Release 0.6.7 (#1244)

Co-authored-by: github-actions[bot] <github-actions[bot]@users.noreply.github.com>

apps/docs/CHANGELOG.md

@@ -1,5 +1,21 @@
 # docs
 
+## 0.0.76
+
+### Patch Changes
+
+- Updated dependencies [23bcc37]
+  - llamaindex@0.6.7
+
+## 0.0.75
+
+### Patch Changes
+
+- Updated dependencies [d902cc3]
+- Updated dependencies [025ffe6]
+- Updated dependencies [a659574]
+  - llamaindex@0.6.6
+
 ## 0.0.74
 
 ### Patch Changes

apps/docs/package.json

@@ -1,6 +1,6 @@
 {
   "name": "docs",
-  "version": "0.0.74",
+  "version": "0.0.76",
   "private": true,
   "scripts": {
     "docusaurus": "docusaurus",

packages/autotool/CHANGELOG.md

@@ -1,5 +1,21 @@
 # @llamaindex/autotool
 
+## 3.0.7
+
+### Patch Changes
+
+- Updated dependencies [23bcc37]
+  - llamaindex@0.6.7
+
+## 3.0.6
+
+### Patch Changes
+
+- Updated dependencies [d902cc3]
+- Updated dependencies [025ffe6]
+- Updated dependencies [a659574]
+  - llamaindex@0.6.6
+
 ## 3.0.5
 
 ### Patch Changes

packages/autotool/examples/01_node/CHANGELOG.md

@@ -1,5 +1,23 @@
 # @llamaindex/autotool-01-node-example
 
+## 0.0.16
+
+### Patch Changes
+
+- Updated dependencies [23bcc37]
+  - llamaindex@0.6.7
+  - @llamaindex/autotool@3.0.7
+
+## 0.0.15
+
+### Patch Changes
+
+- Updated dependencies [d902cc3]
+- Updated dependencies [025ffe6]
+- Updated dependencies [a659574]
+  - llamaindex@0.6.6
+  - @llamaindex/autotool@3.0.6
+
 ## 0.0.14
 
 ### Patch Changes

packages/autotool/examples/01_node/package.json

@@ -13,5 +13,5 @@
   "scripts": {
     "start": "node --import tsx --import @llamaindex/autotool/node ./src/index.ts"
   },
-  "version": "0.0.14"
+  "version": "0.0.16"
 }

packages/autotool/examples/02_nextjs/CHANGELOG.md

@@ -1,5 +1,23 @@
 # @llamaindex/autotool-02-next-example
 
+## 0.1.60
+
+### Patch Changes
+
+- Updated dependencies [23bcc37]
+  - llamaindex@0.6.7
+  - @llamaindex/autotool@3.0.7
+
+## 0.1.59
+
+### Patch Changes
+
+- Updated dependencies [d902cc3]
+- Updated dependencies [025ffe6]
+- Updated dependencies [a659574]
+  - llamaindex@0.6.6
+  - @llamaindex/autotool@3.0.6
+
 ## 0.1.58
 
 ### Patch Changes

packages/autotool/examples/02_nextjs/package.json

@@ -1,7 +1,7 @@
 {
   "name": "@llamaindex/autotool-02-next-example",
   "private": true,
-  "version": "0.1.58",
+  "version": "0.1.60",
   "scripts": {
     "dev": "next dev",
     "build": "next build",

packages/autotool/package.json

@@ -1,7 +1,7 @@
 {
   "name": "@llamaindex/autotool",
   "type": "module",
-  "version": "3.0.5",
+  "version": "3.0.7",
   "description": "auto transpile your JS function to LLM Agent compatible",
   "files": [
     "dist",

packages/community/CHANGELOG.md

@@ -1,5 +1,12 @@
 # @llamaindex/community
 
+## 0.0.39
+
+### Patch Changes
+
+- Updated dependencies [d902cc3]
+  - @llamaindex/core@0.2.5
+
 ## 0.0.38
 
 ### Patch Changes

packages/community/package.json

@@ -1,7 +1,7 @@
 {
   "name": "@llamaindex/community",
   "description": "Community package for LlamaIndexTS",
-  "version": "0.0.38",
+  "version": "0.0.39",
   "type": "module",
   "types": "dist/type/index.d.ts",
   "main": "dist/cjs/index.js",

packages/core/CHANGELOG.md

@@ -1,5 +1,11 @@
 # @llamaindex/core
 
+## 0.2.5
+
+### Patch Changes
+
+- d902cc3: Fix context not being sent using ContextChatEngine
+
 ## 0.2.4
 
 ### Patch Changes

packages/core/package.json

@@ -1,7 +1,7 @@
 {
   "name": "@llamaindex/core",
   "type": "module",
-  "version": "0.2.4",
+  "version": "0.2.5",
   "description": "LlamaIndex Core Module",
   "exports": {
     "./node-parser": {

packages/core/src/memory/base.ts

@@ -1,5 +1,5 @@
 import { Settings } from "../global";
-import type { ChatMessage, MessageContent } from "../llms";
+import type { ChatMessage } from "../llms";
 import { type BaseChatStore, SimpleChatStore } from "../storage/chat-store";
 import { extractText } from "../utils";
 
@@ -12,15 +12,36 @@ export const DEFAULT_CHAT_STORE_KEY = "chat_history";
 export abstract class BaseMemory<
   AdditionalMessageOptions extends object = object,
 > {
+  /**
+   * Retrieves messages from the memory, optionally including transient messages.
+   * Compared to getAllMessages, this method a) allows for transient messages to be included in the retrieval and b) may return a subset of the total messages by applying a token limit.
+   * @param transientMessages Optional array of temporary messages to be included in the retrieval.
+   * These messages are not stored in the memory but are considered for the current interaction.
+   * @returns An array of chat messages, either synchronously or as a Promise.
+   */
   abstract getMessages(
-    input?: MessageContent | undefined,
+    transientMessages?: ChatMessage<AdditionalMessageOptions>[] | undefined,
   ):
     | ChatMessage<AdditionalMessageOptions>[]
     | Promise<ChatMessage<AdditionalMessageOptions>[]>;
+
+  /**
+   * Retrieves all messages stored in the memory.
+   * @returns An array of all chat messages, either synchronously or as a Promise.
+   */
   abstract getAllMessages():
     | ChatMessage<AdditionalMessageOptions>[]
     | Promise<ChatMessage<AdditionalMessageOptions>[]>;
+
+  /**
+   * Adds a new message to the memory.
+   * @param messages The chat message to be added to the memory.
+   */
   abstract put(messages: ChatMessage<AdditionalMessageOptions>): void;
+
+  /**
+   * Clears all messages from the memory.
+   */
   abstract reset(): void;
 
   protected _tokenCountForMessages(messages: ChatMessage[]): number {

packages/core/src/memory/chat-memory-buffer.ts

@@ -1,5 +1,5 @@
 import { Settings } from "../global";
-import type { ChatMessage, LLM, MessageContent } from "../llms";
+import type { ChatMessage, LLM } from "../llms";
 import { type BaseChatStore } from "../storage/chat-store";
 import { BaseChatStoreMemory, DEFAULT_TOKEN_LIMIT_RATIO } from "./base";
 
@@ -34,7 +34,7 @@ export class ChatMemoryBuffer<
   }
 
   getMessages(
-    input?: MessageContent | undefined,
+    transientMessages?: ChatMessage<AdditionalMessageOptions>[] | undefined,
     initialTokenCount: number = 0,
   ) {
     const messages = this.getAllMessages();
@@ -43,16 +43,22 @@ export class ChatMemoryBuffer<
       throw new Error("Initial token count exceeds token limit");
     }
 
-    let messageCount = messages.length;
-    let currentMessages = messages.slice(-messageCount);
-    let tokenCount = this._tokenCountForMessages(messages) + initialTokenCount;
+    // Add input messages as transient messages
+    const messagesWithInput = transientMessages
+      ? [...transientMessages, ...messages]
+      : messages;
+
+    let messageCount = messagesWithInput.length;
+    let currentMessages = messagesWithInput.slice(-messageCount);
+    let tokenCount =
+      this._tokenCountForMessages(messagesWithInput) + initialTokenCount;
 
     while (tokenCount > this.tokenLimit && messageCount > 1) {
       messageCount -= 1;
-      if (messages.at(-messageCount)!.role === "assistant") {
+      if (messagesWithInput.at(-messageCount)!.role === "assistant") {
         messageCount -= 1;
       }
-      currentMessages = messages.slice(-messageCount);
+      currentMessages = messagesWithInput.slice(-messageCount);
       tokenCount =
         this._tokenCountForMessages(currentMessages) + initialTokenCount;
     }
@@ -60,6 +66,6 @@ export class ChatMemoryBuffer<
     if (tokenCount > this.tokenLimit && messageCount <= 0) {
       return [];
     }
-    return messages.slice(-messageCount);
+    return messagesWithInput.slice(-messageCount);
   }
 }

packages/core/src/memory/summary-memory.ts

@@ -114,18 +114,22 @@ export class ChatSummaryMemoryBuffer extends BaseMemory {
     }
   }
 
-  private calcCurrentRequestMessages() {
-    // TODO: check order: currently, we're sending:
+  private calcCurrentRequestMessages(transientMessages?: ChatMessage[]) {
+    // currently, we're sending:
     // system messages first, then transient messages and then the messages that describe the conversation so far
-    return [...this.systemMessages, ...this.calcConversationMessages(true)];
+    return [
+      ...this.systemMessages,
+      ...(transientMessages ? transientMessages : []),
+      ...this.calcConversationMessages(true),
+    ];
   }
 
   reset() {
     this.messages = [];
   }
 
-  async getMessages(): Promise<ChatMessage[]> {
-    const requestMessages = this.calcCurrentRequestMessages();
+  async getMessages(transientMessages?: ChatMessage[]): Promise<ChatMessage[]> {
+    const requestMessages = this.calcCurrentRequestMessages(transientMessages);
 
     // get tokens of current request messages and the transient messages
     const tokens = requestMessages.reduce(
@@ -149,7 +153,7 @@ export class ChatSummaryMemoryBuffer extends BaseMemory {
       // TODO: we still might have too many tokens
       // e.g. too large system messages or transient messages
       // how should we deal with that?
-      return this.calcCurrentRequestMessages();
+      return this.calcCurrentRequestMessages(transientMessages);
     }
     return requestMessages;
   }

packages/core/tests/memory/chat-memory-buffer.test.ts

@@ -0,0 +1,74 @@
+import { Settings } from "@llamaindex/core/global";
+import type { ChatMessage } from "@llamaindex/core/llms";
+import { ChatMemoryBuffer } from "@llamaindex/core/memory";
+import { beforeEach, describe, expect, test } from "vitest";
+
+describe("ChatMemoryBuffer", () => {
+  beforeEach(() => {
+    // Mock the Settings.llm
+    (Settings.llm as any) = {
+      metadata: {
+        contextWindow: 1000,
+      },
+    };
+  });
+
+  test("constructor initializes with custom token limit", () => {
+    const buffer = new ChatMemoryBuffer({ tokenLimit: 500 });
+    expect(buffer.tokenLimit).toBe(500);
+  });
+
+  test("getMessages returns all messages when under token limit", () => {
+    const messages: ChatMessage[] = [
+      { role: "user", content: "Hello" },
+      { role: "assistant", content: "Hi there!" },
+      { role: "user", content: "How are you?" },
+    ];
+    const buffer = new ChatMemoryBuffer({
+      tokenLimit: 1000,
+      chatHistory: messages,
+    });
+
+    const result = buffer.getMessages();
+    expect(result).toEqual(messages);
+  });
+
+  test("getMessages truncates messages when over token limit", () => {
+    const messages: ChatMessage[] = [
+      { role: "user", content: "This is a long message" },
+      { role: "assistant", content: "This is also a long reply" },
+      { role: "user", content: "Short" },
+    ];
+    const buffer = new ChatMemoryBuffer({
+      tokenLimit: 5, // limit to only allow the last message
+      chatHistory: messages,
+    });
+
+    const result = buffer.getMessages();
+    expect(result).toEqual([{ role: "user", content: "Short" }]);
+  });
+
+  test("getMessages handles input messages", () => {
+    const storedMessages: ChatMessage[] = [
+      { role: "user", content: "Hello" },
+      { role: "assistant", content: "Hi there!" },
+    ];
+    const buffer = new ChatMemoryBuffer({
+      tokenLimit: 50,
+      chatHistory: storedMessages,
+    });
+
+    const inputMessages: ChatMessage[] = [
+      { role: "user", content: "New message" },
+    ];
+    const result = buffer.getMessages(inputMessages);
+    expect(result).toEqual([...inputMessages, ...storedMessages]);
+  });
+
+  test("getMessages throws error when initial token count exceeds limit", () => {
+    const buffer = new ChatMemoryBuffer({ tokenLimit: 10 });
+    expect(() => buffer.getMessages(undefined, 20)).toThrow(
+      "Initial token count exceeds token limit",
+    );
+  });
+});

packages/experimental/CHANGELOG.md

@@ -1,5 +1,21 @@
 # @llamaindex/experimental
 
+## 0.0.85
+
+### Patch Changes
+
+- Updated dependencies [23bcc37]
+  - llamaindex@0.6.7
+
+## 0.0.84
+
+### Patch Changes
+
+- Updated dependencies [d902cc3]
+- Updated dependencies [025ffe6]
+- Updated dependencies [a659574]
+  - llamaindex@0.6.6
+
 ## 0.0.83
 
 ### Patch Changes

packages/experimental/package.json

@@ -1,7 +1,7 @@
 {
   "name": "@llamaindex/experimental",
   "description": "Experimental package for LlamaIndexTS",
-  "version": "0.0.83",
+  "version": "0.0.85",
   "type": "module",
   "types": "dist/type/index.d.ts",
   "main": "dist/cjs/index.js",

packages/llamaindex/CHANGELOG.md

@@ -1,5 +1,25 @@
 # llamaindex
 
+## 0.6.7
+
+### Patch Changes
+
+- 23bcc37: fix: add `serializer` in doc store
+
+  `PostgresDocumentStore` now will not use JSON.stringify for better performance
+
+## 0.6.6
+
+### Patch Changes
+
+- d902cc3: Fix context not being sent using ContextChatEngine
+- 025ffe6: fix: update `PostgresKVStore` constructor params
+- a659574: Adds upstash vector store as a storage
+- Updated dependencies [d902cc3]
+  - @llamaindex/core@0.2.5
+  - @llamaindex/openai@0.1.7
+  - @llamaindex/groq@0.0.6
+
 ## 0.6.5
 
 ### Patch Changes

packages/llamaindex/e2e/examples/cloudflare-worker-agent/CHANGELOG.md

@@ -1,5 +1,21 @@
 # @llamaindex/cloudflare-worker-agent-test
 
+## 0.0.69
+
+### Patch Changes
+
+- Updated dependencies [23bcc37]
+  - llamaindex@0.6.7
+
+## 0.0.68
+
+### Patch Changes
+
+- Updated dependencies [d902cc3]
+- Updated dependencies [025ffe6]
+- Updated dependencies [a659574]
+  - llamaindex@0.6.6
+
 ## 0.0.67
 
 ### Patch Changes

packages/llamaindex/e2e/examples/cloudflare-worker-agent/package.json

@@ -1,6 +1,6 @@
 {
   "name": "@llamaindex/cloudflare-worker-agent-test",
-  "version": "0.0.67",
+  "version": "0.0.69",
   "type": "module",
   "private": true,
   "scripts": {

packages/llamaindex/e2e/examples/nextjs-agent/CHANGELOG.md

@@ -1,5 +1,21 @@
 # @llamaindex/next-agent-test
 
+## 0.1.69
+
+### Patch Changes
+
+- Updated dependencies [23bcc37]
+  - llamaindex@0.6.7
+
+## 0.1.68
+
+### Patch Changes
+
+- Updated dependencies [d902cc3]
+- Updated dependencies [025ffe6]
+- Updated dependencies [a659574]
+  - llamaindex@0.6.6
+
 ## 0.1.67
 
 ### Patch Changes

packages/llamaindex/e2e/examples/nextjs-agent/package.json

@@ -1,6 +1,6 @@
 {
   "name": "@llamaindex/next-agent-test",
-  "version": "0.1.67",
+  "version": "0.1.69",
   "private": true,
   "scripts": {
     "dev": "next dev",

packages/llamaindex/e2e/examples/nextjs-edge-runtime/CHANGELOG.md

@@ -1,5 +1,21 @@
 # test-edge-runtime
 
+## 0.1.68
+
+### Patch Changes
+
+- Updated dependencies [23bcc37]
+  - llamaindex@0.6.7
+
+## 0.1.67
+
+### Patch Changes
+
+- Updated dependencies [d902cc3]
+- Updated dependencies [025ffe6]
+- Updated dependencies [a659574]
+  - llamaindex@0.6.6
+
 ## 0.1.66
 
 ### Patch Changes

packages/llamaindex/e2e/examples/nextjs-edge-runtime/package.json

@@ -1,6 +1,6 @@
 {
   "name": "@llamaindex/nextjs-edge-runtime-test",
-  "version": "0.1.66",
+  "version": "0.1.68",
   "private": true,
   "scripts": {
     "dev": "next dev",

packages/llamaindex/e2e/examples/nextjs-node-runtime/CHANGELOG.md

@@ -1,5 +1,21 @@
 # @llamaindex/next-node-runtime
 
+## 0.0.50
+
+### Patch Changes
+
+- Updated dependencies [23bcc37]
+  - llamaindex@0.6.7
+
+## 0.0.49
+
+### Patch Changes
+
+- Updated dependencies [d902cc3]
+- Updated dependencies [025ffe6]
+- Updated dependencies [a659574]
+  - llamaindex@0.6.6
+
 ## 0.0.48
 
 ### Patch Changes

packages/llamaindex/e2e/examples/nextjs-node-runtime/package.json

@@ -1,6 +1,6 @@
 {
   "name": "@llamaindex/next-node-runtime-test",
-  "version": "0.0.48",
+  "version": "0.0.50",
   "private": true,
   "scripts": {
     "dev": "next dev",

packages/llamaindex/e2e/examples/waku-query-engine/CHANGELOG.md

@@ -1,5 +1,21 @@
 # @llamaindex/waku-query-engine-test
 
+## 0.0.69
+
+### Patch Changes
+
+- Updated dependencies [23bcc37]
+  - llamaindex@0.6.7
+
+## 0.0.68
+
+### Patch Changes
+
+- Updated dependencies [d902cc3]
+- Updated dependencies [025ffe6]
+- Updated dependencies [a659574]
+  - llamaindex@0.6.6
+
 ## 0.0.67
 
 ### Patch Changes

packages/llamaindex/e2e/examples/waku-query-engine/package.json

@@ -1,6 +1,6 @@
 {
   "name": "@llamaindex/waku-query-engine-test",
-  "version": "0.0.67",
+  "version": "0.0.69",
   "type": "module",
   "private": true,
   "scripts": {

packages/llamaindex/package.json

@@ -1,6 +1,6 @@
 {
   "name": "llamaindex",
-  "version": "0.6.5",
+  "version": "0.6.7",
   "license": "MIT",
   "type": "module",
   "keywords": [
@@ -43,6 +43,7 @@
     "@types/node": "^22.5.1",
     "@types/papaparse": "^5.3.14",
     "@types/pg": "^8.11.8",
+    "@upstash/vector": "^1.1.5",
     "@zilliz/milvus2-sdk-node": "^2.4.6",
     "ajv": "^8.17.1",
     "assemblyai": "^4.7.0",

packages/llamaindex/src/agent/base.ts

@@ -356,9 +356,8 @@ export abstract class AgentRunner<
     let chatHistory: ChatMessage<AdditionalMessageOptions>[] = [];
 
     if (params.chatHistory instanceof BaseMemory) {
-      chatHistory = (await params.chatHistory.getMessages(
-        params.message,
-      )) as ChatMessage<AdditionalMessageOptions>[];
+      chatHistory =
+        (await params.chatHistory.getMessages()) as ChatMessage<AdditionalMessageOptions>[];
     } else {
       chatHistory =
         params.chatHistory as ChatMessage<AdditionalMessageOptions>[];

packages/llamaindex/src/engines/chat/CondenseQuestionChatEngine.ts

@@ -78,9 +78,7 @@ export class CondenseQuestionChatEngine
   }
 
   private async condenseQuestion(chatHistory: BaseMemory, question: string) {
-    const chatHistoryStr = messagesToHistory(
-      await chatHistory.getMessages(question),
-    );
+    const chatHistoryStr = messagesToHistory(await chatHistory.getMessages());
 
     return this.llm.complete({
       prompt: this.condenseMessagePrompt.format({
@@ -103,7 +101,7 @@ export class CondenseQuestionChatEngine
       ? new ChatMemoryBuffer({
           chatHistory:
             params.chatHistory instanceof BaseMemory
-              ? await params.chatHistory.getMessages(message)
+              ? await params.chatHistory.getMessages()
               : params.chatHistory,
         })
       : this.chatHistory;

packages/llamaindex/src/engines/chat/ContextChatEngine.ts

@@ -92,7 +92,7 @@ export class ContextChatEngine extends PromptMixin implements ChatEngine {
       ? new ChatMemoryBuffer({
           chatHistory:
             params.chatHistory instanceof BaseMemory
-              ? await params.chatHistory.getMessages(message)
+              ? await params.chatHistory.getMessages()
               : params.chatHistory,
         })
       : this.chatHistory;
@@ -139,7 +139,7 @@ export class ContextChatEngine extends PromptMixin implements ChatEngine {
     const textOnly = extractText(message);
     const context = await this.contextGenerator.generate(textOnly);
     const systemMessage = this.prependSystemPrompt(context.message);
-    const messages = await chatHistory.getMessages(systemMessage.content);
+    const messages = await chatHistory.getMessages([systemMessage]);
     return { nodes: context.nodes, messages };
   }
 

packages/llamaindex/src/engines/chat/SimpleChatEngine.ts

@@ -40,7 +40,7 @@ export class SimpleChatEngine implements ChatEngine {
       ? new ChatMemoryBuffer({
           chatHistory:
             params.chatHistory instanceof BaseMemory
-              ? await params.chatHistory.getMessages(message)
+              ? await params.chatHistory.getMessages()
               : params.chatHistory,
         })
       : this.chatHistory;
@@ -48,7 +48,7 @@ export class SimpleChatEngine implements ChatEngine {
 
     if (stream) {
       const stream = await this.llm.chat({
-        messages: await chatHistory.getMessages(params.message),
+        messages: await chatHistory.getMessages(),
         stream: true,
       });
       return streamConverter(
@@ -66,7 +66,7 @@ export class SimpleChatEngine implements ChatEngine {
 
     const response = await this.llm.chat({
       stream: false,
-      messages: await chatHistory.getMessages(params.message),
+      messages: await chatHistory.getMessages(),
     });
     chatHistory.put(response.message);
     return EngineResponse.fromChatResponse(response);

packages/llamaindex/src/ingestion/IngestionCache.ts

@@ -1,7 +1,11 @@
 import type { BaseNode, TransformComponent } from "@llamaindex/core/schema";
 import { MetadataMode } from "@llamaindex/core/schema";
 import { createSHA256 } from "@llamaindex/env";
-import { docToJson, jsonToDoc } from "../storage/docStore/utils.js";
+import {
+  docToJson,
+  jsonSerializer,
+  jsonToDoc,
+} from "../storage/docStore/utils.js";
 import { SimpleKVStore } from "../storage/kvStore/SimpleKVStore.js";
 import type { BaseKVStore } from "../storage/kvStore/types.js";
 
@@ -53,7 +57,7 @@ export class IngestionCache {
 
   async put(hash: string, nodes: BaseNode[]) {
     const val = {
-      [this.nodesKey]: nodes.map((node) => docToJson(node)),
+      [this.nodesKey]: nodes.map((node) => docToJson(node, jsonSerializer)),
     };
     await this.cache.put(hash, val, this.collection);
   }
@@ -63,6 +67,8 @@ export class IngestionCache {
     if (!json || !json[this.nodesKey] || !Array.isArray(json[this.nodesKey])) {
       return undefined;
     }
-    return json[this.nodesKey].map((doc: any) => jsonToDoc(doc));
+    return json[this.nodesKey].map((doc: any) =>
+      jsonToDoc(doc, jsonSerializer),
+    );
   }
 }

packages/llamaindex/src/storage/docStore/KVDocumentStore.ts

@@ -29,7 +29,7 @@ export class KVDocumentStore extends BaseDocumentStore {
     for (const key in jsonDict) {
       const value = jsonDict[key];
       if (isValidDocJson(value)) {
-        docs[key] = jsonToDoc(value);
+        docs[key] = jsonToDoc(value, this.serializer);
       } else {
         console.warn(`Invalid JSON for docId ${key}`);
       }
@@ -52,7 +52,7 @@ export class KVDocumentStore extends BaseDocumentStore {
         );
       }
       const nodeKey = doc.id_;
-      const data = docToJson(doc);
+      const data = docToJson(doc, this.serializer);
       await this.kvstore.put(nodeKey, data, this.nodeCollection);
       const metadata: DocMetaData = { docHash: doc.hash };
 
@@ -94,7 +94,7 @@ export class KVDocumentStore extends BaseDocumentStore {
     if (!isValidDocJson(json)) {
       throw new Error(`Invalid JSON for docId ${docId}`);
     }
-    return jsonToDoc(json);
+    return jsonToDoc(json, this.serializer);
   }
 
   async getRefDocInfo(refDocId: string): Promise<RefDocInfo | undefined> {

packages/llamaindex/src/storage/docStore/PostgresDocumentStore.ts

@@ -1,19 +1,32 @@
 import { DEFAULT_NAMESPACE } from "@llamaindex/core/global";
-import { PostgresKVStore } from "../kvStore/PostgresKVStore.js";
+import {
+  PostgresKVStore,
+  type PostgresKVStoreConfig,
+} from "../kvStore/PostgresKVStore.js";
 import { KVDocumentStore } from "./KVDocumentStore.js";
+import { noneSerializer } from "./utils.js";
 
 const DEFAULT_TABLE_NAME = "llamaindex_doc_store";
 
+export type PostgresDocumentStoreConfig = PostgresKVStoreConfig & {
+  namespace?: string;
+};
+
 export class PostgresDocumentStore extends KVDocumentStore {
-  constructor(config?: {
-    schemaName?: string;
-    tableName?: string;
-    connectionString?: string;
-    namespace?: string;
-  }) {
+  serializer = noneSerializer;
+
+  constructor(config?: PostgresDocumentStoreConfig) {
     const kvStore = new PostgresKVStore({
       schemaName: config?.schemaName,
       tableName: config?.tableName || DEFAULT_TABLE_NAME,
+      ...(config && "clientConfig" in config
+        ? { clientConfig: config.clientConfig }
+        : config && "client" in config
+          ? {
+              client: config.client,
+              shouldConnect: config.shouldConnect ?? false,
+            }
+          : {}),
     });
     const namespace = config?.namespace || DEFAULT_NAMESPACE;
     super(kvStore, namespace);

packages/llamaindex/src/storage/docStore/types.ts

@@ -3,6 +3,7 @@ import {
   DEFAULT_PERSIST_DIR,
 } from "@llamaindex/core/global";
 import { BaseNode } from "@llamaindex/core/schema";
+import { jsonSerializer, type Serializer } from "./utils.js";
 
 const defaultPersistPath = `${DEFAULT_PERSIST_DIR}/${DEFAULT_DOC_STORE_PERSIST_FILENAME}`;
 
@@ -12,6 +13,8 @@ export interface RefDocInfo {
 }
 
 export abstract class BaseDocumentStore {
+  serializer: Serializer<any> = jsonSerializer;
+
   // Save/load
   persist(persistPath: string = defaultPersistPath): void {
     // Persist the docstore to a file.

packages/llamaindex/src/storage/docStore/utils.ts

@@ -4,12 +4,35 @@ import { Document, ObjectType, TextNode } from "@llamaindex/core/schema";
 const TYPE_KEY = "__type__";
 const DATA_KEY = "__data__";
 
-type DocJson = {
-  [TYPE_KEY]: ObjectType;
-  [DATA_KEY]: string;
+export interface Serializer<T> {
+  toPersistence(data: Record<string, unknown>): T;
+  fromPersistence(data: T): Record<string, unknown>;
+}
+
+export const jsonSerializer: Serializer<string> = {
+  toPersistence(data) {
+    return JSON.stringify(data);
+  },
+  fromPersistence(data) {
+    return JSON.parse(data);
+  },
 };
 
-export function isValidDocJson(docJson: any): docJson is DocJson {
+export const noneSerializer: Serializer<Record<string, unknown>> = {
+  toPersistence(data) {
+    return data;
+  },
+  fromPersistence(data) {
+    return data;
+  },
+};
+
+type DocJson<Data> = {
+  [TYPE_KEY]: ObjectType;
+  [DATA_KEY]: Data;
+};
+
+export function isValidDocJson(docJson: any): docJson is DocJson<unknown> {
   return (
     typeof docJson === "object" &&
     docJson !== null &&
@@ -18,16 +41,22 @@ export function isValidDocJson(docJson: any): docJson is DocJson {
   );
 }
 
-export function docToJson(doc: BaseNode): DocJson {
+export function docToJson(
+  doc: BaseNode,
+  serializer: Serializer<unknown>,
+): DocJson<unknown> {
   return {
-    [DATA_KEY]: JSON.stringify(doc.toJSON()),
+    [DATA_KEY]: serializer.toPersistence(doc.toJSON()),
     [TYPE_KEY]: doc.type,
   };
 }
 
-export function jsonToDoc(docDict: DocJson): BaseNode {
+export function jsonToDoc<Data>(
+  docDict: DocJson<Data>,
+  serializer: Serializer<Data>,
+): BaseNode {
   const docType = docDict[TYPE_KEY];
-  const dataDict = JSON.parse(docDict[DATA_KEY]);
+  const dataDict = serializer.fromPersistence(docDict[DATA_KEY]) as any;
   let doc: BaseNode;
 
   if (docType === ObjectType.DOCUMENT) {

packages/llamaindex/src/storage/indexStore/PostgresIndexStore.ts

@@ -1,19 +1,29 @@
 import { DEFAULT_NAMESPACE } from "@llamaindex/core/global";
-import { PostgresKVStore } from "../kvStore/PostgresKVStore.js";
+import {
+  PostgresKVStore,
+  type PostgresKVStoreConfig,
+} from "../kvStore/PostgresKVStore.js";
 import { KVIndexStore } from "./KVIndexStore.js";
 
 const DEFAULT_TABLE_NAME = "llamaindex_index_store";
 
+export type PostgresIndexStoreConfig = PostgresKVStoreConfig & {
+  namespace?: string;
+};
+
 export class PostgresIndexStore extends KVIndexStore {
-  constructor(config?: {
-    schemaName?: string;
-    tableName?: string;
-    connectionString?: string;
-    namespace?: string;
-  }) {
+  constructor(config?: PostgresIndexStoreConfig) {
     const kvStore = new PostgresKVStore({
       schemaName: config?.schemaName,
       tableName: config?.tableName || DEFAULT_TABLE_NAME,
+      ...(config && "clientConfig" in config
+        ? { clientConfig: config.clientConfig }
+        : config && "client" in config
+          ? {
+              client: config.client,
+              shouldConnect: config.shouldConnect ?? false,
+            }
+          : {}),
     });
     const namespace = config?.namespace || DEFAULT_NAMESPACE;
     super(kvStore, namespace);

packages/llamaindex/src/storage/kvStore/PostgresKVStore.ts

@@ -7,41 +7,76 @@ export type DataType = Record<string, Record<string, any>>;
 const DEFAULT_SCHEMA_NAME = "public";
 const DEFAULT_TABLE_NAME = "llamaindex_kv_store";
 
+export type PostgresKVStoreBaseConfig = {
+  schemaName?: string | undefined;
+  tableName?: string | undefined;
+};
+
+export type PostgresKVStoreClientConfig =
+  | {
+      /**
+       * Client configuration options for the pg client.
+       *
+       * {@link https://node-postgres.com/apis/client#new-client PostgresSQL Client API}
+       */
+      clientConfig?: pg.ClientConfig | undefined;
+    }
+  | {
+      /**
+       * A pg client or pool client instance.
+       * If provided, make sure it is not connected to the database yet, or it will throw an error.
+       */
+      shouldConnect?: boolean | undefined;
+      client?: pg.Client | pg.PoolClient;
+    };
+
+export type PostgresKVStoreConfig = PostgresKVStoreBaseConfig &
+  PostgresKVStoreClientConfig;
+
 export class PostgresKVStore extends BaseKVStore {
   private schemaName: string;
   private tableName: string;
-  private connectionString: string | undefined = undefined;
-  private db?: pg.Client;
 
-  constructor(config?: {
-    schemaName?: string | undefined;
-    tableName?: string | undefined;
-    connectionString?: string | undefined;
-  }) {
+  private isDBConnected: boolean = false;
+  private clientConfig: pg.ClientConfig | undefined = undefined;
+  private db?: pg.ClientBase | undefined = undefined;
+
+  constructor(config?: PostgresKVStoreConfig) {
     super();
     this.schemaName = config?.schemaName || DEFAULT_SCHEMA_NAME;
     this.tableName = config?.tableName || DEFAULT_TABLE_NAME;
-    this.connectionString = config?.connectionString;
-  }
-
-  private async getDb(): Promise<pg.Client> {
-    if (!this.db) {
-      try {
-        const pg = await import("pg");
-        const { Client } = pg.default ? pg.default : pg;
-        const db = new Client({ connectionString: this.connectionString });
-        await db.connect();
-        await this.checkSchema(db);
-        this.db = db;
-      } catch (err) {
-        console.error(err);
-        return Promise.reject(err instanceof Error ? err : new Error(`${err}`));
+    if (config) {
+      if ("clientConfig" in config) {
+        this.clientConfig = config.clientConfig;
+      } else if ("client" in config) {
+        this.isDBConnected =
+          config?.shouldConnect !== undefined ? !config.shouldConnect : false;
+        this.db = config.client;
       }
     }
-    return Promise.resolve(this.db);
   }
 
-  private async checkSchema(db: pg.Client) {
+  private async getDb(): Promise<pg.ClientBase> {
+    if (!this.db) {
+      const pg = await import("pg");
+      const { Client } = pg.default ? pg.default : pg;
+      const db = new Client({ ...this.clientConfig });
+      await db.connect();
+      this.isDBConnected = true;
+      this.db = db;
+    }
+    if (this.db && !this.isDBConnected) {
+      await this.db.connect();
+      this.isDBConnected = true;
+    }
+    this.db.on("end", () => {
+      this.isDBConnected = false;
+    });
+    await this.checkSchema(this.db);
+    return this.db;
+  }
+
+  private async checkSchema(db: pg.ClientBase) {
     await db.query(`CREATE SCHEMA IF NOT EXISTS ${this.schemaName}`);
     const tbl = `CREATE TABLE IF NOT EXISTS ${this.schemaName}.${this.tableName} (
       id uuid DEFAULT gen_random_uuid() PRIMARY KEY,
@@ -97,7 +132,7 @@ export class PostgresKVStore extends BaseKVStore {
       const sql = `SELECT * FROM ${this.schemaName}.${this.tableName} WHERE key = $1 AND collection = $2`;
       const result = await db.query(sql, [key, collection]);
       await db.query("COMMIT");
-      return result.rows[0].value;
+      return result.rows[0]?.value;
     } catch (error) {
       await db.query("ROLLBACK");
       throw error;

packages/llamaindex/src/vector-store/UpstashVectorStore.ts

@@ -0,0 +1,237 @@
+import {
+  VectorStoreBase,
+  type IEmbedModel,
+  type MetadataFilter,
+  type MetadataFilters,
+  type VectorStoreNoEmbedModel,
+  type VectorStoreQuery,
+  type VectorStoreQueryResult,
+} from "./types.js";
+
+import type { BaseNode, Metadata, TextNode } from "@llamaindex/core/schema";
+import { getEnv } from "@llamaindex/env";
+import { Index } from "@upstash/vector";
+import { metadataDictToNode, nodeToMetadata } from "./utils.js";
+
+type UpstashParams = {
+  namespace?: string;
+  token?: string;
+  endpoint?: string;
+  maxBatchSize?: number;
+} & IEmbedModel;
+
+/**
+ * Provides support for writing and querying vector data in Upstash.
+ */
+export class UpstashVectorStore
+  extends VectorStoreBase
+  implements VectorStoreNoEmbedModel
+{
+  storesText: boolean = true;
+
+  private db: Index;
+  private maxBatchSize: number;
+  namespace: string;
+
+  /**
+   * @param namespace namespace to use
+   * @param token upstash vector token. if not set, `process.env.UPSTASH_VECTOR_REST_TOKEN` is used.
+   * @param endpoint upstash vector endpoint. If not set, `process.env.UPSTASH_VECTOR_REST_URL` is used.
+   * @param maxBatchSize maximum number of vectors upserted at once. Default is 1000.
+   *
+   * @example
+   * ```ts
+   * const vectorStore = new UpstashVectorStore({ namespace: "my-namespace" })
+   * ```
+   */
+  constructor(params?: UpstashParams) {
+    super(params?.embedModel);
+    this.namespace = params?.namespace ?? "";
+    this.maxBatchSize = params?.maxBatchSize ?? 1000;
+    const token = params?.token ?? getEnv("UPSTASH_VECTOR_REST_TOKEN");
+    const endpoint = params?.endpoint ?? getEnv("UPSTASH_VECTOR_REST_URL");
+
+    if (!token) {
+      throw new Error(
+        "Must specify UPSTASH_VECTOR_REST_TOKEN via env variable.",
+      );
+    }
+    if (!endpoint) {
+      throw new Error("Must specify UPSTASH_VECTOR_REST_URL via env variable.");
+    }
+    this.db = new Index({ token, url: endpoint });
+  }
+
+  private async getDb(): Promise<Index> {
+    if (!this.db) {
+      const { Index } = await import("@upstash/vector");
+      this.db = new Index();
+    }
+
+    return this.db;
+  }
+
+  /**
+   * Connects to the database specified in environment vars.
+   * @returns A connection to the database, or the error encountered while connecting/setting up.
+   */
+  client(): Promise<Index> {
+    return this.getDb();
+  }
+
+  /**
+   * Adds vector record(s) to the table.
+   * @param embeddingResults The Nodes to be inserted, optionally including metadata tuples.
+   * @returns ids of the embeddings (infered from the id_ field of embeddingResults objects)
+   */
+  async add(embeddingResults: BaseNode<Metadata>[]): Promise<string[]> {
+    if (embeddingResults.length == 0) {
+      return [];
+    }
+
+    const nodes = embeddingResults.map(this.nodeToRecord);
+    const result = await this.upsertInBatches(nodes);
+    if (result != "OK") {
+      throw new Error("Failed to save chunk");
+    }
+    return nodes.map((node) => node.id);
+  }
+
+  /**
+   * Adds plain text record(s) to the table. Upstash take cares of embedding conversion.
+   * @param text The Nodes to be inserted, optionally including metadata tuples.
+   * @returns ids of the embeddings (infered from the id_ field of embeddingResults objects)
+   */
+  async addPlainText(text: TextNode<Metadata>[]): Promise<string[]> {
+    if (text.length == 0) {
+      return [];
+    }
+
+    const nodes = text.map(this.textNodeToRecord);
+    const result = await this.upsertInBatches(nodes);
+    if (result != "OK") {
+      throw new Error("Failed to save chunk");
+    }
+    return nodes.map((node) => node.id);
+  }
+
+  private async upsertInBatches(
+    nodes:
+      | ReturnType<UpstashVectorStore["textNodeToRecord"]>[]
+      | ReturnType<UpstashVectorStore["nodeToRecord"]>[],
+  ) {
+    const promises: Promise<string>[] = [];
+    for (let i = 0; i < nodes.length; i += this.maxBatchSize) {
+      const batch = nodes.slice(i, i + this.maxBatchSize);
+      promises.push(this.db.upsert(batch, { namespace: this.namespace }));
+    }
+    const results = await Promise.all(promises);
+    return results.every((result) => result === "OK") ? "OK" : "NOT-OK";
+  }
+
+  /**
+   * Deletes a single record from the database by id.
+   * NOTE: Uses the collection property controlled by setCollection/getCollection.
+   * @param refDocId Unique identifier for the record to delete.
+   * @returns Promise that resolves if the delete query did not throw an error.
+   */
+  async delete(refDocId: string): Promise<void> {
+    await this.db.namespace(this.namespace).delete(refDocId);
+  }
+
+  /**
+   * Deletes a single record from the database by id.
+   * NOTE: Uses the collection property controlled by setCollection/getCollection.
+   * @param refDocId Unique identifier for the record to delete.
+   * @param deleteKwargs Required by VectorStore interface.  Currently ignored.
+   * @returns Promise that resolves if the delete query did not throw an error.
+   */
+  async deleteMany(refDocId: string[]): Promise<void> {
+    await this.db.namespace(this.namespace).delete(refDocId);
+  }
+
+  /**
+   * Query the vector store for the closest matching data to the query embeddings
+   * @param query The VectorStoreQuery to be used
+   * @param options Required by VectorStore interface.  Currently ignored.
+   * @returns Zero or more Document instances with data from the vector store.
+   */
+  async query(
+    query: VectorStoreQuery,
+    _options?: any,
+  ): Promise<VectorStoreQueryResult> {
+    const filter = this.toUpstashFilter(query.filters);
+
+    const defaultOptions: any = {
+      vector: query.queryEmbedding,
+      topK: query.similarityTopK,
+      includeVectors: true,
+      includeMetadata: true,
+      filter,
+    };
+
+    const db = this.db;
+    const results = await db.query(defaultOptions, {
+      namespace: this.namespace,
+    });
+
+    const nodes = results.map((result) => {
+      const node = metadataDictToNode(result.metadata as Record<string, any>, {
+        fallback: {
+          id: result.id,
+          metadata: result.metadata,
+          embedding: result.vector,
+        },
+      });
+      return node;
+    });
+
+    const ret = {
+      nodes: nodes,
+      similarities: results.map((row) => row.score || 999),
+      ids: results.map((row) => String(row.id)),
+    };
+
+    return ret;
+  }
+
+  toFilterString(filter: MetadataFilter) {
+    return `${filter.key} ${filter.operator} ${filter.value}`;
+  }
+
+  toUpstashFilter(stdFilters?: MetadataFilters) {
+    if (!stdFilters?.filters) return;
+
+    for (const item of stdFilters.filters) {
+      if (item.operator === "==") {
+        //@ts-expect-error Upstash equal operator uses only one equal sign, so we have to replace it.
+        item.operator = "=";
+      }
+    }
+
+    const filterStrings = stdFilters.filters.map(this.toFilterString);
+
+    if (filterStrings.length === 1) {
+      return filterStrings[0];
+    }
+    return filterStrings.join(` ${stdFilters.condition ?? "and"} `);
+  }
+
+  nodeToRecord(node: BaseNode<Metadata>) {
+    const id: any = node.id_.length ? node.id_ : null;
+    return {
+      id: id,
+      vector: node.getEmbedding(),
+      metadata: nodeToMetadata(node),
+    };
+  }
+
+  textNodeToRecord(node: TextNode<Metadata>) {
+    const id: any = node.id_.length ? node.id_ : null;
+    return {
+      id,
+      data: node.text,
+      metadata: nodeToMetadata(node),
+    };
+  }
+}

packages/llm/groq/CHANGELOG.md

@@ -1,5 +1,11 @@
 # @llamaindex/groq
 
+## 0.0.6
+
+### Patch Changes
+
+- @llamaindex/openai@0.1.7
+
 ## 0.0.5
 
 ### Patch Changes

packages/llm/groq/package.json

@@ -1,7 +1,7 @@
 {
   "name": "@llamaindex/groq",
   "description": "Groq Adapter for LlamaIndex",
-  "version": "0.0.5",
+  "version": "0.0.6",
   "type": "module",
   "main": "./dist/index.cjs",
   "module": "./dist/index.js",

packages/llm/openai/CHANGELOG.md

@@ -1,5 +1,12 @@
 # @llamaindex/openai
 
+## 0.1.7
+
+### Patch Changes
+
+- Updated dependencies [d902cc3]
+  - @llamaindex/core@0.2.5
+
 ## 0.1.6
 
 ### Patch Changes

packages/llm/openai/package.json

@@ -1,7 +1,7 @@
 {
   "name": "@llamaindex/openai",
   "description": "OpenAI Adapter for LlamaIndex",
-  "version": "0.1.6",
+  "version": "0.1.7",
   "type": "module",
   "main": "./dist/index.cjs",
   "module": "./dist/index.js",

pnpm-lock.yaml

@@ -565,6 +565,9 @@ importers:
       '@types/pg':
         specifier: ^8.11.8
         version: 8.11.8
+      '@upstash/vector':
+        specifier: ^1.1.5
+        version: 1.1.5
       '@zilliz/milvus2-sdk-node':
         specifier: ^2.4.6
         version: 2.4.6
@@ -4511,6 +4514,9 @@ packages:
   '@ungap/structured-clone@1.2.0':
     resolution: {integrity: sha512-zuVdFrMJiuCDQUMCzQaD6KL28MjnqqN8XnAqiEq9PNm/hCPTSGfrXCOfwj1ow4LFb/tNymJPwsNbVePc1xFqrQ==}
 
+  '@upstash/vector@1.1.5':
+    resolution: {integrity: sha512-55+Beu/kCwjcnzg6fnMN06v9PYU1lv9NQfQwpjrJAQTH8GOprcRsQeyXBdNHKNzoQvRnVS0ENd5CDgFoljfrAw==}
+
   '@vitejs/plugin-react@4.3.1':
     resolution: {integrity: sha512-m/V2syj5CuVnaxcUJOQRel/Wr31FFXRFlnOoq1TVtkCxsY5veGMTEmpWHndrhB2U8ScHtCQB1e+4hWYExQc6Lg==}
     engines: {node: ^14.18.0 || >=16.0.0}
@@ -16426,6 +16432,8 @@ snapshots:
 
   '@ungap/structured-clone@1.2.0': {}
 
+  '@upstash/vector@1.1.5': {}
+
   '@vitejs/plugin-react@4.3.1(vite@5.4.2(@types/node@22.5.5)(terser@5.32.0))':
     dependencies:
       '@babel/core': 7.25.2