1# Preset Deployment Workflow — Step-by-Step
2 
3Condensed implementation reference for preset (optimal region) model deployment. See [SKILL.md](../SKILL.md) for overview.
4 
5**Table of Contents:** [Phase 1: Verify Authentication](#phase-1-verify-authentication) · [Phase 2: Get Current Project](#phase-2-get-current-project) · [Phase 3: Get Model Name](#phase-3-get-model-name) · [Phase 4: Check Current Region Capacity](#phase-4-check-current-region-capacity) · [Phase 5: Query Multi-Region Capacity](#phase-5-query-multi-region-capacity) · [Phase 6: Select Region and Project](#phase-6-select-region-and-project) · [Phase 7: Deploy Model](#phase-7-deploy-model)
6 
7---
8 
9## Phase 1: Verify Authentication
10 
11```bash
12az account show --query "{Subscription:name, User:user.name}" -o table
13```
14 
15If not logged in: `az login`
16 
17Switch subscription:
18 
19```bash
20az account list --query "[].[name,id,state]" -o table
21az account set --subscription <subscription-id>
22```
23 
24---
25 
26## Phase 2: Get Current Project
27 
28Read `PROJECT_RESOURCE_ID` from env or prompt user. Format:
29`/subscriptions/{sub-id}/resourceGroups/{rg}/providers/Microsoft.CognitiveServices/accounts/{account}/projects/{project}`
30 
31Parse ARM ID components:
32 
33```bash
34SUBSCRIPTION_ID=$(echo "$PROJECT_RESOURCE_ID" | sed -n 's|.*/subscriptions/\([^/]*\).*|\1|p')
35RESOURCE_GROUP=$(echo "$PROJECT_RESOURCE_ID" | sed -n 's|.*/resourceGroups/\([^/]*\).*|\1|p')
36ACCOUNT_NAME=$(echo "$PROJECT_RESOURCE_ID" | sed -n 's|.*/accounts/\([^/]*\)/projects.*|\1|p')
37PROJECT_NAME=$(echo "$PROJECT_RESOURCE_ID" | sed -n 's|.*/projects/\([^/?]*\).*|\1|p')
38```
39 
40Verify project exists and get region:
41 
42```bash
43az account set --subscription "$SUBSCRIPTION_ID"
44 
45PROJECT_REGION=$(az cognitiveservices account show \
46  --name "$PROJECT_NAME" \
47  --resource-group "$RESOURCE_GROUP" \
48  --query location -o tsv)
49```
50 
51---
52 
53## Phase 3: Get Model Name
54 
55If model not provided as parameter, list available models:
56 
57```bash
58az cognitiveservices account list-models \
59  --name "$PROJECT_NAME" \
60  --resource-group "$RESOURCE_GROUP" \
61  --query "[].name" -o tsv | sort -u
62```
63 
64Get versions for selected model:
65 
66```bash
67az cognitiveservices account list-models \
68  --name "$PROJECT_NAME" \
69  --resource-group "$RESOURCE_GROUP" \
70  --query "[?name=='$MODEL_NAME'].{Name:name, Version:version, Format:format}" \
71  -o table
72```
73 
74---
75 
76## Phase 4: Check Current Region Capacity
77 
78```bash
79CAPACITY_JSON=$(az rest --method GET \
80  --url "https://management.azure.com/subscriptions/$SUBSCRIPTION_ID/providers/Microsoft.CognitiveServices/locations/$PROJECT_REGION/modelCapacities?api-version=2024-10-01&modelFormat=OpenAI&modelName=$MODEL_NAME&modelVersion=$MODEL_VERSION")
81 
82CURRENT_CAPACITY=$(echo "$CAPACITY_JSON" | jq -r '.value[] | select(.properties.skuName=="GlobalStandard") | .properties.availableCapacity')
83```
84 
85If `CURRENT_CAPACITY > 0` → skip to Phase 7. Otherwise continue to Phase 5.
86 
87---
88 
89## Phase 5: Query Multi-Region Capacity
90 
91```bash
92ALL_REGIONS_JSON=$(az rest --method GET \
93  --url "https://management.azure.com/subscriptions/$SUBSCRIPTION_ID/providers/Microsoft.CognitiveServices/modelCapacities?api-version=2024-10-01&modelFormat=OpenAI&modelName=$MODEL_NAME&modelVersion=$MODEL_VERSION")
94```
95 
96Extract available regions (capacity > 0):
97 
98```bash
99AVAILABLE_REGIONS=$(echo "$ALL_REGIONS_JSON" | jq -r '.value[] | select(.properties.skuName=="GlobalStandard" and .properties.availableCapacity > 0) | "\(.location)|\(.properties.availableCapacity)"')
100```
101 
102Extract unavailable regions:
103 
104```bash
105UNAVAILABLE_REGIONS=$(echo "$ALL_REGIONS_JSON" | jq -r '.value[] | select(.properties.skuName=="GlobalStandard" and (.properties.availableCapacity == 0 or .properties.availableCapacity == null)) | "\(.location)|0"')
106```
107 
108If no regions have capacity, defer to the [quota skill](../../../../quota/quota.md) for increase requests. Suggest checking existing deployments or trying alternative models like `gpt-4o-mini`.
109 
110---
111 
112## Phase 6: Select Region and Project
113 
114Present available regions to user. Store selection as `SELECTED_REGION`.
115 
116Find projects in selected region:
117 
118```bash
119PROJECTS_IN_REGION=$(az cognitiveservices account list \
120  --query "[?kind=='AIProject' && location=='$SELECTED_REGION'].{Name:name, ResourceGroup:resourceGroup}" \
121  --output json)
122```
123 
124**If no projects exist — create new:**
125 
126```bash
127az cognitiveservices account create \
128  --name "$HUB_NAME" \
129  --resource-group "$RESOURCE_GROUP" \
130  --location "$SELECTED_REGION" \
131  --kind "AIServices" \
132  --sku "S0" --yes
133 
134az cognitiveservices account create \
135  --name "$NEW_PROJECT_NAME" \
136  --resource-group "$RESOURCE_GROUP" \
137  --location "$SELECTED_REGION" \
138  --kind "AIProject" \
139  --sku "S0" --yes
140```
141 
142---
143 
144## Phase 7: Deploy Model
145 
146Generate unique deployment name using `scripts/generate_deployment_name.sh`:
147 
148```bash
149DEPLOYMENT_NAME=$(bash scripts/generate_deployment_name.sh "$ACCOUNT_NAME" "$RESOURCE_GROUP" "$MODEL_NAME")
150```
151 
152Calculate capacity — 50% of available, minimum 50 TPM:
153 
154```bash
155SELECTED_CAPACITY=$(echo "$ALL_REGIONS_JSON" | jq -r ".value[] | select(.location==\"$SELECTED_REGION\" and .properties.skuName==\"GlobalStandard\") | .properties.availableCapacity")
156DEPLOY_CAPACITY=$(( SELECTED_CAPACITY / 2 ))
157[ "$DEPLOY_CAPACITY" -lt 50 ] && DEPLOY_CAPACITY=50
158```
159 
160Create deployment:
161 
162```bash
163az cognitiveservices account deployment create \
164  --name "$ACCOUNT_NAME" \
165  --resource-group "$RESOURCE_GROUP" \
166  --deployment-name "$DEPLOYMENT_NAME" \
167  --model-name "$MODEL_NAME" \
168  --model-version "$MODEL_VERSION" \
169  --model-format "OpenAI" \
170  --sku-name "GlobalStandard" \
171  --sku-capacity "$DEPLOY_CAPACITY"
172```
173 
174Monitor with `az cognitiveservices account deployment show ... --query "properties.provisioningState"` until `Succeeded` or `Failed`.
175