fix(ai): add session affinity and compat fixes for Fireworks provider caching

Fireworks prompt caching is enabled by default (automatic prefix matching),
but on serverless infrastructure, requests hit random replicas. Without
session affinity, the per-replica cache misses, negating cache hit rates
and the discounted cacheRead pricing.

Changes:

- Add sendSessionAffinityHeaders and supportsCacheControlOnTools
  to AnthropicMessagesCompat interface
- Send x-session-affinity header for Fireworks (and Cloudflare AI
  Gateway Anthropic) when sessionId is available and caching is enabled
- Omit cache_control on tool definitions for Fireworks (unsupported
  per https://docs.fireworks.ai/tools-sdks/anthropic-compatibility)
- Default supportsEagerToolInputStreaming to false for Fireworks
  (unsupported field)
- Default supportsLongCacheRetention to false for Fireworks
  (cache_control.ttl not supported)
- Add compat settings to Fireworks models in generate-models.ts
- Update generated models with Fireworks compat settings
- Add integration tests for session affinity and tool compat

Refs: https://docs.fireworks.ai/guides/prompt-caching
Refs: https://docs.fireworks.ai/tools-sdks/anthropic-compatibility

packages/ai/src/providers/anthropic.ts

@@ -165,9 +165,16 @@ const FINE_GRAINED_TOOL_STREAMING_BETA = "fine-grained-tool-streaming-2025-05-14
 const INTERLEAVED_THINKING_BETA = "interleaved-thinking-2025-05-14";
 
 function getAnthropicCompat(model: Model<"anthropic-messages">): Required<AnthropicMessagesCompat> {
+	// Auto-detect session affinity and cache control support from provider
+	const isFireworks = model.provider === "fireworks";
+	const isCloudflareAiGatewayAnthropic =
+		model.provider === "cloudflare-ai-gateway" && model.baseUrl.includes("anthropic");
 	return {
-		supportsEagerToolInputStreaming: model.compat?.supportsEagerToolInputStreaming ?? true,
-		supportsLongCacheRetention: model.compat?.supportsLongCacheRetention ?? true,
+		supportsEagerToolInputStreaming: model.compat?.supportsEagerToolInputStreaming ?? !isFireworks,
+		supportsLongCacheRetention: model.compat?.supportsLongCacheRetention ?? !isFireworks,
+		sendSessionAffinityHeaders:
+			model.compat?.sendSessionAffinityHeaders ?? !!(isFireworks || isCloudflareAiGatewayAnthropic),
+		supportsCacheControlOnTools: model.compat?.supportsCacheControlOnTools ?? !isFireworks,
 	};
 }
 
@@ -463,6 +470,9 @@ export const streamAnthropic: StreamFunction<"anthropic-messages", AnthropicOpti
 					});
 				}
 
+				const cacheRetention = options?.cacheRetention ?? resolveCacheRetention();
+				const cacheSessionId = cacheRetention === "none" ? undefined : options?.sessionId;
+
 				const created = createClient(
 					model,
 					apiKey,
@@ -470,6 +480,7 @@ export const streamAnthropic: StreamFunction<"anthropic-messages", AnthropicOpti
 					shouldUseFineGrainedToolStreamingBeta(model, context),
 					options?.headers,
 					copilotDynamicHeaders,
+					cacheSessionId,
 				);
 				client = created.client;
 				isOAuth = created.isOAuthToken;
@@ -766,6 +777,7 @@ function createClient(
 	useFineGrainedToolStreamingBeta: boolean,
 	optionsHeaders?: Record<string, string>,
 	dynamicHeaders?: Record<string, string>,
+	sessionId?: string,
 ): { client: Anthropic; isOAuthToken: boolean } {
 	// Adaptive thinking models (Opus 4.6, Sonnet 4.6) have interleaved thinking built-in.
 	// The beta header is deprecated on Opus 4.6 and redundant on Sonnet 4.6, so skip it.
@@ -847,6 +859,8 @@ function createClient(
 	}
 
 	// API key auth
+	const sessionAffinityHeaders: Record<string, string | null> =
+		sessionId && getAnthropicCompat(model).sendSessionAffinityHeaders ? { "x-session-affinity": sessionId } : {};
 	const client = new Anthropic({
 		apiKey,
 		baseURL: model.baseUrl,
@@ -857,6 +871,7 @@ function createClient(
 				"anthropic-dangerous-direct-browser-access": "true",
 				...(betaFeatures.length > 0 ? { "anthropic-beta": betaFeatures.join(",") } : {}),
 			},
+			sessionAffinityHeaders,
 			model.headers,
 			optionsHeaders,
 		),
@@ -912,11 +927,12 @@ function buildParams(
 	}
 
 	if (context.tools && context.tools.length > 0) {
+		const compat = getAnthropicCompat(model);
 		params.tools = convertTools(
 			context.tools,
 			isOAuthToken,
-			getAnthropicCompat(model).supportsEagerToolInputStreaming,
-			cacheControl,
+			compat.supportsEagerToolInputStreaming,
+			compat.supportsCacheControlOnTools ? cacheControl : undefined,
 		);
 	}