feat(js/plugins/vertexai): add initial attempt at context cache vertex

js/plugins/vertexai/package.json

@@ -53,6 +53,7 @@
   },
   "devDependencies": {
     "@types/node": "^20.11.16",
+    "google-gax": "^4.4.1",
     "npm-run-all": "^4.1.5",
     "rimraf": "^6.0.1",
     "tsup": "^8.3.5",

js/plugins/vertexai/src/context_caching/constants.ts

@@ -0,0 +1,29 @@
+/**
+ * Copyright 2024 Google LLC
+ *
+ * Licensed under the Apache License, Version 2.0 (the "License");
+ * you may not use this file except in compliance with the License.
+ * You may obtain a copy of the License at
+ *
+ *     http://www.apache.org/licenses/LICENSE-2.0
+ *
+ * Unless required by applicable law or agreed to in writing, software
+ * distributed under the License is distributed on an "AS IS" BASIS,
+ * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ * See the License for the specific language governing permissions and
+ * limitations under the License.
+ */
+
+export const CONTEXT_CACHE_SUPPORTED_MODELS = [
+  'gemini-1.5-flash-001',
+  'gemini-1.5-pro-001',
+];
+
+export const INVALID_ARGUMENT_MESSAGES = {
+  modelVersion: `Model version is required for context caching, supported only in ${CONTEXT_CACHE_SUPPORTED_MODELS.join(',')} models.`,
+  tools: 'Context caching cannot be used simultaneously with tools.',
+  codeExecution:
+    'Context caching cannot be used simultaneously with code execution.',
+};
+
+export const DEFAULT_TTL = 300;

js/plugins/vertexai/src/context_caching/index.ts

@@ -0,0 +1,97 @@
+/**
+ * @license
+ * Copyright 2024 Google LLC
+ *
+ * Licensed under the Apache License, Version 2.0 (the "License");
+ * you may not use this file except in compliance with the License.
+ * You may obtain a copy of the License at
+ *
+ *     https://www.apache.org/licenses/LICENSE-2.0
+ *
+ * Unless required by applicable law or agreed to in writing, software
+ * distributed on an "AS IS" BASIS,
+ * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ */
+
+import { CachedContent, StartChatParams } from '@google-cloud/vertexai';
+import {
+  ApiClient,
+  CachedContents,
+} from '@google-cloud/vertexai/build/src/resources';
+import { GenerateRequest, GenkitError, z } from 'genkit';
+import { logger } from 'genkit/logging';
+import type { CacheConfigDetails } from './types.js';
+import {
+  calculateTTL,
+  generateCacheKey,
+  getContentForCache,
+  lookupContextCache,
+} from './utils.js';
+
+/**
+ * Handles context caching and transforms the chatRequest for Vertex AI.
+ * @param apiKey
+ * @param request
+ * @param chatRequest
+ * @param modelVersion
+ * @returns
+ */
+export async function handleContextCache(
+  apiClient: ApiClient,
+  request: GenerateRequest<z.ZodTypeAny>,
+  chatRequest: StartChatParams,
+  modelVersion: string,
+  cacheConfigDetails: CacheConfigDetails
+): Promise<{ cache: CachedContent; newChatRequest: StartChatParams }> {
+  const cachedContentsClient = new CachedContents(apiClient);
+
+  const { cachedContent, chatRequest: newChatRequest } = getContentForCache(
+    request,
+    chatRequest,
+    modelVersion,
+    cacheConfigDetails
+  );
+  cachedContent.model = modelVersion;
+  const cacheKey = generateCacheKey(cachedContent);
+
+  cachedContent.displayName = cacheKey;
+
+  let cache;
+  try {
+    cache = await lookupContextCache(cachedContentsClient, cacheKey);
+    logger.debug(`Cache hit: ${cache ? 'true' : 'false'}`);
+  } catch (error) {
+    logger.debug('No cache found, creating one.');
+  }
+
+  if (!cache) {
+    try {
+      const createParams: CachedContent = {
+        ...cachedContent,
+        // TODO: make this neater - idk why they chose to stringify the ttl...
+        ttl: JSON.stringify(calculateTTL(cacheConfigDetails)) + 's',
+      };
+      cache = await cachedContentsClient.create(createParams);
+      logger.debug(`Created new cache entry with key: ${cacheKey}`);
+    } catch (cacheError) {
+      logger.error(
+        `Failed to create cache with key ${cacheKey}: ${cacheError}`
+      );
+      throw new GenkitError({
+        status: 'INTERNAL',
+        message: `Failed to create cache: ${cacheError}`,
+      });
+    }
+  }
+
+  if (!cache) {
+    throw new GenkitError({
+      status: 'INTERNAL',
+      message: 'Failed to use context cache feature',
+    });
+  }
+  // This isn't necessary, but it's nice to have for debugging purposes.
+  newChatRequest.cachedContent = cache.name;
+
+  return { cache, newChatRequest };
+}

js/plugins/vertexai/src/context_caching/types.ts

@@ -0,0 +1,31 @@
+/**
+ * Copyright 2024 Google LLC
+ *
+ * Licensed under the Apache License, Version 2.0 (the "License");
+ * you may not use this file except in compliance with the License.
+ * You may obtain a copy of the License at
+ *
+ *     http://www.apache.org/licenses/LICENSE-2.0
+ *
+ * Unless required by applicable law or agreed to in writing, software
+ * distributed under the License is distributed on an "AS IS" BASIS,
+ * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ * See the License for the specific language governing permissions and
+ * limitations under the License.
+ */
+
+import { z } from 'genkit';
+
+export const cacheConfigSchema = z.union([
+  z.boolean(),
+  z.object({ ttlSeconds: z.number().optional() }).passthrough(),
+]);
+
+export type CacheConfig = z.infer<typeof cacheConfigSchema>;
+
+export const cacheConfigDetailsSchema = z.object({
+  cacheConfig: cacheConfigSchema,
+  endOfCachedContents: z.number(),
+});
+
+export type CacheConfigDetails = z.infer<typeof cacheConfigDetailsSchema>;