Release 0.6.17 (#1305)

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

apps/docs/CHANGELOG.md

@@ -1,5 +1,33 @@
 # docs
 
+## 0.0.86
+
+### Patch Changes
+
+- Updated dependencies [ee697fb]
+  - llamaindex@0.6.17
+
+## 0.0.85
+
+### Patch Changes
+
+- Updated dependencies [63e9846]
+- Updated dependencies [6f3a31c]
+  - llamaindex@0.6.16
+
+## 0.0.84
+
+### Patch Changes
+
+- Updated dependencies [2a82413]
+  - llamaindex@0.6.15
+
+## 0.0.83
+
+### Patch Changes
+
+- llamaindex@0.6.14
+
 ## 0.0.82
 
 ### Patch Changes

apps/docs/package.json

@@ -1,6 +1,6 @@
 {
   "name": "docs",
-  "version": "0.0.82",
+  "version": "0.0.86",
   "private": true,
   "scripts": {
     "docusaurus": "docusaurus",

examples/agent/large_toolcall_with_gpt4o.ts

@@ -13,7 +13,7 @@ import { FunctionTool, OpenAI, ToolCallOptions } from "llamaindex";
   }
 })();
 
-async function callLLM(init: Partial<OpenAI>) {
+async function callLLM(init: { model: string }) {
   const csvData =
     "Country,Average Height (cm)\nNetherlands,156\nDenmark,158\nNorway,160";
 

examples/metadata-filter/preFilters.ts

@@ -0,0 +1,51 @@
+import {
+  Document,
+  MetadataFilters,
+  Settings,
+  SimpleDocumentStore,
+  VectorStoreIndex,
+  storageContextFromDefaults,
+} from "llamaindex";
+
+async function getDataSource() {
+  const docs = [
+    new Document({ text: "The dog is brown", metadata: { dogId: "1" } }),
+    new Document({ text: "The dog is yellow", metadata: { dogId: "2" } }),
+  ];
+  const storageContext = await storageContextFromDefaults({
+    persistDir: "./cache",
+  });
+  const numberOfDocs = Object.keys(
+    (storageContext.docStore as SimpleDocumentStore).toDict(),
+  ).length;
+  if (numberOfDocs === 0) {
+    return await VectorStoreIndex.fromDocuments(docs, { storageContext });
+  }
+  return await VectorStoreIndex.init({
+    storageContext,
+  });
+}
+
+Settings.callbackManager.on("retrieve-end", (event) => {
+  const { nodes, query } = event.detail;
+  console.log(`${query.query} - Number of retrieved nodes:`, nodes.length);
+});
+
+async function main() {
+  const index = await getDataSource();
+  const filters: MetadataFilters = {
+    filters: [{ key: "dogId", value: "2", operator: "==" }],
+  };
+
+  const retriever = index.asRetriever({ similarityTopK: 3, filters });
+  const queryEngine = index.asQueryEngine({
+    similarityTopK: 3,
+    preFilters: filters,
+  });
+
+  console.log("Retriever and query engine should only retrieve 1 node:");
+  await retriever.retrieve({ query: "Retriever: get dog" });
+  await queryEngine.query({ query: "QueryEngine: get dog" });
+}
+
+void main();

examples/qdrantdb/preFilters.ts

@@ -39,6 +39,12 @@ async function main() {
           dogId: "2",
         },
       }),
+      new Document({
+        text: "The dog is black",
+        metadata: {
+          dogId: "3",
+        },
+      }),
     ];
     console.log("Creating QdrantDB vector store");
     const qdrantVs = new QdrantVectorStore({ url: qdrantUrl, collectionName });
@@ -73,6 +79,42 @@ async function main() {
       query: "What is the color of the dog?",
     });
     console.log("Filter with dogId 2 response:", response.toString());
+
+    console.log("Querying index with dogId !=2: Expected output: Not red");
+    const queryEngineNotDogId2 = index.asQueryEngine({
+      preFilters: {
+        filters: [
+          {
+            key: "dogId",
+            value: "2",
+            operator: "!=",
+          },
+        ],
+      },
+    });
+    const responseNotDogId2 = await queryEngineNotDogId2.query({
+      query: "What is the color of the dog?",
+    });
+    console.log(responseNotDogId2.toString());
+
+    console.log(
+      "Querying index with dogId 2 or 3: Expected output: Red, Black",
+    );
+    const queryEngineIn = index.asQueryEngine({
+      preFilters: {
+        filters: [
+          {
+            key: "dogId",
+            value: ["2", "3"],
+            operator: "in",
+          },
+        ],
+      },
+    });
+    const responseIn = await queryEngineIn.query({
+      query: "List all dogs",
+    });
+    console.log(responseIn.toString());
   } catch (e) {
     console.error(e);
   }

examples/vectorIndexCustomize.ts

@@ -25,12 +25,9 @@ async function main() {
     similarityCutoff: 0.7,
   });
   // TODO: cannot pass responseSynthesizer into retriever query engine
-  const queryEngine = new RetrieverQueryEngine(
-    retriever,
-    undefined,
-    undefined,
-    [nodePostprocessor],
-  );
+  const queryEngine = new RetrieverQueryEngine(retriever, undefined, [
+    nodePostprocessor,
+  ]);
 
   const response = await queryEngine.query({
     query: "What did the author do growing up?",

examples/vectorIndexFromVectorStore.ts

@@ -165,9 +165,7 @@ async function main() {
     });
 
     const responseSynthesizer = getResponseSynthesizer("tree_summarize");
-    return new RetrieverQueryEngine(retriever, responseSynthesizer, {
-      filter,
-    });
+    return new RetrieverQueryEngine(retriever, responseSynthesizer);
   };
 
   // whatever is a key from your metadata

packages/autotool/CHANGELOG.md

@@ -1,5 +1,33 @@
 # @llamaindex/autotool
 
+## 3.0.17
+
+### Patch Changes
+
+- Updated dependencies [ee697fb]
+  - llamaindex@0.6.17
+
+## 3.0.16
+
+### Patch Changes
+
+- Updated dependencies [63e9846]
+- Updated dependencies [6f3a31c]
+  - llamaindex@0.6.16
+
+## 3.0.15
+
+### Patch Changes
+
+- Updated dependencies [2a82413]
+  - llamaindex@0.6.15
+
+## 3.0.14
+
+### Patch Changes
+
+- llamaindex@0.6.14
+
 ## 3.0.13
 
 ### Patch Changes

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

@@ -1,5 +1,37 @@
 # @llamaindex/autotool-01-node-example
 
+## 0.0.26
+
+### Patch Changes
+
+- Updated dependencies [ee697fb]
+  - llamaindex@0.6.17
+  - @llamaindex/autotool@3.0.17
+
+## 0.0.25
+
+### Patch Changes
+
+- Updated dependencies [63e9846]
+- Updated dependencies [6f3a31c]
+  - llamaindex@0.6.16
+  - @llamaindex/autotool@3.0.16
+
+## 0.0.24
+
+### Patch Changes
+
+- Updated dependencies [2a82413]
+  - llamaindex@0.6.15
+  - @llamaindex/autotool@3.0.15
+
+## 0.0.23
+
+### Patch Changes
+
+- llamaindex@0.6.14
+- @llamaindex/autotool@3.0.14
+
 ## 0.0.22
 
 ### 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": "0.0.22"
+  "version": "0.0.26"
 }

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

@@ -1,5 +1,37 @@
 # @llamaindex/autotool-02-next-example
 
+## 0.1.70
+
+### Patch Changes
+
+- Updated dependencies [ee697fb]
+  - llamaindex@0.6.17
+  - @llamaindex/autotool@3.0.17
+
+## 0.1.69
+
+### Patch Changes
+
+- Updated dependencies [63e9846]
+- Updated dependencies [6f3a31c]
+  - llamaindex@0.6.16
+  - @llamaindex/autotool@3.0.16
+
+## 0.1.68
+
+### Patch Changes
+
+- Updated dependencies [2a82413]
+  - llamaindex@0.6.15
+  - @llamaindex/autotool@3.0.15
+
+## 0.1.67
+
+### Patch Changes
+
+- llamaindex@0.6.14
+- @llamaindex/autotool@3.0.14
+
 ## 0.1.66
 
 ### Patch Changes

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

@@ -1,7 +1,7 @@
 {
   "name": "@llamaindex/autotool-02-next-example",
   "private": true,
-  "version": "0.1.66",
+  "version": "0.1.70",
   "scripts": {
     "dev": "next dev",
     "build": "next build",

packages/autotool/package.json

@@ -1,7 +1,7 @@
 {
   "name": "@llamaindex/autotool",
   "type": "module",
-  "version": "3.0.13",
+  "version": "3.0.17",
   "description": "auto transpile your JS function to LLM Agent compatible",
   "files": [
     "dist",

packages/cloud/CHANGELOG.md

@@ -1,5 +1,26 @@
 # @llamaindex/cloud
 
+## 0.2.13
+
+### Patch Changes
+
+- Updated dependencies [ee697fb]
+  - @llamaindex/core@0.2.11
+
+## 0.2.12
+
+### Patch Changes
+
+- Updated dependencies [3489e7d]
+- Updated dependencies [468bda5]
+  - @llamaindex/core@0.2.10
+
+## 0.2.11
+
+### Patch Changes
+
+- 0b20ff9: fix: package.json format
+
 ## 0.2.10
 
 ### Patch Changes

packages/cloud/package.json

@@ -1,6 +1,6 @@
 {
   "name": "@llamaindex/cloud",
-  "version": "0.2.10",
+  "version": "0.2.13",
   "type": "module",
   "license": "MIT",
   "scripts": {
@@ -51,13 +51,13 @@
   "devDependencies": {
     "@hey-api/client-fetch": "^0.2.4",
     "@hey-api/openapi-ts": "^0.53.0",
-    "@llamaindex/core": "workspace:^0.2.8",
-    "@llamaindex/env": "workspace:^0.1.13",
+    "@llamaindex/core": "workspace:*",
+    "@llamaindex/env": "workspace:*",
     "bunchee": "5.3.2"
   },
   "peerDependencies": {
-    "@llamaindex/core": "workspace:^0.2.8",
-    "@llamaindex/env": "workspace:^0.1.13"
+    "@llamaindex/core": "workspace:*",
+    "@llamaindex/env": "workspace:*"
   },
   "dependencies": {
     "magic-bytes.js": "^1.10.0"

packages/community/CHANGELOG.md

@@ -1,5 +1,27 @@
 # @llamaindex/community
 
+## 0.0.46
+
+### Patch Changes
+
+- Updated dependencies [ee697fb]
+  - @llamaindex/core@0.2.11
+
+## 0.0.45
+
+### Patch Changes
+
+- Updated dependencies [3489e7d]
+- Updated dependencies [468bda5]
+  - @llamaindex/core@0.2.10
+
+## 0.0.44
+
+### Patch Changes
+
+- Updated dependencies [b17d439]
+  - @llamaindex/core@0.2.9
+
 ## 0.0.43
 
 ### Patch Changes

packages/community/package.json

@@ -1,7 +1,7 @@
 {
   "name": "@llamaindex/community",
   "description": "Community package for LlamaIndexTS",
-  "version": "0.0.43",
+  "version": "0.0.46",
   "type": "module",
   "types": "dist/type/index.d.ts",
   "main": "dist/cjs/index.js",

packages/core/CHANGELOG.md

@@ -1,5 +1,24 @@
 # @llamaindex/core
 
+## 0.2.11
+
+### Patch Changes
+
+- ee697fb: fix: generate uuid when inserting to Qdrant
+
+## 0.2.10
+
+### Patch Changes
+
+- 3489e7d: fix: num output incorrect in prompt helper
+- 468bda5: fix: correct warning when chunk size smaller than 0
+
+## 0.2.9
+
+### Patch Changes
+
+- b17d439: Fix #1278: resolved issue where the id\_ was not correctly passed as the id when creating a TextNode. As a result, the upsert operation to the vector database was using a generated ID instead of the provided document ID, if available.
+
 ## 0.2.8
 
 ### Patch Changes

packages/core/package.json

@@ -1,7 +1,7 @@
 {
   "name": "@llamaindex/core",
   "type": "module",
-  "version": "0.2.8",
+  "version": "0.2.11",
   "description": "LlamaIndex Core Module",
   "exports": {
     "./agent": {

packages/core/src/indices/prompt-helper.ts

@@ -8,18 +8,16 @@ import {
   Settings,
 } from "../global";
 import type { LLMMetadata } from "../llms";
-import { SentenceSplitter } from "../node-parser";
-import type { PromptTemplate } from "../prompts";
+import { TextSplitter, TokenTextSplitter, truncateText } from "../node-parser";
+import { BasePromptTemplate, PromptTemplate } from "../prompts";
 
 /**
  * Get the empty prompt text given a prompt.
  */
-function getEmptyPromptTxt(prompt: PromptTemplate) {
-  return prompt.format({
-    ...Object.fromEntries(
-      [...prompt.templateVars.keys()].map((key) => [key, ""]),
-    ),
-  });
+function getEmptyPromptTxt(prompt: PromptTemplate): string {
+  return prompt.format(
+    Object.fromEntries([...prompt.templateVars.keys()].map((key) => [key, ""])),
+  );
 }
 
 /**
@@ -35,24 +33,24 @@ export function getBiggestPrompt(prompts: PromptTemplate[]): PromptTemplate {
 }
 
 export type PromptHelperOptions = {
-  contextWindow?: number;
-  numOutput?: number;
-  chunkOverlapRatio?: number;
-  chunkSizeLimit?: number;
-  tokenizer?: Tokenizer;
-  separator?: string;
+  contextWindow?: number | undefined;
+  numOutput?: number | undefined;
+  chunkOverlapRatio?: number | undefined;
+  chunkSizeLimit?: number | undefined;
+  tokenizer?: Tokenizer | undefined;
+  separator?: string | undefined;
 };
 
 /**
  * A collection of helper functions for working with prompts.
  */
 export class PromptHelper {
-  contextWindow = DEFAULT_CONTEXT_WINDOW;
-  numOutput = DEFAULT_NUM_OUTPUTS;
-  chunkOverlapRatio = DEFAULT_CHUNK_OVERLAP_RATIO;
+  contextWindow: number;
+  numOutput: number;
+  chunkOverlapRatio: number;
   chunkSizeLimit: number | undefined;
   tokenizer: Tokenizer;
-  separator = " ";
+  separator: string;
 
   constructor(options: PromptHelperOptions = {}) {
     const {
@@ -72,68 +70,93 @@ export class PromptHelper {
   }
 
   /**
-   * Given a prompt, return the maximum size of the inputs to the prompt.
-   * @param prompt
-   * @returns
+   * Calculate the available context size based on the number of prompt tokens.
    */
-  private getAvailableContextSize(prompt: PromptTemplate) {
-    const emptyPromptText = getEmptyPromptTxt(prompt);
-    const promptTokens = this.tokenizer.encode(emptyPromptText);
-    const numPromptTokens = promptTokens.length;
-
-    return this.contextWindow - numPromptTokens - this.numOutput;
-  }
-
-  /**
-   * Find the maximum size of each chunk given a prompt.
-   */
-  private getAvailableChunkSize(
-    prompt: PromptTemplate,
-    numChunks = 1,
-    padding = 5,
-  ): number {
-    const availableContextSize = this.getAvailableContextSize(prompt);
-
-    const result = Math.floor(availableContextSize / numChunks) - padding;
-
-    if (this.chunkSizeLimit) {
-      return Math.min(this.chunkSizeLimit, result);
-    } else {
-      return result;
+  #getAvailableContextSize(numPromptTokens: number): number {
+    const contextSizeTokens =
+      this.contextWindow - numPromptTokens - this.numOutput;
+    if (contextSizeTokens < 0) {
+      throw new Error(
+        `Calculated available context size ${contextSizeTokens} is not non-negative.`,
+      );
     }
+    return contextSizeTokens;
   }
 
   /**
-   * Creates a text splitter with the correct chunk sizes and overlaps given a prompt.
+   * Calculate the available chunk size based on the prompt and other parameters.
+   */
+  #getAvailableChunkSize<Template extends BasePromptTemplate>(
+    prompt: Template,
+    numChunks: number = 1,
+    padding: number = 5,
+  ): number {
+    let numPromptTokens = 0;
+
+    if (prompt instanceof PromptTemplate) {
+      numPromptTokens = this.tokenizer.encode(getEmptyPromptTxt(prompt)).length;
+    }
+
+    const availableContextSize = this.#getAvailableContextSize(numPromptTokens);
+    let result = Math.floor(availableContextSize / numChunks) - padding;
+
+    if (this.chunkSizeLimit !== undefined) {
+      result = Math.min(this.chunkSizeLimit, result);
+    }
+
+    return result;
+  }
+
+  /**
+   * Creates a text splitter configured to maximally pack the available context window.
    */
   getTextSplitterGivenPrompt(
-    prompt: PromptTemplate,
-    numChunks = 1,
-    padding = DEFAULT_PADDING,
-  ) {
-    const chunkSize = this.getAvailableChunkSize(prompt, numChunks, padding);
-    if (chunkSize === 0) {
-      throw new Error("Got 0 as available chunk size");
+    prompt: BasePromptTemplate,
+    numChunks: number = 1,
+    padding: number = DEFAULT_PADDING,
+  ): TextSplitter {
+    const chunkSize = this.#getAvailableChunkSize(prompt, numChunks, padding);
+    if (chunkSize <= 0) {
+      throw new TypeError(`Chunk size ${chunkSize} is not positive.`);
     }
-    const chunkOverlap = this.chunkOverlapRatio * chunkSize;
-    return new SentenceSplitter({
+    const chunkOverlap = Math.floor(this.chunkOverlapRatio * chunkSize);
+    return new TokenTextSplitter({
+      separator: this.separator,
       chunkSize,
       chunkOverlap,
-      separator: this.separator,
       tokenizer: this.tokenizer,
     });
   }
 
   /**
-   * Repack resplits the strings based on the optimal text splitter.
+   * Truncate text chunks to fit within the available context window.
+   */
+  truncate(
+    prompt: BasePromptTemplate,
+    textChunks: string[],
+    padding: number = DEFAULT_PADDING,
+  ): string[] {
+    const textSplitter = this.getTextSplitterGivenPrompt(
+      prompt,
+      textChunks.length,
+      padding,
+    );
+    return textChunks.map((chunk) => truncateText(chunk, textSplitter));
+  }
+
+  /**
+   * Repack text chunks to better utilize the available context window.
    */
   repack(
-    prompt: PromptTemplate,
+    prompt: BasePromptTemplate,
     textChunks: string[],
-    padding = DEFAULT_PADDING,
-  ) {
+    padding: number = DEFAULT_PADDING,
+  ): string[] {
     const textSplitter = this.getTextSplitterGivenPrompt(prompt, 1, padding);
-    const combinedStr = textChunks.join("\n\n");
+    const combinedStr = textChunks
+      .map((c) => c.trim())
+      .filter((c) => c.length > 0)
+      .join("\n\n");
     return textSplitter.splitText(combinedStr);
   }
 
@@ -154,7 +177,8 @@ export class PromptHelper {
     } = options ?? {};
     return new PromptHelper({
       contextWindow: metadata.contextWindow,
-      numOutput: metadata.maxTokens ?? DEFAULT_NUM_OUTPUTS,
+      // fixme: numOutput is not in LLMMetadata
+      numOutput: DEFAULT_NUM_OUTPUTS,
       chunkOverlapRatio,
       chunkSizeLimit,
       tokenizer,

packages/core/src/node-parser/index.ts

@@ -13,6 +13,7 @@ export { MetadataAwareTextSplitter, NodeParser, TextSplitter } from "./base";
 export { MarkdownNodeParser } from "./markdown";
 export { SentenceSplitter } from "./sentence-splitter";
 export { SentenceWindowNodeParser } from "./sentence-window";
+export { TokenTextSplitter } from "./token-text-splitter";
 export type { SplitterParams } from "./type";
 export {
   splitByChar,
@@ -20,5 +21,6 @@ export {
   splitByRegex,
   splitBySentenceTokenizer,
   splitBySep,
+  truncateText,
 } from "./utils";
 export type { TextSplitterFn } from "./utils";

packages/core/src/node-parser/token-text-splitter.ts

@@ -0,0 +1,206 @@
+import type { Tokenizer } from "@llamaindex/env";
+import { z } from "zod";
+import { DEFAULT_CHUNK_OVERLAP, DEFAULT_CHUNK_SIZE, Settings } from "../global";
+import { MetadataAwareTextSplitter } from "./base";
+import type { SplitterParams } from "./type";
+import { splitByChar, splitBySep } from "./utils";
+
+const DEFAULT_METADATA_FORMAT_LEN = 2;
+
+const tokenTextSplitterSchema = z.object({
+  chunkSize: z.number().positive().default(DEFAULT_CHUNK_SIZE),
+  chunkOverlap: z.number().nonnegative().default(DEFAULT_CHUNK_OVERLAP),
+  separator: z.string().default(" "),
+  backupSeparators: z.array(z.string()).default(["\n"]),
+});
+
+export class TokenTextSplitter extends MetadataAwareTextSplitter {
+  chunkSize: number = DEFAULT_CHUNK_SIZE;
+  chunkOverlap: number = DEFAULT_CHUNK_OVERLAP;
+  separator: string = " ";
+  backupSeparators: string[] = ["\n"];
+  #tokenizer: Tokenizer;
+  #splitFns: Array<(text: string) => string[]> = [];
+
+  constructor(
+    params?: SplitterParams & Partial<z.infer<typeof tokenTextSplitterSchema>>,
+  ) {
+    super();
+
+    if (params) {
+      const parsedParams = tokenTextSplitterSchema.parse(params);
+      this.chunkSize = parsedParams.chunkSize;
+      this.chunkOverlap = parsedParams.chunkOverlap;
+      this.separator = parsedParams.separator;
+      this.backupSeparators = parsedParams.backupSeparators;
+    }
+
+    if (this.chunkOverlap > this.chunkSize) {
+      throw new Error(
+        `Got a larger chunk overlap (${this.chunkOverlap}) than chunk size (${this.chunkSize}), should be smaller.`,
+      );
+    }
+
+    this.#tokenizer = params?.tokenizer ?? Settings.tokenizer;
+
+    const allSeparators = [this.separator, ...this.backupSeparators];
+    this.#splitFns = allSeparators.map((sep) => splitBySep(sep));
+    this.#splitFns.push(splitByChar());
+  }
+
+  /**
+   * Split text into chunks, reserving space required for metadata string.
+   * @param text The text to split.
+   * @param metadata The metadata string.
+   * @returns An array of text chunks.
+   */
+  splitTextMetadataAware(text: string, metadata: string): string[] {
+    const metadataLength =
+      this.tokenSize(metadata) + DEFAULT_METADATA_FORMAT_LEN;
+    const effectiveChunkSize = this.chunkSize - metadataLength;
+
+    if (effectiveChunkSize <= 0) {
+      throw new Error(
+        `Metadata length (${metadataLength}) is longer than chunk size (${this.chunkSize}). ` +
+          `Consider increasing the chunk size or decreasing the size of your metadata to avoid this.`,
+      );
+    } else if (effectiveChunkSize < 50) {
+      console.warn(
+        `Metadata length (${metadataLength}) is close to chunk size (${this.chunkSize}). ` +
+          `Resulting chunks are less than 50 tokens. Consider increasing the chunk size or decreasing the size of your metadata to avoid this.`,
+      );
+    }
+
+    return this._splitText(text, effectiveChunkSize);
+  }
+
+  /**
+   * Split text into chunks.
+   * @param text The text to split.
+   * @returns An array of text chunks.
+   */
+  splitText(text: string): string[] {
+    return this._splitText(text, this.chunkSize);
+  }
+
+  /**
+   * Internal method to split text into chunks up to a specified size.
+   * @param text The text to split.
+   * @param chunkSize The maximum size of each chunk.
+   * @returns An array of text chunks.
+   */
+  private _splitText(text: string, chunkSize: number): string[] {
+    if (text === "") return [text];
+
+    // Dispatch chunking start event
+    Settings.callbackManager.dispatchEvent("chunking-start", { text: [text] });
+
+    const splits = this._split(text, chunkSize);
+    const chunks = this._merge(splits, chunkSize);
+
+    Settings.callbackManager.dispatchEvent("chunking-end", { chunks });
+
+    return chunks;
+  }
+
+  /**
+   * Break text into splits that are smaller than the chunk size.
+   * @param text The text to split.
+   * @param chunkSize The maximum size of each split.
+   * @returns An array of text splits.
+   */
+  private _split(text: string, chunkSize: number): string[] {
+    if (this.tokenSize(text) <= chunkSize) {
+      return [text];
+    }
+
+    for (const splitFn of this.#splitFns) {
+      const splits = splitFn(text);
+      if (splits.length > 1) {
+        const newSplits: string[] = [];
+        for (const split of splits) {
+          const splitLen = this.tokenSize(split);
+          if (splitLen <= chunkSize) {
+            newSplits.push(split);
+          } else {
+            newSplits.push(...this._split(split, chunkSize));
+          }
+        }
+        return newSplits;
+      }
+    }
+
+    return [text];
+  }
+
+  /**
+   * Merge splits into chunks with overlap.
+   * @param splits The array of text splits.
+   * @param chunkSize The maximum size of each chunk.
+   * @returns An array of merged text chunks.
+   */
+  private _merge(splits: string[], chunkSize: number): string[] {
+    const chunks: string[] = [];
+    let currentChunk: string[] = [];
+    let currentLength = 0;
+
+    for (const split of splits) {
+      const splitLength = this.tokenSize(split);
+
+      if (splitLength > chunkSize) {
+        console.warn(
+          `Got a split of size ${splitLength}, larger than chunk size ${chunkSize}.`,
+        );
+      }
+
+      if (currentLength + splitLength > chunkSize) {
+        const chunk = currentChunk.join("").trim();
+        if (chunk) {
+          chunks.push(chunk);
+        }
+
+        currentChunk = [];
+        currentLength = 0;
+
+        const overlapTokens = this.chunkOverlap;
+        const overlapSplits: string[] = [];
+
+        let overlapLength = 0;
+        while (
+          overlapSplits.length < splits.length &&
+          overlapLength < overlapTokens
+        ) {
+          const overlapSplit = currentChunk.shift();
+          if (!overlapSplit) break;
+          overlapSplits.push(overlapSplit);
+          overlapLength += this.tokenSize(overlapSplit);
+        }
+
+        for (const overlapSplit of overlapSplits.reverse()) {
+          currentChunk.push(overlapSplit);
+          currentLength += this.tokenSize(overlapSplit);
+          if (currentLength >= overlapTokens) break;
+        }
+      }
+
+      currentChunk.push(split);
+      currentLength += splitLength;
+    }
+
+    const finalChunk = currentChunk.join("").trim();
+    if (finalChunk) {
+      chunks.push(finalChunk);
+    }
+
+    return chunks;
+  }
+
+  /**
+   * Calculate the number of tokens in the text using the tokenizer.
+   * @param text The text to tokenize.
+   * @returns The number of tokens.
+   */
+  private tokenSize(text: string): number {
+    return this.#tokenizer.encode(text).length;
+  }
+}

packages/core/src/node-parser/utils.ts

@@ -3,7 +3,10 @@ import SentenceTokenizer from "./sentence_tokenizer";
 
 export type TextSplitterFn = (text: string) => string[];
 
-const truncateText = (text: string, textSplitter: TextSplitter): string => {
+export const truncateText = (
+  text: string,
+  textSplitter: TextSplitter,
+): string => {
   const chunks = textSplitter.splitText(text);
   return chunks[0] ?? text;
 };

packages/core/src/prompts/prompt.ts

@@ -64,11 +64,13 @@ export const defaultRefinePrompt: RefinePrompt = new PromptTemplate({
   templateVars: ["query", "existingAnswer", "context"],
   template: `The original query is as follows: {query}
 We have provided an existing answer: {existingAnswer}
-We have the opportunity to refine the existing answer (only if needed) with some more context below.
+We have the opportunity to refine the existing answer
+(only if needed) with some more context below.
 ------------
 {context}
 ------------
-Given the new context, refine the original answer to better answer the query. If the context isn't useful, return the original answer.
+Given the new context, refine the original answer to better answer the query.
+If the context isn't useful, return the original answer.
 Refined Answer:`,
 });
 

packages/core/src/response-synthesizers/factory.ts

@@ -403,27 +403,27 @@ class MultiModal extends BaseSynthesizer {
   }
 }
 
-export function getResponseSynthesizer(
-  mode: ResponseMode,
+const modeToSynthesizer = {
+  compact: CompactAndRefine,
+  refine: Refine,
+  tree_summarize: TreeSummarize,
+  multi_modal: MultiModal,
+} as const;
+
+export function getResponseSynthesizer<Mode extends ResponseMode>(
+  mode: Mode,
   options: BaseSynthesizerOptions & {
     textQATemplate?: TextQAPrompt;
     refineTemplate?: RefinePrompt;
     summaryTemplate?: TreeSummarizePrompt;
     metadataMode?: MetadataMode;
   } = {},
-) {
-  switch (mode) {
-    case "compact": {
-      return new CompactAndRefine(options);
-    }
-    case "refine": {
-      return new Refine(options);
-    }
-    case "tree_summarize": {
-      return new TreeSummarize(options);
-    }
-    case "multi_modal": {
-      return new MultiModal(options);
-    }
+): InstanceType<(typeof modeToSynthesizer)[Mode]> {
+  const Synthesizer: (typeof modeToSynthesizer)[Mode] = modeToSynthesizer[mode];
+  if (!Synthesizer) {
+    throw new Error(`Invalid response mode: ${mode}`);
   }
+  return new Synthesizer(options) as InstanceType<
+    (typeof modeToSynthesizer)[Mode]
+  >;
 }

packages/core/src/schema/node.ts

@@ -479,7 +479,7 @@ export function buildNodeFromSplits(
     ) {
       const imageDoc = doc as ImageNode;
       const imageNode = new ImageNode({
-        id_: idGenerator(i, imageDoc),
+        id_: imageDoc.id_ ?? idGenerator(i, imageDoc),
         text: textChunk,
         image: imageDoc.image,
         embedding: imageDoc.embedding,
@@ -496,7 +496,7 @@ export function buildNodeFromSplits(
     ) {
       const textDoc = doc as TextNode;
       const node = new TextNode({
-        id_: idGenerator(i, textDoc),
+        id_: textDoc.id_ ?? idGenerator(i, textDoc),
         text: textChunk,
         embedding: textDoc.embedding,
         excludedEmbedMetadataKeys: [...textDoc.excludedEmbedMetadataKeys],

packages/core/src/utils/index.ts

@@ -80,3 +80,4 @@ export {
 } from "./llms";
 
 export { objectEntries } from "./object-entries";
+export { UUIDFromString } from "./uuid";

packages/core/src/utils/uuid.ts

@@ -0,0 +1,22 @@
+import { createSHA256 } from "@llamaindex/env";
+
+export function UUIDFromString(input: string) {
+  const hashFunction = createSHA256();
+  hashFunction.update(input);
+  const base64Hash = hashFunction.digest();
+
+  // Convert base64 to hex
+  const hexHash = Buffer.from(base64Hash, "base64").toString("hex");
+
+  // Format the hash to resemble a UUID (version 5 style)
+  const uuid = [
+    hexHash.substring(0, 8),
+    hexHash.substring(8, 12),
+    "5" + hexHash.substring(12, 15), // Set the version to 5 (name-based)
+    ((parseInt(hexHash.substring(15, 17), 16) & 0x3f) | 0x80).toString(16) +
+      hexHash.substring(17, 19), // Set the variant
+    hexHash.substring(19, 31),
+  ].join("-");
+
+  return uuid;
+}

packages/core/tests/utils/uuid.test.ts

@@ -0,0 +1,37 @@
+import { UUIDFromString } from "@llamaindex/core/utils";
+import { describe, expect, it } from "vitest";
+
+const UUID_REGEX =
+  /^[0-9a-f]{8}-[0-9a-f]{4}-5[0-9a-f]{3}-[89ab][0-9a-f]{3}-[0-9a-f]{12}$/i;
+
+describe("UUIDFromString", () => {
+  it("should convert string to UUID", () => {
+    const string = "document_id_1";
+    const result = UUIDFromString(string);
+    expect(result).toBeDefined();
+    expect(result).toMatch(UUID_REGEX);
+  });
+
+  it("should return the same UUID for the same input string", () => {
+    const string = "document_id_1";
+    const result1 = UUIDFromString(string);
+    const result2 = UUIDFromString(string);
+    expect(result1).toEqual(result2);
+  });
+
+  it("should return the different UUID for different input strings", () => {
+    const string1 = "document_id_1";
+    const string2 = "document_id_2";
+    const result1 = UUIDFromString(string1);
+    const result2 = UUIDFromString(string2);
+    expect(result1).not.toEqual(result2);
+  });
+
+  it("should handle case-sensitive input strings", () => {
+    const string1 = "document_id_1";
+    const string2 = "Document_Id_1";
+    const result1 = UUIDFromString(string1);
+    const result2 = UUIDFromString(string2);
+    expect(result1).not.toEqual(result2);
+  });
+});

packages/experimental/CHANGELOG.md

@@ -1,5 +1,33 @@
 # @llamaindex/experimental
 
+## 0.0.95
+
+### Patch Changes
+
+- Updated dependencies [ee697fb]
+  - llamaindex@0.6.17
+
+## 0.0.94
+
+### Patch Changes
+
+- Updated dependencies [63e9846]
+- Updated dependencies [6f3a31c]
+  - llamaindex@0.6.16
+
+## 0.0.93
+
+### Patch Changes
+
+- Updated dependencies [2a82413]
+  - llamaindex@0.6.15
+
+## 0.0.92
+
+### Patch Changes
+
+- llamaindex@0.6.14
+
 ## 0.0.91
 
 ### Patch Changes

packages/experimental/package.json

@@ -1,7 +1,7 @@
 {
   "name": "@llamaindex/experimental",
   "description": "Experimental package for LlamaIndexTS",
-  "version": "0.0.91",
+  "version": "0.0.95",
   "type": "module",
   "types": "dist/type/index.d.ts",
   "main": "dist/cjs/index.js",

packages/llamaindex/CHANGELOG.md

@@ -1,5 +1,52 @@
 # llamaindex
 
+## 0.6.17
+
+### Patch Changes
+
+- ee697fb: fix: generate uuid when inserting to Qdrant
+- Updated dependencies [ee697fb]
+  - @llamaindex/core@0.2.11
+  - @llamaindex/cloud@0.2.13
+  - @llamaindex/ollama@0.0.6
+  - @llamaindex/openai@0.1.14
+  - @llamaindex/groq@0.0.13
+
+## 0.6.16
+
+### Patch Changes
+
+- 63e9846: fix: preFilters does not work with asQueryEngine
+- 6f3a31c: feat: add metadata filters for Qdrant vector store
+- Updated dependencies [3489e7d]
+- Updated dependencies [468bda5]
+  - @llamaindex/core@0.2.10
+  - @llamaindex/cloud@0.2.12
+  - @llamaindex/ollama@0.0.5
+  - @llamaindex/openai@0.1.13
+  - @llamaindex/groq@0.0.12
+
+## 0.6.15
+
+### Patch Changes
+
+- 2a82413: fix(core): set `Settings.llm` to OpenAI by default and support lazy load openai
+- Updated dependencies [2a82413]
+- Updated dependencies [0b20ff9]
+  - @llamaindex/groq@0.0.11
+  - @llamaindex/openai@0.1.12
+  - @llamaindex/cloud@0.2.11
+
+## 0.6.14
+
+### Patch Changes
+
+- Updated dependencies [b17d439]
+  - @llamaindex/core@0.2.9
+  - @llamaindex/ollama@0.0.4
+  - @llamaindex/openai@0.1.11
+  - @llamaindex/groq@0.0.10
+
 ## 0.6.13
 
 ### Patch Changes

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

@@ -1,5 +1,33 @@
 # @llamaindex/cloudflare-worker-agent-test
 
+## 0.0.79
+
+### Patch Changes
+
+- Updated dependencies [ee697fb]
+  - llamaindex@0.6.17
+
+## 0.0.78
+
+### Patch Changes
+
+- Updated dependencies [63e9846]
+- Updated dependencies [6f3a31c]
+  - llamaindex@0.6.16
+
+## 0.0.77
+
+### Patch Changes
+
+- Updated dependencies [2a82413]
+  - llamaindex@0.6.15
+
+## 0.0.76
+
+### Patch Changes
+
+- llamaindex@0.6.14
+
 ## 0.0.75
 
 ### Patch Changes

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

@@ -1,6 +1,6 @@
 {
   "name": "@llamaindex/cloudflare-worker-agent-test",
-  "version": "0.0.75",
+  "version": "0.0.79",
   "type": "module",
   "private": true,
   "scripts": {

packages/llamaindex/e2e/examples/llama-parse-browser/CHANGELOG.md

@@ -1,5 +1,24 @@
 # @llamaindex/llama-parse-browser-test
 
+## 0.0.9
+
+### Patch Changes
+
+- @llamaindex/cloud@0.2.13
+
+## 0.0.8
+
+### Patch Changes
+
+- @llamaindex/cloud@0.2.12
+
+## 0.0.7
+
+### Patch Changes
+
+- Updated dependencies [0b20ff9]
+  - @llamaindex/cloud@0.2.11
+
 ## 0.0.6
 
 ### Patch Changes

packages/llamaindex/e2e/examples/llama-parse-browser/package.json

@@ -1,7 +1,7 @@
 {
   "name": "@llamaindex/llama-parse-browser-test",
   "private": true,
-  "version": "0.0.6",
+  "version": "0.0.9",
   "type": "module",
   "scripts": {
     "dev": "vite",

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

@@ -1,5 +1,33 @@
 # @llamaindex/next-agent-test
 
+## 0.1.79
+
+### Patch Changes
+
+- Updated dependencies [ee697fb]
+  - llamaindex@0.6.17
+
+## 0.1.78
+
+### Patch Changes
+
+- Updated dependencies [63e9846]
+- Updated dependencies [6f3a31c]
+  - llamaindex@0.6.16
+
+## 0.1.77
+
+### Patch Changes
+
+- Updated dependencies [2a82413]
+  - llamaindex@0.6.15
+
+## 0.1.76
+
+### Patch Changes
+
+- llamaindex@0.6.14
+
 ## 0.1.75
 
 ### Patch Changes

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

@@ -1,6 +1,6 @@
 {
   "name": "@llamaindex/next-agent-test",
-  "version": "0.1.75",
+  "version": "0.1.79",
   "private": true,
   "scripts": {
     "dev": "next dev",

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

@@ -1,5 +1,33 @@
 # test-edge-runtime
 
+## 0.1.78
+
+### Patch Changes
+
+- Updated dependencies [ee697fb]
+  - llamaindex@0.6.17
+
+## 0.1.77
+
+### Patch Changes
+
+- Updated dependencies [63e9846]
+- Updated dependencies [6f3a31c]
+  - llamaindex@0.6.16
+
+## 0.1.76
+
+### Patch Changes
+
+- Updated dependencies [2a82413]
+  - llamaindex@0.6.15
+
+## 0.1.75
+
+### Patch Changes
+
+- llamaindex@0.6.14
+
 ## 0.1.74
 
 ### Patch Changes

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

@@ -1,6 +1,6 @@
 {
   "name": "@llamaindex/nextjs-edge-runtime-test",
-  "version": "0.1.74",
+  "version": "0.1.78",
   "private": true,
   "scripts": {
     "dev": "next dev",

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

@@ -1,5 +1,33 @@
 # @llamaindex/next-node-runtime
 
+## 0.0.60
+
+### Patch Changes
+
+- Updated dependencies [ee697fb]
+  - llamaindex@0.6.17
+
+## 0.0.59
+
+### Patch Changes
+
+- Updated dependencies [63e9846]
+- Updated dependencies [6f3a31c]
+  - llamaindex@0.6.16
+
+## 0.0.58
+
+### Patch Changes
+
+- Updated dependencies [2a82413]
+  - llamaindex@0.6.15
+
+## 0.0.57
+
+### Patch Changes
+
+- llamaindex@0.6.14
+
 ## 0.0.56
 
 ### Patch Changes

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

@@ -1,6 +1,6 @@
 {
   "name": "@llamaindex/next-node-runtime-test",
-  "version": "0.0.56",
+  "version": "0.0.60",
   "private": true,
   "scripts": {
     "dev": "next dev",

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

@@ -1,5 +1,33 @@
 # @llamaindex/waku-query-engine-test
 
+## 0.0.79
+
+### Patch Changes
+
+- Updated dependencies [ee697fb]
+  - llamaindex@0.6.17
+
+## 0.0.78
+
+### Patch Changes
+
+- Updated dependencies [63e9846]
+- Updated dependencies [6f3a31c]
+  - llamaindex@0.6.16
+
+## 0.0.77
+
+### Patch Changes
+
+- Updated dependencies [2a82413]
+  - llamaindex@0.6.15
+
+## 0.0.76
+
+### Patch Changes
+
+- llamaindex@0.6.14
+
 ## 0.0.75
 
 ### Patch Changes

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

@@ -1,6 +1,6 @@
 {
   "name": "@llamaindex/waku-query-engine-test",
-  "version": "0.0.75",
+  "version": "0.0.79",
   "type": "module",
   "private": true,
   "scripts": {

packages/llamaindex/package.json

@@ -1,6 +1,6 @@
 {
   "name": "llamaindex",
-  "version": "0.6.13",
+  "version": "0.6.17",
   "license": "MIT",
   "type": "module",
   "keywords": [

packages/llamaindex/src/Settings.ts

@@ -2,7 +2,6 @@ import {
   type CallbackManager,
   Settings as CoreSettings,
 } from "@llamaindex/core/global";
-import { OpenAI } from "@llamaindex/openai";
 
 import { PromptHelper } from "@llamaindex/core/indices";
 
@@ -61,12 +60,6 @@ class GlobalSettings implements Config {
   }
 
   get llm(): LLM {
-    // fixme: we might need check internal error instead of try-catch here
-    try {
-      CoreSettings.llm;
-    } catch (error) {
-      CoreSettings.llm = new OpenAI();
-    }
     return CoreSettings.llm;
   }
 

packages/llamaindex/src/cloud/LlamaCloudIndex.ts

@@ -308,7 +308,6 @@ export class LlamaCloudIndex {
     return new RetrieverQueryEngine(
       retriever,
       params?.responseSynthesizer,
-      params?.preFilters,
       params?.nodePostprocessors,
     );
   }

packages/llamaindex/src/embeddings/fireworks.ts

@@ -2,7 +2,7 @@ import { getEnv } from "@llamaindex/env";
 import { OpenAIEmbedding } from "@llamaindex/openai";
 
 export class FireworksEmbedding extends OpenAIEmbedding {
-  constructor(init?: Partial<OpenAIEmbedding>) {
+  constructor(init?: Omit<Partial<OpenAIEmbedding>, "session">) {
     const {
       apiKey = getEnv("FIREWORKS_API_KEY"),
       additionalSessionOptions = {},

packages/llamaindex/src/embeddings/together.ts

@@ -2,7 +2,7 @@ import { getEnv } from "@llamaindex/env";
 import { OpenAIEmbedding } from "@llamaindex/openai";
 
 export class TogetherEmbedding extends OpenAIEmbedding {
-  constructor(init?: Partial<OpenAIEmbedding>) {
+  constructor(init?: Omit<Partial<OpenAIEmbedding>, "session">) {
     const {
       apiKey = getEnv("TOGETHER_API_KEY"),
       additionalSessionOptions = {},

packages/llamaindex/src/engines/query/RetrieverQueryEngine.ts

@@ -14,12 +14,10 @@ export class RetrieverQueryEngine extends BaseQueryEngine {
   retriever: BaseRetriever;
   responseSynthesizer: BaseSynthesizer;
   nodePostprocessors: BaseNodePostprocessor[];
-  preFilters?: unknown;
 
   constructor(
     retriever: BaseRetriever,
     responseSynthesizer?: BaseSynthesizer,
-    preFilters?: unknown,
     nodePostprocessors?: BaseNodePostprocessor[],
   ) {
     super(async (strOrQueryBundle, stream) => {
@@ -52,7 +50,6 @@ export class RetrieverQueryEngine extends BaseQueryEngine {
     this.retriever = retriever;
     this.responseSynthesizer =
       responseSynthesizer || getResponseSynthesizer("compact");
-    this.preFilters = preFilters;
     this.nodePostprocessors = nodePostprocessors || [];
   }
 

packages/llamaindex/src/index.edge.ts

@@ -1,3 +1,15 @@
+//#region initial setup for OpenAI
+import { OpenAI } from "@llamaindex/openai";
+import { Settings } from "./Settings.js";
+
+try {
+  Settings.llm;
+} catch {
+  Settings.llm = new OpenAI();
+}
+
+//#endregion
+
 export {
   LlamaParseReader,
   type Language,
@@ -28,12 +40,12 @@ export type {
   JSONArray,
   JSONObject,
   JSONValue,
+  LlamaIndexEventMaps,
   LLMEndEvent,
   LLMStartEvent,
   LLMStreamEvent,
   LLMToolCallEvent,
   LLMToolResultEvent,
-  LlamaIndexEventMaps,
 } from "@llamaindex/core/global";
 export * from "@llamaindex/core/indices";
 export * from "@llamaindex/core/llms";
@@ -61,7 +73,7 @@ export * from "./postprocessors/index.js";
 export * from "./QuestionGenerator.js";
 export * from "./selectors/index.js";
 export * from "./ServiceContext.js";
-export { Settings } from "./Settings.js";
 export * from "./storage/StorageContext.js";
 export * from "./tools/index.js";
 export * from "./types.js";
+export { Settings };

packages/llamaindex/src/indices/keyword/index.ts

@@ -246,7 +246,6 @@ export class KeywordTableIndex extends BaseIndex<KeywordTable> {
     return new RetrieverQueryEngine(
       retriever ?? this.asRetriever(),
       responseSynthesizer,
-      options?.preFilters,
       options?.nodePostprocessors,
     );
   }

packages/llamaindex/src/indices/summary/index.ts

@@ -189,7 +189,6 @@ export class SummaryIndex extends BaseIndex<IndexList> {
     return new RetrieverQueryEngine(
       retriever,
       responseSynthesizer,
-      options?.preFilters,
       options?.nodePostprocessors,
     );
   }

packages/llamaindex/src/indices/vectorStore/index.ts

@@ -298,9 +298,8 @@ export class VectorStoreIndex extends BaseIndex<IndexDict> {
       similarityTopK,
     } = options ?? {};
     return new RetrieverQueryEngine(
-      retriever ?? this.asRetriever({ similarityTopK }),
+      retriever ?? this.asRetriever({ similarityTopK, filters: preFilters }),
       responseSynthesizer,
-      preFilters,
       nodePostprocessors,
     );
   }
@@ -387,7 +386,7 @@ export type VectorIndexRetrieverOptions = {
   index: VectorStoreIndex;
   similarityTopK?: number | undefined;
   topK?: TopKMap | undefined;
-  filters?: MetadataFilters;
+  filters?: MetadataFilters | undefined;
 };
 
 export class VectorIndexRetriever extends BaseRetriever {

packages/llamaindex/src/llm/deepinfra.ts

@@ -6,7 +6,7 @@ const DEFAULT_MODEL = "mistralai/Mixtral-8x22B-Instruct-v0.1";
 const BASE_URL = "https://api.deepinfra.com/v1/openai";
 
 export class DeepInfra extends OpenAI {
-  constructor(init?: Partial<OpenAI>) {
+  constructor(init?: Omit<Partial<OpenAI>, "session">) {
     const {
       apiKey = getEnv(ENV_VARIABLE_NAME),
       additionalSessionOptions = {},

packages/llamaindex/src/llm/deepseek.ts

@@ -10,7 +10,9 @@ type DeepSeekModelName = keyof typeof DEEPSEEK_MODELS;
 const DEFAULT_MODEL: DeepSeekModelName = "deepseek-coder";
 
 export class DeepSeekLLM extends OpenAI {
-  constructor(init?: Partial<OpenAI> & { model?: DeepSeekModelName }) {
+  constructor(
+    init?: Omit<Partial<OpenAI>, "session"> & { model?: DeepSeekModelName },
+  ) {
     const {
       apiKey = getEnv("DEEPSEEK_API_KEY"),
       additionalSessionOptions = {},

packages/llamaindex/src/llm/fireworks.ts

@@ -2,7 +2,7 @@ import { getEnv } from "@llamaindex/env";
 import { OpenAI } from "@llamaindex/openai";
 
 export class FireworksLLM extends OpenAI {
-  constructor(init?: Partial<OpenAI>) {
+  constructor(init?: Omit<Partial<OpenAI>, "session">) {
     const {
       apiKey = getEnv("FIREWORKS_API_KEY"),
       additionalSessionOptions = {},

packages/llamaindex/src/llm/together.ts

@@ -2,7 +2,7 @@ import { getEnv } from "@llamaindex/env";
 import { OpenAI } from "@llamaindex/openai";
 
 export class TogetherLLM extends OpenAI {
-  constructor(init?: Partial<OpenAI>) {
+  constructor(init?: Omit<Partial<OpenAI>, "session">) {
     const {
       apiKey = getEnv("TOGETHER_API_KEY"),
       additionalSessionOptions = {},

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

@@ -1,16 +1,23 @@
 import type { BaseNode } from "@llamaindex/core/schema";
 import {
+  FilterCondition,
+  FilterOperator,
   VectorStoreBase,
   type IEmbedModel,
+  type MetadataFilters,
   type VectorStoreNoEmbedModel,
   type VectorStoreQuery,
   type VectorStoreQueryResult,
 } from "./types.js";
 
-import type { QdrantClientParams } from "@qdrant/js-client-rest";
+import { UUIDFromString } from "@llamaindex/core/utils";
+import type { QdrantClientParams, Schemas } from "@qdrant/js-client-rest";
 import { QdrantClient } from "@qdrant/js-client-rest";
 import { metadataDictToNode, nodeToMetadata } from "./utils.js";
 
+type QdrantFilter = Schemas["Filter"];
+type QdrantMustConditions = QdrantFilter["must"];
+
 type PointStruct = {
   id: string;
   payload: Record<string, string>;
@@ -164,7 +171,7 @@ export class QdrantVectorStore
 
       for (let k = 0; k < nodeIds.length; k++) {
         const point: PointStruct = {
-          id: nodeIds[k]!.id_,
+          id: UUIDFromString(nodeIds[k]!.id_),
           payload: payloads[k]!,
           vector: vectors[k]!,
         };
@@ -272,7 +279,7 @@ export class QdrantVectorStore
   ): Promise<VectorStoreQueryResult> {
     const qdrantFilters = options?.qdrant_filters;
 
-    let queryFilters;
+    let queryFilters: QdrantFilter | undefined;
 
     if (!query.queryEmbedding) {
       throw new Error("No query embedding provided");
@@ -281,7 +288,7 @@ export class QdrantVectorStore
     if (qdrantFilters) {
       queryFilters = qdrantFilters;
     } else {
-      queryFilters = await this.buildQueryFilter(query);
+      queryFilters = buildQueryFilter(query);
     }
 
     const result = (await this.db.search(this.collectionName, {
@@ -292,58 +299,118 @@ export class QdrantVectorStore
 
     return this.parseToQueryResult(result);
   }
+}
 
-  /**
-   * Qdrant filter builder
-   * @param query The VectorStoreQuery to be used
-   */
-  private async buildQueryFilter(query: VectorStoreQuery) {
-    if (!query.docIds && !query.queryStr && !query.filters) {
-      return null;
-    }
+/**
+ * Qdrant filter builder
+ * @param query The VectorStoreQuery to be used
+ */
+function buildQueryFilter(query: VectorStoreQuery): QdrantFilter | undefined {
+  if (!query.docIds && !query.queryStr && !query.filters) return undefined;
 
-    const mustConditions = [];
+  const mustConditions: QdrantMustConditions = [];
+  if (query.docIds) {
+    mustConditions.push({
+      key: "doc_id",
+      match: { any: query.docIds },
+    });
+  }
 
-    if (query.docIds) {
-      mustConditions.push({
-        key: "doc_id",
-        match: {
-          any: query.docIds,
-        },
-      });
-    }
+  const metadataFilters = toQdrantMetadataFilters(query.filters);
+  if (metadataFilters) {
+    mustConditions.push(metadataFilters);
+  }
 
-    if (!query.filters) {
-      return {
-        must: mustConditions,
-      };
-    }
+  return { must: mustConditions };
+}
 
-    const metadataFilters = query.filters.filters;
+/**
+ * Converts metadata filters to Qdrant-compatible filters
+ * @param subFilters The metadata filters to be converted
+ * @returns A QdrantFilter object or undefined if no valid filters are provided
+ */
+function toQdrantMetadataFilters(
+  subFilters?: MetadataFilters,
+): QdrantFilter | undefined {
+  if (!subFilters?.filters.length) return undefined;
 
-    for (let i = 0; i < metadataFilters.length; i++) {
-      const filter = metadataFilters[i]!;
+  const conditions: QdrantMustConditions = [];
 
-      if (typeof filter.key === "number") {
-        mustConditions.push({
-          key: filter.key,
-          match: {
-            gt: filter.value,
-            lt: filter.value,
+  for (const subfilter of subFilters.filters) {
+    if (subfilter.operator === FilterOperator.EQ) {
+      if (typeof subfilter.value === "number") {
+        conditions.push({
+          key: subfilter.key,
+          range: {
+            gte: subfilter.value,
+            lte: subfilter.value,
           },
         });
       } else {
-        mustConditions.push({
-          key: filter.key,
-          match: {
-            value: filter.value,
-          },
+        conditions.push({
+          key: subfilter.key,
+          match: { value: subfilter.value },
         });
       }
+    } else if (subfilter.operator === FilterOperator.LT) {
+      conditions.push({
+        key: subfilter.key,
+        range: { lt: subfilter.value },
+      });
+    } else if (subfilter.operator === FilterOperator.GT) {
+      conditions.push({
+        key: subfilter.key,
+        range: { gt: subfilter.value },
+      });
+    } else if (subfilter.operator === FilterOperator.GTE) {
+      conditions.push({
+        key: subfilter.key,
+        range: { gte: subfilter.value },
+      });
+    } else if (subfilter.operator === FilterOperator.LTE) {
+      conditions.push({
+        key: subfilter.key,
+        range: { lte: subfilter.value },
+      });
+    } else if (subfilter.operator === FilterOperator.TEXT_MATCH) {
+      conditions.push({
+        key: subfilter.key,
+        match: { text: subfilter.value },
+      });
+    } else if (subfilter.operator === FilterOperator.NE) {
+      conditions.push({
+        key: subfilter.key,
+        match: { except: [subfilter.value] },
+      });
+    } else if (subfilter.operator === FilterOperator.IN) {
+      const values = Array.isArray(subfilter.value)
+        ? subfilter.value.map(String)
+        : String(subfilter.value).split(",");
+      conditions.push({
+        key: subfilter.key,
+        match: { any: values },
+      });
+    } else if (subfilter.operator === FilterOperator.NIN) {
+      const values = Array.isArray(subfilter.value)
+        ? subfilter.value.map(String)
+        : String(subfilter.value).split(",");
+      conditions.push({
+        key: subfilter.key,
+        match: { except: values },
+      });
+    } else if (subfilter.operator === FilterOperator.IS_EMPTY) {
+      conditions.push({
+        is_empty: { key: subfilter.key },
+      });
     }
-
-    return {
-      must: mustConditions,
-    };
   }
+
+  const filter: QdrantFilter = {};
+  if (subFilters.condition === FilterCondition.OR) {
+    filter.should = conditions;
+  } else {
+    filter.must = conditions;
+  }
+
+  return filter;
 }

packages/llamaindex/tests/indices/VectorStoreIndex.test.ts

@@ -27,7 +27,7 @@ describe("VectorStoreIndex", () => {
       runs: number = 2,
     ): Promise<Array<number>> => {
       const documents = [new Document({ text: "lorem ipsem", id_: "1" })];
-      const entries = [];
+      const entries: number[] = [];
       for (let i = 0; i < runs; i++) {
         await VectorStoreIndex.fromDocuments(documents, {
           serviceContext,
@@ -43,7 +43,7 @@ describe("VectorStoreIndex", () => {
 
   test("fromDocuments stores duplicates without a doc store strategy", async () => {
     const entries = await testStrategy(DocStoreStrategy.NONE);
-    expect(entries[0]! + 1).toBe(entries[1]);
+    expect(entries[0]).toBe(entries[1]);
   });
 
   test("fromDocuments ignores duplicates with upserts doc store strategy", async () => {

packages/llamaindex/tests/init.test.ts

@@ -0,0 +1,7 @@
+import { expect, test, vi } from "vitest";
+
+test("init without error", async () => {
+  vi.stubEnv("OPENAI_API_KEY", undefined);
+  const { Settings } = await import("llamaindex");
+  expect(Settings.llm).toBeDefined();
+});

packages/llm/groq/CHANGELOG.md

@@ -1,5 +1,31 @@
 # @llamaindex/groq
 
+## 0.0.13
+
+### Patch Changes
+
+- @llamaindex/openai@0.1.14
+
+## 0.0.12
+
+### Patch Changes
+
+- @llamaindex/openai@0.1.13
+
+## 0.0.11
+
+### Patch Changes
+
+- 2a82413: fix(core): set `Settings.llm` to OpenAI by default and support lazy load openai
+- Updated dependencies [2a82413]
+  - @llamaindex/openai@0.1.12
+
+## 0.0.10
+
+### Patch Changes
+
+- @llamaindex/openai@0.1.11
+
 ## 0.0.9
 
 ### Patch Changes

packages/llm/groq/package.json

@@ -1,7 +1,7 @@
 {
   "name": "@llamaindex/groq",
   "description": "Groq Adapter for LlamaIndex",
-  "version": "0.0.9",
+  "version": "0.0.13",
   "type": "module",
   "main": "./dist/index.cjs",
   "module": "./dist/index.js",

packages/llm/groq/src/llm.ts

@@ -4,7 +4,7 @@ import GroqSDK, { type ClientOptions } from "groq-sdk";
 
 export class Groq extends OpenAI {
   constructor(
-    init?: Partial<OpenAI> & {
+    init?: Omit<Partial<OpenAI>, "session"> & {
       additionalSessionOptions?: ClientOptions;
     },
   ) {
@@ -22,9 +22,10 @@ export class Groq extends OpenAI {
       ...rest,
     });
 
-    this.session.openai = new GroqSDK({
-      apiKey,
-      ...init?.additionalSessionOptions,
-    }) as any;
+    this.lazySession = async () =>
+      new GroqSDK({
+        apiKey,
+        ...init?.additionalSessionOptions,
+      }) as any;
   }
 }

packages/llm/ollama/CHANGELOG.md

@@ -1,5 +1,27 @@
 # @llamaindex/ollama
 
+## 0.0.6
+
+### Patch Changes
+
+- Updated dependencies [ee697fb]
+  - @llamaindex/core@0.2.11
+
+## 0.0.5
+
+### Patch Changes
+
+- Updated dependencies [3489e7d]
+- Updated dependencies [468bda5]
+  - @llamaindex/core@0.2.10
+
+## 0.0.4
+
+### Patch Changes
+
+- Updated dependencies [b17d439]
+  - @llamaindex/core@0.2.9
+
 ## 0.0.3
 
 ### Patch Changes

packages/llm/ollama/package.json

@@ -1,7 +1,7 @@
 {
   "name": "@llamaindex/ollama",
   "description": "Ollama Adapter for LlamaIndex",
-  "version": "0.0.3",
+  "version": "0.0.6",
   "type": "module",
   "main": "./dist/index.cjs",
   "module": "./dist/index.js",

packages/llm/openai/CHANGELOG.md

@@ -1,5 +1,33 @@
 # @llamaindex/openai
 
+## 0.1.14
+
+### Patch Changes
+
+- Updated dependencies [ee697fb]
+  - @llamaindex/core@0.2.11
+
+## 0.1.13
+
+### Patch Changes
+
+- Updated dependencies [3489e7d]
+- Updated dependencies [468bda5]
+  - @llamaindex/core@0.2.10
+
+## 0.1.12
+
+### Patch Changes
+
+- 2a82413: fix(core): set `Settings.llm` to OpenAI by default and support lazy load openai
+
+## 0.1.11
+
+### Patch Changes
+
+- Updated dependencies [b17d439]
+  - @llamaindex/core@0.2.9
+
 ## 0.1.10
 
 ### Patch Changes

packages/llm/openai/package.json

@@ -1,7 +1,7 @@
 {
   "name": "@llamaindex/openai",
   "description": "OpenAI Adapter for LlamaIndex",
-  "version": "0.1.10",
+  "version": "0.1.14",
   "type": "module",
   "main": "./dist/index.cjs",
   "module": "./dist/index.js",

packages/llm/openai/src/azure.ts

@@ -2,11 +2,6 @@ import { getEnv } from "@llamaindex/env";
 
 import type { AzureClientOptions } from "openai";
 
-export interface AzureOpenAIConfig extends AzureClientOptions {
-  /** @deprecated use "deployment" instead */
-  deploymentName?: string | undefined;
-}
-
 // NOTE we're not supporting the legacy models as they're not available for new deployments
 // https://learn.microsoft.com/en-us/azure/ai-services/openai/concepts/legacy-models
 // If you have a need for them, please open an issue on GitHub
@@ -85,14 +80,15 @@ const DEFAULT_API_VERSION = "2023-05-15";
 //^ NOTE: this will change over time, if you want to pin it, use a specific version
 
 export function getAzureConfigFromEnv(
-  init?: Partial<AzureOpenAIConfig> & { model?: string },
-): AzureOpenAIConfig {
+  init?: Partial<AzureClientOptions> & { model?: string },
+): AzureClientOptions {
   const deployment =
-    init?.deploymentName ??
-    init?.deployment ??
-    getEnv("AZURE_OPENAI_DEPLOYMENT") ?? // From Azure docs
-    getEnv("AZURE_OPENAI_API_DEPLOYMENT_NAME") ?? // LCJS compatible
-    init?.model; // Fall back to model name, Python compatible
+    init && "deploymentName" in init && typeof init.deploymentName === "string"
+      ? init?.deploymentName
+      : (init?.deployment ??
+        getEnv("AZURE_OPENAI_DEPLOYMENT") ?? // From Azure docs
+        getEnv("AZURE_OPENAI_API_DEPLOYMENT_NAME") ?? // LCJS compatible
+        init?.model); // Fall back to model name, Python compatible
   return {
     apiKey:
       init?.apiKey ??
@@ -110,15 +106,10 @@ export function getAzureConfigFromEnv(
       getEnv("OPENAI_API_VERSION") ?? // Python compatible
       getEnv("AZURE_OPENAI_API_VERSION") ?? // LCJS compatible
       DEFAULT_API_VERSION,
-    deploymentName: deployment, // LCJS compatible
     deployment, // For Azure OpenAI
   };
 }
 
-export function getAzureBaseUrl(config: AzureOpenAIConfig): string {
-  return `${config.endpoint}/openai/deployments/${config.deploymentName}`;
-}
-
 export function getAzureModel(openAIModel: string) {
   for (const [key, value] of Object.entries(
     ALL_AZURE_OPENAI_EMBEDDING_MODELS,

packages/llm/openai/src/embedding.ts

@@ -1,14 +1,16 @@
 import { BaseEmbedding } from "@llamaindex/core/embeddings";
-import { Tokenizers } from "@llamaindex/env";
-import type { ClientOptions as OpenAIClientOptions } from "openai";
-import type { AzureOpenAIConfig } from "./azure.js";
+import { getEnv, Tokenizers } from "@llamaindex/env";
+import type {
+  AzureClientOptions,
+  AzureOpenAI as AzureOpenAILLM,
+  ClientOptions as OpenAIClientOptions,
+  OpenAI as OpenAILLM,
+} from "openai";
 import {
   getAzureConfigFromEnv,
   getAzureModel,
   shouldUseAzure,
 } from "./azure.js";
-import type { OpenAISession } from "./llm.js";
-import { getOpenAISession } from "./llm.js";
 
 export const ALL_OPENAI_EMBEDDING_MODELS = {
   "text-embedding-ada-002": {
@@ -32,6 +34,8 @@ export const ALL_OPENAI_EMBEDDING_MODELS = {
 
 type ModelKeys = keyof typeof ALL_OPENAI_EMBEDDING_MODELS;
 
+type LLMInstance = Pick<AzureOpenAILLM | OpenAILLM, "embeddings" | "apiKey">;
+
 export class OpenAIEmbedding extends BaseEmbedding {
   /** embeddding model. defaults to "text-embedding-ada-002" */
   model: string;
@@ -51,14 +55,26 @@ export class OpenAIEmbedding extends BaseEmbedding {
     | Omit<Partial<OpenAIClientOptions>, "apiKey" | "maxRetries" | "timeout">
     | undefined;
 
-  /** session object */
-  session: OpenAISession;
+  // use lazy here to avoid check OPENAI_API_KEY immediately
+  lazySession: () => Promise<LLMInstance>;
+  #session: Promise<LLMInstance> | null = null;
+  get session() {
+    if (!this.#session) {
+      this.#session = this.lazySession();
+    }
+    return this.#session;
+  }
 
   /**
    * OpenAI Embedding
    * @param init - initial parameters
    */
-  constructor(init?: Partial<OpenAIEmbedding> & { azure?: AzureOpenAIConfig }) {
+  constructor(
+    init?: Omit<Partial<OpenAIEmbedding>, "lazySession"> & {
+      session?: LLMInstance | undefined;
+      azure?: AzureClientOptions;
+    },
+  ) {
     super();
 
     this.model = init?.model ?? "text-embedding-ada-002";
@@ -77,7 +93,6 @@ export class OpenAIEmbedding extends BaseEmbedding {
     if (key) {
       this.embedInfo = ALL_OPENAI_EMBEDDING_MODELS[key];
     }
-
     if (init?.azure || shouldUseAzure()) {
       const azureConfig = {
         ...getAzureConfigFromEnv({
@@ -85,26 +100,32 @@ export class OpenAIEmbedding extends BaseEmbedding {
         }),
         ...init?.azure,
       };
-
-      this.apiKey = azureConfig.apiKey;
-      this.session =
-        init?.session ??
-        getOpenAISession({
-          azure: true,
-          maxRetries: this.maxRetries,
-          timeout: this.timeout,
-          ...this.additionalSessionOptions,
-          ...azureConfig,
-        });
+      this.apiKey =
+        init?.session?.apiKey ?? azureConfig.apiKey ?? getEnv("OPENAI_API_KEY");
+      this.lazySession = async () =>
+        import("openai").then(
+          async ({ AzureOpenAI }) =>
+            init?.session ??
+            new AzureOpenAI({
+              maxRetries: this.maxRetries,
+              timeout: this.timeout!,
+              ...this.additionalSessionOptions,
+              ...azureConfig,
+            }),
+        );
     } else {
-      this.apiKey = init?.apiKey ?? undefined;
-      this.session =
-        init?.session ??
-        getOpenAISession({
-          apiKey: this.apiKey,
-          maxRetries: this.maxRetries,
-          timeout: this.timeout,
-          ...this.additionalSessionOptions,
+      this.apiKey = init?.session?.apiKey ?? getEnv("OPENAI_API_KEY");
+      this.lazySession = async () =>
+        import("openai").then(({ OpenAI }) => {
+          return (
+            init?.session ??
+            new OpenAI({
+              apiKey: this.apiKey,
+              maxRetries: this.maxRetries,
+              timeout: this.timeout!,
+              ...this.additionalSessionOptions,
+            })
+          );
         });
     }
   }
@@ -118,7 +139,9 @@ export class OpenAIEmbedding extends BaseEmbedding {
     // TODO: ensure this for every sub class by calling it in the base class
     input = this.truncateMaxTokens(input);
 
-    const { data } = await this.session.openai.embeddings.create(
+    const { data } = await (
+      await this.session
+    ).embeddings.create(
       this.dimensions
         ? {
             model: this.model,

packages/llm/openai/src/index.ts

@@ -10,9 +10,6 @@ export {
   GPT4_MODELS,
   O1_MODELS,
   OpenAI,
-  OpenAISession,
   type OpenAIAdditionalChatOptions,
   type OpenAIAdditionalMetadata,
 } from "./llm";
-
-export { type AzureOpenAIConfig } from "./azure";

packages/llm/openai/src/llm.ts

@@ -1,12 +1,11 @@
 import { getEnv } from "@llamaindex/env";
-import type OpenAILLM from "openai";
 import type {
-  ClientOptions,
+  AzureClientOptions,
+  AzureOpenAI as AzureOpenAILLM,
   ClientOptions as OpenAIClientOptions,
+  OpenAI as OpenAILLM,
 } from "openai";
-import { AzureOpenAI, OpenAI as OrigOpenAI } from "openai";
 import type { ChatModel } from "openai/resources/chat/chat";
-import { isDeepEqual } from "remeda";
 
 import { wrapEventCaller, wrapLLMEvent } from "@llamaindex/core/decorator";
 import {
@@ -35,64 +34,12 @@ import type {
   ChatCompletionUserMessageParam,
 } from "openai/resources/chat/completions";
 import type { ChatCompletionMessageParam } from "openai/resources/index.js";
-import type { AzureOpenAIConfig } from "./azure.js";
 import {
   getAzureConfigFromEnv,
   getAzureModel,
   shouldUseAzure,
 } from "./azure.js";
 
-export class OpenAISession {
-  openai: Pick<OrigOpenAI, "chat" | "embeddings">;
-
-  constructor(options: ClientOptions & { azure?: boolean } = {}) {
-    if (options.azure) {
-      this.openai = new AzureOpenAI(options as AzureOpenAIConfig);
-    } else {
-      if (!options.apiKey) {
-        options.apiKey = getEnv("OPENAI_API_KEY");
-      }
-
-      if (!options.apiKey) {
-        throw new Error("Set OpenAI Key in OPENAI_API_KEY env variable"); // Overriding OpenAI package's error message
-      }
-
-      this.openai = new OrigOpenAI({
-        ...options,
-      });
-    }
-  }
-}
-
-// I'm not 100% sure this is necessary vs. just starting a new session
-// every time we make a call. They say they try to reuse connections
-// so in theory this is more efficient, but we should test it in the future.
-const defaultOpenAISession: {
-  session: OpenAISession;
-  options: ClientOptions;
-}[] = [];
-
-/**
- * Get a session for the OpenAI API. If one already exists with the same options,
- * it will be returned. Otherwise, a new session will be created.
- * @param options
- * @returns
- */
-export function getOpenAISession(
-  options: ClientOptions & { azure?: boolean } = {},
-) {
-  let session = defaultOpenAISession.find((session) => {
-    return isDeepEqual(session.options, options);
-  })?.session;
-
-  if (!session) {
-    session = new OpenAISession(options);
-    defaultOpenAISession.push({ session, options });
-  }
-
-  return session;
-}
-
 export const GPT4_MODELS = {
   "chatgpt-4o-latest": {
     contextWindow: 128000,
@@ -182,6 +129,8 @@ export type OpenAIAdditionalChatOptions = Omit<
   | "toolChoice"
 >;
 
+type LLMInstance = Pick<AzureOpenAILLM | OpenAILLM, "chat" | "apiKey">;
+
 export class OpenAI extends ToolCallLLM<OpenAIAdditionalChatOptions> {
   model:
     | ChatModel
@@ -196,14 +145,24 @@ export class OpenAI extends ToolCallLLM<OpenAIAdditionalChatOptions> {
   apiKey?: string | undefined = undefined;
   maxRetries: number;
   timeout?: number;
-  session: OpenAISession;
   additionalSessionOptions?:
     | undefined
     | Omit<Partial<OpenAIClientOptions>, "apiKey" | "maxRetries" | "timeout">;
 
+  // use lazy here to avoid check OPENAI_API_KEY immediately
+  lazySession: () => Promise<LLMInstance>;
+  #session: Promise<LLMInstance> | null = null;
+  get session() {
+    if (!this.#session) {
+      this.#session = this.lazySession();
+    }
+    return this.#session;
+  }
+
   constructor(
-    init?: Partial<OpenAI> & {
-      azure?: AzureOpenAIConfig;
+    init?: Omit<Partial<OpenAI>, "session"> & {
+      session?: LLMInstance | undefined;
+      azure?: AzureClientOptions;
     },
   ) {
     super();
@@ -216,6 +175,8 @@ export class OpenAI extends ToolCallLLM<OpenAIAdditionalChatOptions> {
     this.timeout = init?.timeout ?? 60 * 1000; // Default is 60 seconds
     this.additionalChatOptions = init?.additionalChatOptions;
     this.additionalSessionOptions = init?.additionalSessionOptions;
+    this.apiKey =
+      init?.session?.apiKey ?? init?.apiKey ?? getEnv("OPENAI_API_KEY");
 
     if (init?.azure || shouldUseAzure()) {
       const azureConfig = {
@@ -225,25 +186,26 @@ export class OpenAI extends ToolCallLLM<OpenAIAdditionalChatOptions> {
         ...init?.azure,
       };
 
-      this.apiKey = azureConfig.apiKey;
-      this.session =
+      this.lazySession = async () =>
         init?.session ??
-        getOpenAISession({
-          azure: true,
-          maxRetries: this.maxRetries,
-          timeout: this.timeout,
-          ...this.additionalSessionOptions,
-          ...azureConfig,
+        import("openai").then(({ AzureOpenAI }) => {
+          return new AzureOpenAI({
+            maxRetries: this.maxRetries,
+            timeout: this.timeout!,
+            ...this.additionalSessionOptions,
+            ...azureConfig,
+          });
         });
     } else {
-      this.apiKey = init?.apiKey ?? undefined;
-      this.session =
+      this.lazySession = async () =>
         init?.session ??
-        getOpenAISession({
-          apiKey: this.apiKey,
-          maxRetries: this.maxRetries,
-          timeout: this.timeout,
-          ...this.additionalSessionOptions,
+        import("openai").then(({ OpenAI }) => {
+          return new OpenAI({
+            apiKey: this.apiKey,
+            maxRetries: this.maxRetries,
+            timeout: this.timeout!,
+            ...this.additionalSessionOptions,
+          });
         });
     }
   }
@@ -382,7 +344,9 @@ export class OpenAI extends ToolCallLLM<OpenAIAdditionalChatOptions> {
     }
 
     // Non-streaming
-    const response = await this.session.openai.chat.completions.create({
+    const response = await (
+      await this.session
+    ).chat.completions.create({
       ...baseRequestParams,
       stream: false,
     });
@@ -414,11 +378,12 @@ export class OpenAI extends ToolCallLLM<OpenAIAdditionalChatOptions> {
   protected async *streamChat(
     baseRequestParams: OpenAILLM.Chat.ChatCompletionCreateParams,
   ): AsyncIterable<ChatResponseChunk<ToolCallLLMMessageOptions>> {
-    const stream: AsyncIterable<OpenAILLM.Chat.ChatCompletionChunk> =
-      await this.session.openai.chat.completions.create({
-        ...baseRequestParams,
-        stream: true,
-      });
+    const stream: AsyncIterable<OpenAILLM.Chat.ChatCompletionChunk> = await (
+      await this.session
+    ).chat.completions.create({
+      ...baseRequestParams,
+      stream: true,
+    });
 
     // TODO: add callback to streamConverter and use streamConverter here
     // this will be used to keep track of the current tool call, make sure input are valid json object.

pnpm-lock.yaml

@@ -360,10 +360,10 @@ importers:
         specifier: ^0.53.0
         version: 0.53.0(typescript@5.6.2)
       '@llamaindex/core':
-        specifier: workspace:^0.2.8
+        specifier: workspace:*
         version: link:../core
       '@llamaindex/env':
-        specifier: workspace:^0.1.13
+        specifier: workspace:*
         version: link:../env
       bunchee:
         specifier: 5.3.2
@@ -19518,8 +19518,8 @@ snapshots:
       '@typescript-eslint/parser': 7.2.0(eslint@8.57.0)(typescript@5.6.2)
       eslint: 8.57.0
       eslint-import-resolver-node: 0.3.9
-      eslint-import-resolver-typescript: 3.6.3(@typescript-eslint/parser@7.2.0(eslint@8.57.0)(typescript@5.6.2))(eslint-import-resolver-node@0.3.9)(eslint-plugin-import@2.29.1(@typescript-eslint/parser@7.2.0(eslint@8.57.0)(typescript@5.6.2))(eslint@8.57.0))(eslint@8.57.0)
-      eslint-plugin-import: 2.29.1(@typescript-eslint/parser@7.2.0(eslint@8.57.0)(typescript@5.6.2))(eslint-import-resolver-typescript@3.6.3(@typescript-eslint/parser@7.2.0(eslint@8.57.0)(typescript@5.6.2))(eslint-import-resolver-node@0.3.9)(eslint-plugin-import@2.29.1(@typescript-eslint/parser@7.2.0(eslint@8.57.0)(typescript@5.6.2))(eslint@8.57.0))(eslint@8.57.0))(eslint@8.57.0)
+      eslint-import-resolver-typescript: 3.6.3(@typescript-eslint/parser@7.2.0(eslint@8.57.0)(typescript@5.6.2))(eslint-import-resolver-node@0.3.9)(eslint-plugin-import@2.29.1)(eslint@8.57.0)
+      eslint-plugin-import: 2.29.1(@typescript-eslint/parser@7.2.0(eslint@8.57.0)(typescript@5.6.2))(eslint-import-resolver-typescript@3.6.3)(eslint@8.57.0)
       eslint-plugin-jsx-a11y: 6.9.0(eslint@8.57.0)
       eslint-plugin-react: 7.35.0(eslint@8.57.0)
       eslint-plugin-react-hooks: 4.6.2(eslint@8.57.0)
@@ -19566,25 +19566,6 @@ snapshots:
     transitivePeerDependencies:
       - supports-color
 
-  eslint-import-resolver-typescript@3.6.3(@typescript-eslint/parser@7.2.0(eslint@8.57.0)(typescript@5.6.2))(eslint-import-resolver-node@0.3.9)(eslint-plugin-import@2.29.1(@typescript-eslint/parser@7.2.0(eslint@8.57.0)(typescript@5.6.2))(eslint@8.57.0))(eslint@8.57.0):
-    dependencies:
-      '@nolyfill/is-core-module': 1.0.39
-      debug: 4.3.7
-      enhanced-resolve: 5.17.1
-      eslint: 8.57.0
-      eslint-module-utils: 2.8.2(@typescript-eslint/parser@7.2.0(eslint@8.57.0)(typescript@5.6.2))(eslint-import-resolver-node@0.3.9)(eslint-import-resolver-typescript@3.6.3(@typescript-eslint/parser@7.2.0(eslint@8.57.0)(typescript@5.6.2))(eslint-import-resolver-node@0.3.9)(eslint-plugin-import@2.29.1(@typescript-eslint/parser@7.2.0(eslint@8.57.0)(typescript@5.6.2))(eslint@8.57.0))(eslint@8.57.0))(eslint@8.57.0)
-      fast-glob: 3.3.2
-      get-tsconfig: 4.8.0
-      is-bun-module: 1.1.0
-      is-glob: 4.0.3
-    optionalDependencies:
-      eslint-plugin-import: 2.29.1(@typescript-eslint/parser@7.2.0(eslint@8.57.0)(typescript@5.6.2))(eslint-import-resolver-typescript@3.6.3(@typescript-eslint/parser@7.2.0(eslint@8.57.0)(typescript@5.6.2))(eslint-import-resolver-node@0.3.9)(eslint-plugin-import@2.29.1(@typescript-eslint/parser@7.2.0(eslint@8.57.0)(typescript@5.6.2))(eslint@8.57.0))(eslint@8.57.0))(eslint@8.57.0)
-    transitivePeerDependencies:
-      - '@typescript-eslint/parser'
-      - eslint-import-resolver-node
-      - eslint-import-resolver-webpack
-      - supports-color
-
   eslint-import-resolver-typescript@3.6.3(@typescript-eslint/parser@7.2.0(eslint@8.57.0)(typescript@5.6.2))(eslint-import-resolver-node@0.3.9)(eslint-plugin-import@2.29.1)(eslint@8.57.0):
     dependencies:
       '@nolyfill/is-core-module': 1.0.39
@@ -19604,17 +19585,6 @@ snapshots:
       - eslint-import-resolver-webpack
       - supports-color
 
-  eslint-module-utils@2.8.2(@typescript-eslint/parser@7.2.0(eslint@8.57.0)(typescript@5.6.2))(eslint-import-resolver-node@0.3.9)(eslint-import-resolver-typescript@3.6.3(@typescript-eslint/parser@7.2.0(eslint@8.57.0)(typescript@5.6.2))(eslint-import-resolver-node@0.3.9)(eslint-plugin-import@2.29.1(@typescript-eslint/parser@7.2.0(eslint@8.57.0)(typescript@5.6.2))(eslint@8.57.0))(eslint@8.57.0))(eslint@8.57.0):
-    dependencies:
-      debug: 3.2.7
-    optionalDependencies:
-      '@typescript-eslint/parser': 7.2.0(eslint@8.57.0)(typescript@5.6.2)
-      eslint: 8.57.0
-      eslint-import-resolver-node: 0.3.9
-      eslint-import-resolver-typescript: 3.6.3(@typescript-eslint/parser@7.2.0(eslint@8.57.0)(typescript@5.6.2))(eslint-import-resolver-node@0.3.9)(eslint-plugin-import@2.29.1(@typescript-eslint/parser@7.2.0(eslint@8.57.0)(typescript@5.6.2))(eslint@8.57.0))(eslint@8.57.0)
-    transitivePeerDependencies:
-      - supports-color
-
   eslint-module-utils@2.8.2(@typescript-eslint/parser@7.2.0(eslint@8.57.0)(typescript@5.6.2))(eslint-import-resolver-node@0.3.9)(eslint-import-resolver-typescript@3.6.3(@typescript-eslint/parser@7.2.0(eslint@8.57.0)(typescript@5.6.2))(eslint-import-resolver-node@0.3.9)(eslint-plugin-import@2.29.1)(eslint@8.57.0))(eslint@8.57.0):
     dependencies:
       debug: 3.2.7
@@ -19626,33 +19596,6 @@ snapshots:
     transitivePeerDependencies:
       - supports-color
 
-  eslint-plugin-import@2.29.1(@typescript-eslint/parser@7.2.0(eslint@8.57.0)(typescript@5.6.2))(eslint-import-resolver-typescript@3.6.3(@typescript-eslint/parser@7.2.0(eslint@8.57.0)(typescript@5.6.2))(eslint-import-resolver-node@0.3.9)(eslint-plugin-import@2.29.1(@typescript-eslint/parser@7.2.0(eslint@8.57.0)(typescript@5.6.2))(eslint@8.57.0))(eslint@8.57.0))(eslint@8.57.0):
-    dependencies:
-      array-includes: 3.1.8
-      array.prototype.findlastindex: 1.2.5
-      array.prototype.flat: 1.3.2
-      array.prototype.flatmap: 1.3.2
-      debug: 3.2.7
-      doctrine: 2.1.0
-      eslint: 8.57.0
-      eslint-import-resolver-node: 0.3.9
-      eslint-module-utils: 2.8.2(@typescript-eslint/parser@7.2.0(eslint@8.57.0)(typescript@5.6.2))(eslint-import-resolver-node@0.3.9)(eslint-import-resolver-typescript@3.6.3(@typescript-eslint/parser@7.2.0(eslint@8.57.0)(typescript@5.6.2))(eslint-import-resolver-node@0.3.9)(eslint-plugin-import@2.29.1(@typescript-eslint/parser@7.2.0(eslint@8.57.0)(typescript@5.6.2))(eslint@8.57.0))(eslint@8.57.0))(eslint@8.57.0)
-      hasown: 2.0.2
-      is-core-module: 2.15.1
-      is-glob: 4.0.3
-      minimatch: 3.1.2
-      object.fromentries: 2.0.8
-      object.groupby: 1.0.3
-      object.values: 1.2.0
-      semver: 6.3.1
-      tsconfig-paths: 3.15.0
-    optionalDependencies:
-      '@typescript-eslint/parser': 7.2.0(eslint@8.57.0)(typescript@5.6.2)
-    transitivePeerDependencies:
-      - eslint-import-resolver-typescript
-      - eslint-import-resolver-webpack
-      - supports-color
-
   eslint-plugin-import@2.29.1(@typescript-eslint/parser@7.2.0(eslint@8.57.0)(typescript@5.6.2))(eslint-import-resolver-typescript@3.6.3)(eslint@8.57.0):
     dependencies:
       array-includes: 3.1.8
@@ -19679,7 +19622,6 @@ snapshots:
       - eslint-import-resolver-typescript
       - eslint-import-resolver-webpack
       - supports-color
-    optional: true
 
   eslint-plugin-import@2.29.1(@typescript-eslint/parser@8.5.0(eslint@8.57.0)(typescript@5.6.2))(eslint@8.57.0):
     dependencies: