shumengya/sproutclaw
1
0
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity

fix(ai): support long cache retention compat

Browse Source 
closes #3543
This commit is contained in:
Mario Zechner
2026-04-23 23:43:34 +02:00
parent 1312346199
commit 4cd4cfd98e
11 changed files with 331 additions and 29 deletions
Download Patch File Download Diff File Expand all files Collapse all files

7
packages/ai/src/providers/anthropic.ts
View File

@@ -51,14 +51,14 @@ function resolveCacheRetention(cacheRetention?: CacheRetention): CacheRetention
}
function getCacheControl(
baseUrl: string,
model: Model<"anthropic-messages">,
cacheRetention?: CacheRetention,
): { retention: CacheRetention; cacheControl?: CacheControlEphemeral } {
const retention = resolveCacheRetention(cacheRetention);
if (retention === "none") {
return { retention };
}
const ttl = retention === "long" && baseUrl.includes("api.anthropic.com") ? "1h" : undefined;
const ttl = retention === "long" && getAnthropicCompat(model).supportsLongCacheRetention ? "1h" : undefined;
return {
retention,
cacheControl: { type: "ephemeral", ...(ttl && { ttl }) },
@@ -166,6 +166,7 @@ const INTERLEAVED_THINKING_BETA = "interleaved-thinking-2025-05-14";
function getAnthropicCompat(model: Model<"anthropic-messages">): Required<AnthropicMessagesCompat> {
return {
supportsEagerToolInputStreaming: model.compat?.supportsEagerToolInputStreaming ?? true,
supportsLongCacheRetention: model.compat?.supportsLongCacheRetention ?? true,
};
}
@@ -833,7 +834,7 @@ function buildParams(
isOAuthToken: boolean,
options?: AnthropicOptions,
): MessageCreateParamsStreaming {
const { cacheControl } = getCacheControl(model.baseUrl, options?.cacheRetention);
const { cacheControl } = getCacheControl(model, options?.cacheRetention);
const params: MessageCreateParamsStreaming = {
model: model.id,
messages: convertMessages(context.messages, model, isOAuthToken, cacheControl),

31
packages/ai/src/providers/openai-responses.ts
View File

@@ -8,6 +8,7 @@ import type {
CacheRetention,
Context,
Model,
OpenAIResponsesCompat,
SimpleStreamOptions,
StreamFunction,
StreamOptions,
@@ -35,18 +36,18 @@ function resolveCacheRetention(cacheRetention?: CacheRetention): CacheRetention
return "short";
}
/**
* Get prompt cache retention based on cacheRetention and base URL.
* Only applies to direct OpenAI API calls (api.openai.com).
*/
function getPromptCacheRetention(baseUrl: string, cacheRetention: CacheRetention): "24h" | undefined {
if (cacheRetention !== "long") {
return undefined;
}
if (baseUrl.includes("api.openai.com")) {
return "24h";
}
return undefined;
function getCompat(model: Model<"openai-responses">): Required<OpenAIResponsesCompat> {
return {
sendSessionIdHeader: model.compat?.sendSessionIdHeader ?? true,
supportsLongCacheRetention: model.compat?.supportsLongCacheRetention ?? true,
};
}
function getPromptCacheRetention(
compat: Required<OpenAIResponsesCompat>,
cacheRetention: CacheRetention,
): "24h" | undefined {
return cacheRetention === "long" && compat.supportsLongCacheRetention ? "24h" : undefined;
}
// OpenAI Responses-specific options
@@ -169,6 +170,7 @@ function createClient(
apiKey = process.env.OPENAI_API_KEY;
}
const compat = getCompat(model);
const headers = { ...model.headers };
if (model.provider === "github-copilot") {
const hasImages = hasCopilotVisionInput(context.messages);
@@ -180,7 +182,7 @@ function createClient(
}
if (sessionId) {
if (model.compat?.sendSessionIdHeader !== false) {
if (compat.sendSessionIdHeader) {
headers.session_id = sessionId;
}
headers["x-client-request-id"] = sessionId;
@@ -203,12 +205,13 @@ function buildParams(model: Model<"openai-responses">, context: Context, options
const messages = convertResponsesMessages(model, context, OPENAI_TOOL_CALL_PROVIDERS);
const cacheRetention = resolveCacheRetention(options?.cacheRetention);
const compat = getCompat(model);
const params: ResponseCreateParamsStreaming = {
model: model.id,
input: messages,
stream: true,
prompt_cache_key: cacheRetention === "none" ? undefined : options?.sessionId,
prompt_cache_retention: getPromptCacheRetention(model.baseUrl, cacheRetention),
prompt_cache_retention: getPromptCacheRetention(compat, cacheRetention),
store: false,
};

6
packages/ai/src/types.ts
View File

@@ -296,12 +296,16 @@ export interface OpenAICompletionsCompat {
cacheControlFormat?: "anthropic";
/** Whether to send known session-affinity headers (`session_id`, `x-client-request-id`, `x-session-affinity`) from `options.sessionId` when caching is enabled. Default: false. */
sendSessionAffinityHeaders?: boolean;
/** Whether the provider supports long prompt cache retention (`prompt_cache_retention: "24h"` or Anthropic-style `cache_control.ttl: "1h"`, depending on format). Default: true. */
supportsLongCacheRetention?: boolean;
}
/** Compatibility settings for OpenAI Responses APIs. */
export interface OpenAIResponsesCompat {
/** Whether to send the OpenAI `session_id` cache-affinity header from `options.sessionId` when caching is enabled. Default: true. */
sendSessionIdHeader?: boolean;
/** Whether the provider supports `prompt_cache_retention: "24h"`. Default: true. */
supportsLongCacheRetention?: boolean;
}
/** Compatibility settings for Anthropic Messages-compatible APIs. */
@@ -314,6 +318,8 @@ export interface AnthropicMessagesCompat {
* Default: true.
*/
supportsEagerToolInputStreaming?: boolean;
/** Whether the provider supports Anthropic long cache retention (`cache_control.ttl: "1h"`). Default: true. */
supportsLongCacheRetention?: boolean;
}
/**