[management] expose prompt-cache accounting in the agent-network REST API

Access-log entries, session summaries, and usage-overview buckets gain
cached_input_tokens, cache_creation_tokens, and cache_cost_usd so the
cache share of a request's cost is visible instead of hiding inside
total_tokens. Generated with shared/management/http/api/generate.sh.

shared/management/http/api/openapi.yml

@@ -5822,13 +5822,28 @@ components:
         total_tokens:
           type: integer
           format: int64
-          description: Total tokens consumed.
+          description: Total tokens consumed, including prompt-cache tokens.
           example: 1840
+        cached_input_tokens:
+          type: integer
+          format: int64
+          description: Input tokens read from the provider's prompt cache. Additive to input_tokens for Anthropic-shape providers; a subset of input_tokens for OpenAI.
+          example: 0
+        cache_creation_tokens:
+          type: integer
+          format: int64
+          description: Input tokens written to the provider's prompt cache. Zero for providers without a cache-write bucket.
+          example: 30528
         cost_usd:
           type: number
           format: double
           description: Estimated USD cost of the request.
           example: 0.0231
+        cache_cost_usd:
+          type: number
+          format: double
+          description: Portion of cost_usd billed for prompt-cache usage.
+          example: 0.1145
         stream:
           type: boolean
           description: Whether the request was a streaming completion.
@@ -5852,7 +5867,10 @@ components:
         - input_tokens
         - output_tokens
         - total_tokens
+        - cached_input_tokens
+        - cache_creation_tokens
         - cost_usd
+        - cache_cost_usd
     AgentNetworkAccessLogsResponse:
       type: object
       properties:
@@ -5926,13 +5944,28 @@ components:
         total_tokens:
           type: integer
           format: int64
-          description: Total tokens across the session.
+          description: Total tokens across the session, including prompt-cache tokens.
           example: 12880
+        cached_input_tokens:
+          type: integer
+          format: int64
+          description: Total prompt-cache read tokens across the session.
+          example: 0
+        cache_creation_tokens:
+          type: integer
+          format: int64
+          description: Total prompt-cache write tokens across the session.
+          example: 30528
         cost_usd:
           type: number
           format: double
           description: Total estimated USD cost across the session.
           example: 0.1617
+        cache_cost_usd:
+          type: number
+          format: double
+          description: Portion of cost_usd billed for prompt-cache usage across the session.
+          example: 0.1145
         providers:
           type: array
           items:
@@ -5959,7 +5992,10 @@ components:
         - input_tokens
         - output_tokens
         - total_tokens
+        - cached_input_tokens
+        - cache_creation_tokens
         - cost_usd
+        - cache_cost_usd
         - decision
         - entries
     AgentNetworkAccessLogSessionsResponse:
@@ -6013,19 +6049,37 @@ components:
         total_tokens:
           type: integer
           format: int64
-          description: Total tokens in the bucket.
+          description: Total tokens in the bucket, including prompt-cache tokens.
           example: 184000
+        cached_input_tokens:
+          type: integer
+          format: int64
+          description: Total prompt-cache read tokens in the bucket.
+          example: 20000
+        cache_creation_tokens:
+          type: integer
+          format: int64
+          description: Total prompt-cache write tokens in the bucket.
+          example: 45000
         cost_usd:
           type: number
           format: double
           description: Total estimated USD spend in the bucket.
           example: 2.31
+        cache_cost_usd:
+          type: number
+          format: double
+          description: Portion of cost_usd billed for prompt-cache usage in the bucket.
+          example: 0.42
       required:
         - period_start
         - input_tokens
         - output_tokens
         - total_tokens
+        - cached_input_tokens
+        - cache_creation_tokens
         - cost_usd
+        - cache_cost_usd
     AgentNetworkSettings:
       type: object
       description: Per-account Agent Network gateway settings. One row per account; cluster and subdomain are auto-assigned on first provider create and immutable thereafter.

shared/management/http/api/types.gen.go

@@ -1732,6 +1732,15 @@ type AccountSettings struct {
 
 // AgentNetworkAccessLog One per-request agent-network (LLM) access log entry with flattened, queryable LLM dimensions.
 type AgentNetworkAccessLog struct {
+	// CacheCostUsd Portion of cost_usd billed for prompt-cache usage.
+	CacheCostUsd float64 `json:"cache_cost_usd"`
+
+	// CacheCreationTokens Input tokens written to the provider's prompt cache. Zero for providers without a cache-write bucket.
+	CacheCreationTokens int64 `json:"cache_creation_tokens"`
+
+	// CachedInputTokens Input tokens read from the provider's prompt cache. Additive to input_tokens for Anthropic-shape providers; a subset of input_tokens for OpenAI.
+	CachedInputTokens int64 `json:"cached_input_tokens"`
+
 	// CostUsd Estimated USD cost of the request.
 	CostUsd float64 `json:"cost_usd"`
 
@@ -1801,7 +1810,7 @@ type AgentNetworkAccessLog struct {
 	// Timestamp Timestamp when the request was made.
 	Timestamp time.Time `json:"timestamp"`
 
-	// TotalTokens Total tokens consumed.
+	// TotalTokens Total tokens consumed, including prompt-cache tokens.
 	TotalTokens int64 `json:"total_tokens"`
 
 	// UserId NetBird user id of the authenticated caller, if applicable.
@@ -1810,6 +1819,15 @@ type AgentNetworkAccessLog struct {
 
 // AgentNetworkAccessLogSession A session-grouped view of agent-network access logs — all requests sharing a session id (or a single session-less request) folded into one summary plus its ordered entries.
 type AgentNetworkAccessLogSession struct {
+	// CacheCostUsd Portion of cost_usd billed for prompt-cache usage across the session.
+	CacheCostUsd float64 `json:"cache_cost_usd"`
+
+	// CacheCreationTokens Total prompt-cache write tokens across the session.
+	CacheCreationTokens int64 `json:"cache_creation_tokens"`
+
+	// CachedInputTokens Total prompt-cache read tokens across the session.
+	CachedInputTokens int64 `json:"cached_input_tokens"`
+
 	// CostUsd Total estimated USD cost across the session.
 	CostUsd float64 `json:"cost_usd"`
 
@@ -1846,7 +1864,7 @@ type AgentNetworkAccessLogSession struct {
 	// StartedAt Timestamp of the session's earliest request.
 	StartedAt time.Time `json:"started_at"`
 
-	// TotalTokens Total tokens across the session.
+	// TotalTokens Total tokens across the session, including prompt-cache tokens.
 	TotalTokens int64 `json:"total_tokens"`
 
 	// UserId NetBird user id of the session's caller.
@@ -2347,6 +2365,15 @@ type AgentNetworkSettingsRequest struct {
 
 // AgentNetworkUsageBucket One aggregated agent-network usage time bucket (UTC). The bucket width is set by the request's granularity.
 type AgentNetworkUsageBucket struct {
+	// CacheCostUsd Portion of cost_usd billed for prompt-cache usage in the bucket.
+	CacheCostUsd float64 `json:"cache_cost_usd"`
+
+	// CacheCreationTokens Total prompt-cache write tokens in the bucket.
+	CacheCreationTokens int64 `json:"cache_creation_tokens"`
+
+	// CachedInputTokens Total prompt-cache read tokens in the bucket.
+	CachedInputTokens int64 `json:"cached_input_tokens"`
+
 	// CostUsd Total estimated USD spend in the bucket.
 	CostUsd float64 `json:"cost_usd"`
 
@@ -2359,7 +2386,7 @@ type AgentNetworkUsageBucket struct {
 	// PeriodStart Start of the bucket in YYYY-MM-DD (UTC) — the day, the week start (Monday), or the month start, depending on granularity.
 	PeriodStart string `json:"period_start"`
 
-	// TotalTokens Total tokens in the bucket.
+	// TotalTokens Total tokens in the bucket, including prompt-cache tokens.
 	TotalTokens int64 `json:"total_tokens"`
 }