feat(ultraplan-local): v1.6.0 — /ultraresearch-local deep research command
Add /ultraresearch-local for structured research combining local codebase analysis with external knowledge via parallel agent swarms. Produces research briefs with triangulation, confidence ratings, and source quality assessment. New command: /ultraresearch-local with modes --quick, --local, --external, --fg. New agents: research-orchestrator (opus), docs-researcher, community-researcher, security-researcher, contrarian-researcher, gemini-bridge (all sonnet). New template: research-brief-template.md. Integration: --research flag in /ultraplan-local accepts pre-built research briefs (up to 3), enriches the interview and exploration phases. Planning orchestrator cross-references brief findings during synthesis. Design principle: Context Engineering — right information to right agent at right time. Research briefs are structured artifacts in the pipeline: ultraresearch → brief → ultraplan --research → plan → ultraexecute. Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
This commit is contained in:
commit
baa2d0220b
488 changed files with 213221 additions and 0 deletions
|
|
@ -0,0 +1,426 @@
|
|||
# APIM with Azure Front Door for Global AI Distribution
|
||||
|
||||
**Last updated:** 2026-02
|
||||
**Status:** GA
|
||||
**Category:** API Management & AI Gateway
|
||||
|
||||
---
|
||||
|
||||
## Introduksjon
|
||||
|
||||
Nar organisasjoner ruller ut AI-tjenester globalt eller trenger ekstra beskyttelse og ytelsesoptimalisering, er kombinasjonen av Azure Front Door og Azure API Management en kraftig arkitektur. Azure Front Door gir global HTTP(S)-lastbalansering, DDoS-beskyttelse, Web Application Firewall (WAF), edge caching og TLS-offloading -- alt foran APIM som haandterer AI-spesifikk policy-haaandheving, token-ratebegrensning og backend-lastbalansering.
|
||||
|
||||
For norsk offentlig sektor er denne kombinasjonen relevant i flere scenarier: organisasjoner med innbyggertjenester som ma handtere trafikktopper (f.eks. skattemelding-perioden, eksamenssvar), tjenester som eksponeres mot internasjonale brukere, eller organisasjoner som krever ekstra lag med DDoS-beskyttelse. Azure Front Door sine globale PoP-er (Points of Presence) gir lavere latency for brukere narmere edge-lokasjoner.
|
||||
|
||||
Arkitekturen Front Door + APIM + AI Backend gir et tre-lags forsvar: Front Door handterer DDoS og WAF pa nettverksniva, APIM haandterer API-spesifikk sikkerhet og trafikkstyring, og Azure AI-tjenestene handterer modellspesifikk tilgangskontroll. Denne referansen dekker konfigurasjon, sikring og optimalisering av denne arkitekturen.
|
||||
|
||||
---
|
||||
|
||||
## Global lastdistribusjon
|
||||
|
||||
### Arkitekturoversikt
|
||||
|
||||
```
|
||||
Brukere (globalt)
|
||||
|
|
||||
v
|
||||
Azure Front Door (Global L7 Load Balancer)
|
||||
|-- PoP Oslo (naermest norske brukere)
|
||||
|-- PoP London
|
||||
|-- PoP New York
|
||||
|
|
||||
v
|
||||
APIM Instance(r)
|
||||
|-- West Europe (primaer)
|
||||
|-- North Europe (sekundaer)
|
||||
|
|
||||
v
|
||||
AI Backends
|
||||
|-- Azure OpenAI (West Europe)
|
||||
|-- Azure OpenAI (Sweden Central)
|
||||
|-- Microsoft Foundry (North Europe)
|
||||
```
|
||||
|
||||
### Oppsett av Front Door-profil
|
||||
|
||||
Konfigurer Front Door med APIM som origin:
|
||||
|
||||
| Innstilling | Verdi |
|
||||
|------------|-------|
|
||||
| **Origin type** | API Management |
|
||||
| **Origin hostname** | `{apim-name}.azure-api.net` |
|
||||
| **Caching** | Enable caching (for GET-requests) |
|
||||
| **Query string behavior** | Use Query String |
|
||||
| **Health probe path** | `/status-0123456789abcdef` |
|
||||
| **Health probe protocol** | HTTPS |
|
||||
| **Health probe method** | GET |
|
||||
| **Probe interval** | 30 sekunder |
|
||||
|
||||
### Bicep: Front Door med APIM Origin
|
||||
|
||||
```bicep
|
||||
resource frontDoorProfile 'Microsoft.Cdn/profiles@2024-02-01' = {
|
||||
name: frontDoorName
|
||||
location: 'global'
|
||||
sku: {
|
||||
name: 'Premium_AzureFrontDoor' // Premium for Private Link + WAF
|
||||
}
|
||||
}
|
||||
|
||||
resource frontDoorEndpoint 'Microsoft.Cdn/profiles/afdEndpoints@2024-02-01' = {
|
||||
parent: frontDoorProfile
|
||||
name: 'ai-gateway-endpoint'
|
||||
location: 'global'
|
||||
properties: {
|
||||
enabledState: 'Enabled'
|
||||
}
|
||||
}
|
||||
|
||||
resource originGroup 'Microsoft.Cdn/profiles/originGroups@2024-02-01' = {
|
||||
parent: frontDoorProfile
|
||||
name: 'apim-origin-group'
|
||||
properties: {
|
||||
loadBalancingSettings: {
|
||||
sampleSize: 4
|
||||
successfulSamplesRequired: 3
|
||||
additionalLatencyInMilliseconds: 50
|
||||
}
|
||||
healthProbeSettings: {
|
||||
probePath: '/status-0123456789abcdef'
|
||||
probeRequestType: 'GET'
|
||||
probeProtocol: 'Https'
|
||||
probeIntervalInSeconds: 30
|
||||
}
|
||||
sessionAffinityState: 'Disabled'
|
||||
}
|
||||
}
|
||||
|
||||
resource originWestEurope 'Microsoft.Cdn/profiles/originGroups/origins@2024-02-01' = {
|
||||
parent: originGroup
|
||||
name: 'apim-west-europe'
|
||||
properties: {
|
||||
hostName: '${apimNameWestEurope}.azure-api.net'
|
||||
httpPort: 80
|
||||
httpsPort: 443
|
||||
originHostHeader: '${apimNameWestEurope}.azure-api.net'
|
||||
priority: 1
|
||||
weight: 1000
|
||||
enabledState: 'Enabled'
|
||||
enforceCertificateNameCheck: true
|
||||
}
|
||||
}
|
||||
|
||||
resource originNorthEurope 'Microsoft.Cdn/profiles/originGroups/origins@2024-02-01' = {
|
||||
parent: originGroup
|
||||
name: 'apim-north-europe'
|
||||
properties: {
|
||||
hostName: '${apimNameNorthEurope}.azure-api.net'
|
||||
httpPort: 80
|
||||
httpsPort: 443
|
||||
originHostHeader: '${apimNameNorthEurope}.azure-api.net'
|
||||
priority: 2 // Failover
|
||||
weight: 1000
|
||||
enabledState: 'Enabled'
|
||||
enforceCertificateNameCheck: true
|
||||
}
|
||||
}
|
||||
|
||||
resource route 'Microsoft.Cdn/profiles/afdEndpoints/routes@2024-02-01' = {
|
||||
parent: frontDoorEndpoint
|
||||
name: 'ai-gateway-route'
|
||||
properties: {
|
||||
originGroup: {
|
||||
id: originGroup.id
|
||||
}
|
||||
supportedProtocols: [ 'Https' ]
|
||||
patternsToMatch: [ '/ai/*' ]
|
||||
forwardingProtocol: 'HttpsOnly'
|
||||
httpsRedirect: 'Enabled'
|
||||
linkToDefaultDomain: 'Enabled'
|
||||
}
|
||||
}
|
||||
```
|
||||
|
||||
---
|
||||
|
||||
## DDoS-beskyttelse
|
||||
|
||||
### Front Door innebygd DDoS-beskyttelse
|
||||
|
||||
Azure Front Door gir plattform-niva DDoS-beskyttelse automatisk:
|
||||
|
||||
| Beskyttelsestype | Dekning |
|
||||
|-----------------|---------|
|
||||
| L3/L4 DDoS | Automatisk for alle Front Door-profiler |
|
||||
| L7 DDoS | Via WAF-policyer |
|
||||
| Volumetriske angrep | Absorberes av Front Doors globale nettverk |
|
||||
| Protocol-angrep | Filtreres pa edge |
|
||||
| Application-layer | WAF rate limiting + bot protection |
|
||||
|
||||
### Kombinert beskyttelse
|
||||
|
||||
For kritiske AI-tjenester, kombiner Front Door med Azure DDoS Protection:
|
||||
|
||||
```bicep
|
||||
resource ddosProtectionPlan 'Microsoft.Network/ddosProtectionPlans@2023-11-01' = {
|
||||
name: 'ai-gateway-ddos-plan'
|
||||
location: location
|
||||
properties: {}
|
||||
}
|
||||
```
|
||||
|
||||
---
|
||||
|
||||
## Web Application Firewall
|
||||
|
||||
### WAF-policy for AI Gateway
|
||||
|
||||
```bicep
|
||||
resource wafPolicy 'Microsoft.Network/FrontDoorWebApplicationFirewallPolicies@2024-02-01' = {
|
||||
name: 'aiGatewayWafPolicy'
|
||||
location: 'global'
|
||||
sku: {
|
||||
name: 'Premium_AzureFrontDoor'
|
||||
}
|
||||
properties: {
|
||||
policySettings: {
|
||||
enabledState: 'Enabled'
|
||||
mode: 'Prevention'
|
||||
requestBodyCheck: 'Enabled'
|
||||
requestBodyInspectLimitInKB: 128
|
||||
}
|
||||
managedRules: {
|
||||
managedRuleSets: [
|
||||
{
|
||||
ruleSetType: 'Microsoft_DefaultRuleSet'
|
||||
ruleSetVersion: '2.1'
|
||||
ruleGroupOverrides: []
|
||||
}
|
||||
{
|
||||
ruleSetType: 'Microsoft_BotManagerRuleSet'
|
||||
ruleSetVersion: '1.1'
|
||||
}
|
||||
]
|
||||
}
|
||||
customRules: {
|
||||
rules: [
|
||||
{
|
||||
name: 'RateLimitAiRequests'
|
||||
priority: 100
|
||||
enabledState: 'Enabled'
|
||||
ruleType: 'RateLimitRule'
|
||||
rateLimitDurationInMinutes: 1
|
||||
rateLimitThreshold: 100
|
||||
matchConditions: [
|
||||
{
|
||||
matchVariable: 'RequestUri'
|
||||
operator: 'Contains'
|
||||
matchValue: [ '/ai/' ]
|
||||
}
|
||||
]
|
||||
action: 'Block'
|
||||
}
|
||||
{
|
||||
name: 'BlockSuspiciousPayloads'
|
||||
priority: 200
|
||||
enabledState: 'Enabled'
|
||||
ruleType: 'MatchRule'
|
||||
matchConditions: [
|
||||
{
|
||||
matchVariable: 'RequestBody'
|
||||
operator: 'Contains'
|
||||
matchValue: [
|
||||
'ignore previous instructions'
|
||||
'ignore all instructions'
|
||||
'disregard your system prompt'
|
||||
]
|
||||
transforms: [ 'Lowercase' ]
|
||||
}
|
||||
]
|
||||
action: 'Block'
|
||||
}
|
||||
{
|
||||
name: 'GeoBlockNonAllowed'
|
||||
priority: 300
|
||||
enabledState: 'Enabled'
|
||||
ruleType: 'MatchRule'
|
||||
matchConditions: [
|
||||
{
|
||||
matchVariable: 'RemoteAddr'
|
||||
operator: 'GeoMatch'
|
||||
negateCondition: true
|
||||
matchValue: [ 'NO', 'SE', 'DK', 'FI' ] // Nordiske land
|
||||
}
|
||||
]
|
||||
action: 'Log' // Start med Log, bytt til Block etter validering
|
||||
}
|
||||
]
|
||||
}
|
||||
}
|
||||
}
|
||||
```
|
||||
|
||||
### WAF-regler tilpasset AI-trafikk
|
||||
|
||||
| Regel | Type | Handling | Formal |
|
||||
|-------|------|---------|--------|
|
||||
| Rate limit per IP | RateLimit | Block | Maks 100 req/min per IP |
|
||||
| Prompt injection patterns | Match | Block | Blokkerer kjente injeksjonsmonstre |
|
||||
| Geo-filtering | Match | Log/Block | Begrens til tillatte land |
|
||||
| Bot protection | Managed | Block | Identifiserer og blokkerer botter |
|
||||
| OWASP Core Rules | Managed | Block | Standard webapplikasjonsbeskyttelse |
|
||||
| Payload size limit | Match | Block | Maks request body-storrelse |
|
||||
|
||||
---
|
||||
|
||||
## Edge Caching
|
||||
|
||||
### Caching-strategi for AI med Front Door
|
||||
|
||||
For AI-API-er er caching begrenset til GET-requests og statisk innhold. POST-baserte chat completion-kall caches ikke av Front Door, men det finnes bruksomrader:
|
||||
|
||||
| Innholdstype | Cachebar? | Strategi |
|
||||
|-------------|-----------|----------|
|
||||
| Chat completions (POST) | Nei | Bruk APIM semantisk caching |
|
||||
| Model listing (GET) | Ja | Front Door edge cache |
|
||||
| API documentation | Ja | Front Door edge cache |
|
||||
| Health endpoints | Nei | Bypass cache |
|
||||
| Static assets (dev portal) | Ja | Front Door edge cache |
|
||||
|
||||
### Caching for Developer Portal
|
||||
|
||||
```bicep
|
||||
resource devPortalRoute 'Microsoft.Cdn/profiles/afdEndpoints/routes@2024-02-01' = {
|
||||
parent: frontDoorEndpoint
|
||||
name: 'dev-portal-route'
|
||||
properties: {
|
||||
originGroup: {
|
||||
id: devPortalOriginGroup.id
|
||||
}
|
||||
supportedProtocols: [ 'Https' ]
|
||||
patternsToMatch: [ '/developer/*' ]
|
||||
forwardingProtocol: 'HttpsOnly'
|
||||
cacheConfiguration: {
|
||||
queryStringCachingBehavior: 'IgnoreQueryString'
|
||||
compressionSettings: {
|
||||
isCompressionEnabled: true
|
||||
contentTypesToCompress: [
|
||||
'text/html'
|
||||
'text/css'
|
||||
'application/javascript'
|
||||
'application/json'
|
||||
]
|
||||
}
|
||||
}
|
||||
}
|
||||
}
|
||||
```
|
||||
|
||||
---
|
||||
|
||||
## Geografisk ruting
|
||||
|
||||
### Priority-basert ruting med failover
|
||||
|
||||
```
|
||||
Bruker i Norge
|
||||
|
|
||||
v
|
||||
Front Door PoP Oslo
|
||||
|
|
||||
|-- Priority 1: APIM West Europe (Nederland)
|
||||
|-- Priority 2: APIM North Europe (Irland)
|
||||
|
|
||||
v
|
||||
APIM West Europe
|
||||
|
|
||||
|-- Priority 1: Azure OpenAI Sweden Central
|
||||
|-- Priority 2: Azure OpenAI West Europe
|
||||
```
|
||||
|
||||
### Restriksjon: Kun Front Door-trafikk til APIM
|
||||
|
||||
Sikre at APIM kun aksepterer trafikk fra Front Door:
|
||||
|
||||
```xml
|
||||
<policies>
|
||||
<inbound>
|
||||
<base />
|
||||
<!-- Verify Front Door ID header -->
|
||||
<check-header name="X-Azure-FDID"
|
||||
failed-check-httpcode="403"
|
||||
failed-check-error-message="Access denied. Traffic must route through Azure Front Door."
|
||||
ignore-case="false">
|
||||
<value>{{FrontDoorId}}</value>
|
||||
</check-header>
|
||||
|
||||
<!-- Additionally restrict to Front Door IP ranges -->
|
||||
<ip-filter action="allow">
|
||||
<address-range from="147.243.0.0" to="147.243.255.255" />
|
||||
<!-- AzureFrontDoor.Backend service tag ranges -->
|
||||
</ip-filter>
|
||||
</inbound>
|
||||
</policies>
|
||||
```
|
||||
|
||||
### Private Link mellom Front Door og APIM
|
||||
|
||||
For maksimal sikkerhet, bruk Front Door Premium med Private Link:
|
||||
|
||||
```bicep
|
||||
resource privateEndpoint 'Microsoft.Cdn/profiles/originGroups/origins@2024-02-01' = {
|
||||
parent: originGroup
|
||||
name: 'apim-private'
|
||||
properties: {
|
||||
hostName: '${apimName}.azure-api.net'
|
||||
originHostHeader: '${apimName}.azure-api.net'
|
||||
priority: 1
|
||||
weight: 1000
|
||||
enabledState: 'Enabled'
|
||||
sharedPrivateLinkResource: {
|
||||
privateLink: {
|
||||
id: apiManagement.id
|
||||
}
|
||||
privateLinkLocation: location
|
||||
groupId: 'gateway'
|
||||
requestMessage: 'Front Door Private Link to APIM'
|
||||
}
|
||||
}
|
||||
}
|
||||
```
|
||||
|
||||
---
|
||||
|
||||
## Kostnadsestimat for Front Door + APIM
|
||||
|
||||
| Komponent | Manedlig kostnad (NOK) |
|
||||
|-----------|----------------------|
|
||||
| Front Door Premium (base) | ~4 000 |
|
||||
| Front Door: 10M requests | ~1 500 |
|
||||
| Front Door: Data transfer (100 GB) | ~1 000 |
|
||||
| WAF Policy (Premium) | ~4 500 |
|
||||
| APIM Standard v2 | ~20 000 |
|
||||
| **Total** | **~31 000** |
|
||||
|
||||
Merk: Front Door Standard er rimeligere (~60% av Premium-pris) men mangler Private Link og WAF Managed Rules.
|
||||
|
||||
---
|
||||
|
||||
## Referanser
|
||||
|
||||
- [Configure Front Door Standard/Premium in front of Azure API Management](https://learn.microsoft.com/en-us/azure/api-management/front-door-api-management) -- trinnvis veiledning
|
||||
- [What is Azure Front Door?](https://learn.microsoft.com/en-us/azure/frontdoor/front-door-overview) -- oversikt
|
||||
- [Azure Front Door DDoS protection](https://learn.microsoft.com/en-us/azure/frontdoor/front-door-ddos) -- DDoS-beskyttelse
|
||||
- [Web Application Firewall on Azure Front Door](https://learn.microsoft.com/en-us/azure/web-application-firewall/afds/afds-overview) -- WAF-oversikt
|
||||
- [AI gateway in Azure API Management](https://learn.microsoft.com/en-us/azure/api-management/genai-gateway-capabilities) -- AI gateway
|
||||
- [Restrict caller IPs policy](https://learn.microsoft.com/en-us/azure/api-management/ip-filter-policy) -- IP-filtrering
|
||||
- [Check header policy](https://learn.microsoft.com/en-us/azure/api-management/check-header-policy) -- header-validering
|
||||
- [Architecture best practices for Azure Front Door](https://learn.microsoft.com/en-us/azure/well-architected/service-guides/azure-front-door) -- Well-Architected-anbefalinger
|
||||
|
||||
## For Cosmo
|
||||
|
||||
- **Bruk denne referansen** nar kunden trenger global distribusjon av AI-tjenester, ekstra DDoS-beskyttelse, eller WAF foran AI-gateway-en sin.
|
||||
- For de fleste norske offentlige virksomheter er Front Door overkill for rene interne AI-tjenester. Anbefal det primaert for innbyggerrettede tjenester med hoy trafikk eller behov for geographic redundancy.
|
||||
- Front Door Premium er nodvendig for Private Link til APIM og WAF Managed Rules. Standard-tier mangler disse, men er tilstrekkelig for basic lastbalansering og DDoS.
|
||||
- Husk alltid a konfigurere `X-Azure-FDID`-header-sjekk i APIM for a forhindre at noen omgar Front Door og kaller APIM direkte.
|
||||
- Kombiner Front Door WAF-regler for prompt injection-monstre pa nettverksniva med APIM Content Safety policy for AI-spesifikk innholdsmoderasjon -- dette gir forsvar i dybden.
|
||||
Loading…
Add table
Add a link
Reference in a new issue