
||||
```python
|
||||
# Generate scoped STS token per tenant request
|
||||
sts = boto3.client("sts")
|
||||
policy = {"Version": "2012-10-17", "Statement": [{
|
||||
"Effect": "Allow",
|
||||
"Action": ["s3:GetObject", "s3:PutObject"],
|
||||
"Resource": [f"arn:aws:s3:::tenants-bucket/{tenant_id}/*"],
|
||||
}]}
|
||||
creds = sts.assume_role(
|
||||
RoleArn=ROLE, RoleSessionName=f"tenant-{tenant_id}",
|
||||
Policy=json.dumps(policy), DurationSeconds=900,
|
||||
)
|
||||
```
|
||||
|
||||
## 매 결정 기준
|
||||
| 상황 | Approach |
|
||||
|---|---|
|
||||
| Early-stage horizontal | Shared schema + RLS, Stripe per-seat |
|
||||
| Vertical w/ compliance | Schema-per-tenant or DB-per-tenant |
|
||||
| AI-native, variable cost | Usage-based + cap to protect margin |
|
||||
| Enterprise w/ SOC2/HIPAA | Single-tenant deploy option, BYOC |
|
||||
| PLG self-serve | Magic-link + provisioned trial in <60s |
|
||||
|
||||
**기본값**: shared-DB RLS + hybrid pricing (platform fee + usage) + Stripe meters + PLG onboarding.
|
||||
|
||||
## 🔗 Graph
|
||||
|
||||
## 🤖 LLM 활용
|
||||
**언제**: in-product copilot, customer support deflection, churn prediction from usage signals, content/email generation, dynamic onboarding.
|
||||
**언제 X**: pricing/billing computation — must be deterministic for audit and revenue recognition.
|
||||
|
||||
## ❌ 안티패턴
|
||||
- **No tenant isolation**: WHERE tenant_id checked only in app layer → IDOR breach.
|
||||
- **Per-seat pricing for AI**: high-usage user breaks margin; need usage cap or tier.
|
||||
- **Free tier without limits**: abuse → infra cost spirals.
|
||||
- **Single-region SaaS for global**: latency + data residency violations (GDPR).
|
||||
- **No self-serve cancel**: regulatory risk (FTC click-to-cancel 2024) + churn spikes.
|
||||
- **Version skew**: different customers on different versions → support combinatorics explode.
|
||||
|
||||
## 🧪 검증 / 중복
|
||||
- Verified (a16z SaaS metrics, OpenView PLG benchmarks, Stripe Billing docs, AWS SaaS Lens).
|
||||
- 신뢰도 A.
|
||||
|
||||
## 🕓 Changelog
|
||||
| 날짜 | 변경 |
|
||||
|---|---|
|
||||
| 2026-05-08 | Phase 1 |
|
||||
| 2026-05-10 | Manual cleanup — multi-tenancy, pricing models, AI-native SaaS 2026 |
|
||||
Reference in New Issue
Block a user