Release 0.7.0 (#1337)

Co-authored-by: github-actions[bot] <github-actions[bot]@users.noreply.github.com>
Co-authored-by: himself65 <himself65@users.noreply.github.com>

apps/docs/CHANGELOG.md

@@ -1,5 +1,83 @@
 # docs
 
+## 0.0.92
+
+### Patch Changes
+
+- Updated dependencies [1364e8e]
+- Updated dependencies [3b7736f]
+- Updated dependencies [96fc69c]
+  - llamaindex@0.7.0
+  - @llamaindex/examples@0.0.9
+
+## 0.0.91
+
+### Patch Changes
+
+- Updated dependencies [5729bd9]
+  - llamaindex@0.6.22
+
+## 0.0.90
+
+### Patch Changes
+
+- Updated dependencies [6f75306]
+- Updated dependencies [94cb4ad]
+  - llamaindex@0.6.21
+
+## 0.0.89
+
+### Patch Changes
+
+- Updated dependencies [6a9a7b1]
+  - llamaindex@0.6.20
+
+## 0.0.88
+
+### Patch Changes
+
+- Updated dependencies [62cba52]
+- Updated dependencies [d265e96]
+- Updated dependencies [d30bbf7]
+- Updated dependencies [53fd00a]
+  - llamaindex@0.6.19
+
+## 0.0.87
+
+### Patch Changes
+
+- Updated dependencies [5f67820]
+- Updated dependencies [fe08d04]
+  - llamaindex@0.6.18
+
+## 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": "0.0.92",
   "private": true,
   "scripts": {
     "docusaurus": "docusaurus",

examples/CHANGELOG.md

@@ -1,5 +1,16 @@
 # examples
 
+## 0.0.9
+
+### Patch Changes
+
+- Updated dependencies [1364e8e]
+- Updated dependencies [96fc69c]
+- Updated dependencies [3b7736f]
+- Updated dependencies [96fc69c]
+  - llamaindex@0.7.0
+  - @llamaindex/core@0.3.0
+
 ## 0.0.8
 
 ### Patch Changes

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/astradb/example.ts

@@ -1,6 +1,7 @@
 import {
   AstraDBVectorStore,
   Document,
+  MetadataFilters,
   storageContextFromDefaults,
   VectorStoreIndex,
 } from "llamaindex";
@@ -42,8 +43,10 @@ async function main() {
     const index = await VectorStoreIndex.fromDocuments(docs, {
       storageContext: ctx,
     });
-
-    const queryEngine = index.asQueryEngine();
+    const preFilters: MetadataFilters = {
+      filters: [{ key: "id", operator: "in", value: [123, 789] }],
+    }; // try changing the filters to see the different results
+    const queryEngine = index.asQueryEngine({ preFilters });
     const response = await queryEngine.query({
       query: "Describe AstraDB.",
     });

examples/chromadb/preFilters.ts

@@ -1,57 +1,83 @@
 import {
   ChromaVectorStore,
   Document,
+  MetadataFilters,
   VectorStoreIndex,
   storageContextFromDefaults,
 } from "llamaindex";
 
-const collectionName = "dog_colors";
+const collectionName = "dogs_with_color";
 
 async function main() {
   try {
-    const docs = [
-      new Document({
-        text: "The dog is brown",
-        metadata: {
-          dogId: "1",
-        },
-      }),
-      new Document({
-        text: "The dog is red",
-        metadata: {
-          dogId: "2",
-        },
-      }),
-    ];
-
-    console.log("Creating ChromaDB vector store");
     const chromaVS = new ChromaVectorStore({ collectionName });
-    const ctx = await storageContextFromDefaults({ vectorStore: chromaVS });
+    const index = await VectorStoreIndex.fromVectorStore(chromaVS);
 
-    console.log("Embedding documents and adding to index");
-    const index = await VectorStoreIndex.fromDocuments(docs, {
-      storageContext: ctx,
-    });
+    const queryFn = async (filters?: MetadataFilters) => {
+      console.log("\nQuerying dogs by filters: ", JSON.stringify(filters));
+      const query = "List all colors of dogs";
+      const queryEngine = index.asQueryEngine({
+        preFilters: filters,
+        similarityTopK: 3,
+      });
+      const response = await queryEngine.query({ query });
+      console.log(response.toString());
+    };
 
-    console.log("Querying index");
-    const queryEngine = index.asQueryEngine({
-      preFilters: {
-        filters: [
-          {
-            key: "dogId",
-            value: "2",
-            operator: "==",
-          },
-        ],
-      },
-    });
-    const response = await queryEngine.query({
-      query: "What is the color of the dog?",
-    });
-    console.log(response.toString());
+    await queryFn(); // red, brown, yellow
+    await queryFn({ filters: [{ key: "dogId", value: "1", operator: "==" }] }); // brown
+    await queryFn({ filters: [{ key: "dogId", value: "1", operator: "!=" }] }); // red, yellow
+    await queryFn({
+      filters: [
+        { key: "dogId", value: "1", operator: "==" },
+        { key: "dogId", value: "3", operator: "==" },
+      ],
+      condition: "or",
+    }); // brown, yellow
+    await queryFn({
+      filters: [{ key: "dogId", value: ["1", "2"], operator: "in" }],
+    }); // red, brown
   } catch (e) {
     console.error(e);
   }
 }
 
-void main();
+async function generate() {
+  const docs = [
+    new Document({
+      id_: "doc1",
+      text: "The dog is brown",
+      metadata: {
+        dogId: "1",
+      },
+    }),
+    new Document({
+      id_: "doc2",
+      text: "The dog is red",
+      metadata: {
+        dogId: "2",
+      },
+    }),
+    new Document({
+      id_: "doc3",
+      text: "The dog is yellow",
+      metadata: {
+        dogId: "3",
+      },
+    }),
+  ];
+
+  console.log("Creating ChromaDB vector store");
+  const chromaVS = new ChromaVectorStore({ collectionName });
+  const ctx = await storageContextFromDefaults({ vectorStore: chromaVS });
+
+  console.log("Embedding documents and adding to index");
+  await VectorStoreIndex.fromDocuments(docs, {
+    storageContext: ctx,
+  });
+}
+
+(async () => {
+  await generate();
+  await main();
+})();

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/package.json

@@ -1,12 +1,12 @@
 {
   "name": "@llamaindex/examples",
   "private": true,
-  "version": "0.0.8",
+  "version": "0.0.9",
   "dependencies": {
     "@aws-crypto/sha256-js": "^5.2.0",
     "@azure/identity": "^4.4.1",
     "@datastax/astra-db-ts": "^1.4.1",
-    "@llamaindex/core": "^0.2.0",
+    "@llamaindex/core": "^0.3.0",
     "@notionhq/client": "^2.2.15",
     "@pinecone-database/pinecone": "^3.0.2",
     "@vercel/postgres": "^0.10.0",
@@ -15,7 +15,7 @@
     "commander": "^12.1.0",
     "dotenv": "^16.4.5",
     "js-tiktoken": "^1.0.14",
-    "llamaindex": "^0.6.0",
+    "llamaindex": "^0.7.0",
     "mongodb": "^6.7.0",
     "pathe": "^1.1.2",
     "postgres": "^3.4.4"

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

@@ -1,4 +1,5 @@
 import {
+  BaseVectorStore,
   getResponseSynthesizer,
   OpenAI,
   OpenAIEmbedding,
@@ -6,7 +7,6 @@ import {
   Settings,
   TextNode,
   VectorIndexRetriever,
-  VectorStore,
   VectorStoreIndex,
   VectorStoreQuery,
   VectorStoreQueryResult,
@@ -24,7 +24,7 @@ Settings.llm = new OpenAI({
  * Please do not use this class in production; it's only for demonstration purposes.
  */
 class PineconeVectorStore<T extends RecordMetadata = RecordMetadata>
-  implements VectorStore
+  implements BaseVectorStore
 {
   storesText = true;
   isEmbeddingQuery = false;
@@ -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 +1,82 @@
 # @llamaindex/autotool
 
+## 4.0.0
+
+### Patch Changes
+
+- Updated dependencies [1364e8e]
+- Updated dependencies [3b7736f]
+- Updated dependencies [96fc69c]
+  - llamaindex@0.7.0
+
+## 3.0.22
+
+### Patch Changes
+
+- Updated dependencies [5729bd9]
+  - llamaindex@0.6.22
+
+## 3.0.21
+
+### Patch Changes
+
+- Updated dependencies [6f75306]
+- Updated dependencies [94cb4ad]
+  - llamaindex@0.6.21
+
+## 3.0.20
+
+### Patch Changes
+
+- Updated dependencies [6a9a7b1]
+  - llamaindex@0.6.20
+
+## 3.0.19
+
+### Patch Changes
+
+- Updated dependencies [62cba52]
+- Updated dependencies [d265e96]
+- Updated dependencies [d30bbf7]
+- Updated dependencies [53fd00a]
+  - llamaindex@0.6.19
+
+## 3.0.18
+
+### Patch Changes
+
+- Updated dependencies [5f67820]
+- Updated dependencies [fe08d04]
+  - llamaindex@0.6.18
+
+## 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 +1,92 @@
 # @llamaindex/autotool-01-node-example
 
+## 0.0.32
+
+### Patch Changes
+
+- Updated dependencies [1364e8e]
+- Updated dependencies [3b7736f]
+- Updated dependencies [96fc69c]
+  - llamaindex@0.7.0
+  - @llamaindex/autotool@4.0.0
+
+## 0.0.31
+
+### Patch Changes
+
+- Updated dependencies [5729bd9]
+  - llamaindex@0.6.22
+  - @llamaindex/autotool@3.0.22
+
+## 0.0.30
+
+### Patch Changes
+
+- Updated dependencies [6f75306]
+- Updated dependencies [94cb4ad]
+  - llamaindex@0.6.21
+  - @llamaindex/autotool@3.0.21
+
+## 0.0.29
+
+### Patch Changes
+
+- Updated dependencies [6a9a7b1]
+  - llamaindex@0.6.20
+  - @llamaindex/autotool@3.0.20
+
+## 0.0.28
+
+### Patch Changes
+
+- Updated dependencies [62cba52]
+- Updated dependencies [d265e96]
+- Updated dependencies [d30bbf7]
+- Updated dependencies [53fd00a]
+  - llamaindex@0.6.19
+  - @llamaindex/autotool@3.0.19
+
+## 0.0.27
+
+### Patch Changes
+
+- Updated dependencies [5f67820]
+- Updated dependencies [fe08d04]
+  - llamaindex@0.6.18
+  - @llamaindex/autotool@3.0.18
+
+## 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": "0.0.32"
 }

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

@@ -1,5 +1,92 @@
 # @llamaindex/autotool-02-next-example
 
+## 0.1.76
+
+### Patch Changes
+
+- Updated dependencies [1364e8e]
+- Updated dependencies [3b7736f]
+- Updated dependencies [96fc69c]
+  - llamaindex@0.7.0
+  - @llamaindex/autotool@4.0.0
+
+## 0.1.75
+
+### Patch Changes
+
+- Updated dependencies [5729bd9]
+  - llamaindex@0.6.22
+  - @llamaindex/autotool@3.0.22
+
+## 0.1.74
+
+### Patch Changes
+
+- Updated dependencies [6f75306]
+- Updated dependencies [94cb4ad]
+  - llamaindex@0.6.21
+  - @llamaindex/autotool@3.0.21
+
+## 0.1.73
+
+### Patch Changes
+
+- Updated dependencies [6a9a7b1]
+  - llamaindex@0.6.20
+  - @llamaindex/autotool@3.0.20
+
+## 0.1.72
+
+### Patch Changes
+
+- Updated dependencies [62cba52]
+- Updated dependencies [d265e96]
+- Updated dependencies [d30bbf7]
+- Updated dependencies [53fd00a]
+  - llamaindex@0.6.19
+  - @llamaindex/autotool@3.0.19
+
+## 0.1.71
+
+### Patch Changes
+
+- Updated dependencies [5f67820]
+- Updated dependencies [fe08d04]
+  - llamaindex@0.6.18
+  - @llamaindex/autotool@3.0.18
+
+## 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": "0.1.76",
   "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": "4.0.0",
   "description": "auto transpile your JS function to LLM Agent compatible",
   "files": [
     "dist",

packages/cloud/CHANGELOG.md

@@ -1,5 +1,41 @@
 # @llamaindex/cloud
 
+## 1.0.0
+
+### Patch Changes
+
+- Updated dependencies [1364e8e]
+- Updated dependencies [96fc69c]
+  - @llamaindex/core@0.3.0
+
+## 0.2.14
+
+### Patch Changes
+
+- Updated dependencies [5f67820]
+  - @llamaindex/core@0.2.12
+
+## 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": "1.0.0",
   "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 +1,42 @@
 # @llamaindex/community
 
+## 0.0.48
+
+### Patch Changes
+
+- Updated dependencies [1364e8e]
+- Updated dependencies [96fc69c]
+  - @llamaindex/core@0.3.0
+
+## 0.0.47
+
+### Patch Changes
+
+- Updated dependencies [5f67820]
+  - @llamaindex/core@0.2.12
+
+## 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": "0.0.48",
   "type": "module",
   "types": "dist/type/index.d.ts",
   "main": "dist/cjs/index.js",

packages/core/CHANGELOG.md

@@ -1,5 +1,37 @@
 # @llamaindex/core
 
+## 0.3.0
+
+### Minor Changes
+
+- 1364e8e: update metadata extractors to use PromptTemplate
+- 96fc69c: add defaultQuestionExtractPrompt
+
+## 0.2.12
+
+### Patch Changes
+
+- 5f67820: Fix that node parsers generate nodes with UUIDs
+
+## 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": "0.3.0",
   "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/index.ts

@@ -12,11 +12,15 @@ export {
   defaultCondenseQuestionPrompt,
   defaultContextSystemPrompt,
   defaultKeywordExtractPrompt,
+  defaultNodeTextTemplate,
   defaultQueryKeywordExtractPrompt,
+  defaultQuestionExtractPrompt,
   defaultRefinePrompt,
   defaultSubQuestionPrompt,
   defaultSummaryPrompt,
   defaultTextQAPrompt,
+  defaultTitleCombinePromptTemplate,
+  defaultTitleExtractorPromptTemplate,
   defaultTreeSummarizePrompt,
 } from "./prompt";
 export type {
@@ -25,9 +29,12 @@ export type {
   ContextSystemPrompt,
   KeywordExtractPrompt,
   QueryKeywordExtractPrompt,
+  QuestionExtractPrompt,
   RefinePrompt,
   SubQuestionPrompt,
   SummaryPrompt,
   TextQAPrompt,
+  TitleCombinePrompt,
+  TitleExtractorPrompt,
   TreeSummarizePrompt,
 } from "./prompt";

packages/core/src/prompts/prompt.ts

@@ -13,8 +13,12 @@ export type CondenseQuestionPrompt = PromptTemplate<
   ["chatHistory", "question"]
 >;
 export type ContextSystemPrompt = PromptTemplate<["context"]>;
-export type KeywordExtractPrompt = PromptTemplate<["context"]>;
+export type KeywordExtractPrompt = PromptTemplate<["context", "maxKeywords"]>;
 export type QueryKeywordExtractPrompt = PromptTemplate<["question"]>;
+export type QuestionExtractPrompt = PromptTemplate<["context", "numQuestions"]>;
+export type TitleExtractorPrompt = PromptTemplate<["context"]>;
+export type TitleCombinePrompt = PromptTemplate<["context"]>;
+export type KeywordExtractorPrompt = PromptTemplate<["context", "numKeywords"]>;
 
 export const defaultTextQAPrompt: TextQAPrompt = new PromptTemplate({
   templateVars: ["context", "query"],
@@ -64,11 +68,13 @@ 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:`,
 });
 
@@ -251,3 +257,55 @@ export const defaultQueryKeywordExtractPrompt = new PromptTemplate({
 }).partialFormat({
   maxKeywords: "10",
 });
+
+export const defaultQuestionExtractPrompt = new PromptTemplate({
+  templateVars: ["numQuestions", "context"],
+  template: `(
+  "Given the contextual informations below, generate {numQuestions} questions this context can provides specific answers to which are unlikely to be found else where. Higher-level summaries of surrounding context may be provided as well. "
+  "Try using these summaries to generate better questions that this context can answer."
+  "---------------------"
+  "{context}"
+  "---------------------"
+  "Provide questions in the following format: 'QUESTIONS: <questions>'"
+)`,
+}).partialFormat({
+  numQuestions: "5",
+});
+
+export const defaultTitleExtractorPromptTemplate = new PromptTemplate({
+  templateVars: ["context"],
+  template: `{context}
+Give a title that summarizes all of the unique entities, titles or themes found in the context. 
+Title: `,
+});
+
+export const defaultTitleCombinePromptTemplate = new PromptTemplate({
+  templateVars: ["context"],
+  template: `{context} 
+Based on the above candidate titles and contents, what is the comprehensive title for this document? 
+Title: `,
+});
+
+export const defaultKeywordExtractorPromptTemplate = new PromptTemplate({
+  templateVars: ["context", "numKeywords"],
+  template: `{context}
+Give {numKeywords} unique keywords for this document. 
+Format as comma separated. 
+Keywords: `,
+}).partialFormat({
+  keywordCount: "5",
+});
+
+export const defaultNodeTextTemplate = new PromptTemplate({
+  templateVars: ["metadataStr", "content"],
+  template: `[Excerpt from document]
+{metadataStr}
+Excerpt:
+-----
+{content}
+-----
+`,
+}).partialFormat({
+  metadataStr: "",
+  content: "",
+});

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/tests/node-parser/text-splitter.test.ts

@@ -2,6 +2,7 @@ import {
   SentenceSplitter,
   splitBySentenceTokenizer,
 } from "@llamaindex/core/node-parser";
+import { Document } from "@llamaindex/core/schema";
 import { describe, expect, test } from "vitest";
 
 describe("sentence splitter", () => {
@@ -115,4 +116,26 @@ describe("sentence splitter", () => {
     const split = splitBySentenceTokenizer();
     expect(split(text)).toEqual([text]);
   });
+
+  test("split nodes with UUID IDs and correct relationships", () => {
+    const UUID_REGEX =
+      /^[0-9a-f]{8}-[0-9a-f]{4}-4[0-9a-f]{3}-[89ab][0-9a-f]{3}-[0-9a-f]{12}$/i;
+    const sentenceSplitter = new SentenceSplitter();
+    const docId = "test-doc-id";
+    const doc = new Document({
+      id_: docId,
+      text: "This is a test sentence. This is another test sentence.",
+    });
+    const nodes = sentenceSplitter.getNodesFromDocuments([doc]);
+    nodes.forEach((node) => {
+      // test node id should match uuid regex
+      expect(node.id_).toMatch(UUID_REGEX);
+
+      // test source reference to the doc ID
+      const source = node.relationships?.SOURCE;
+      expect(source).toBeDefined();
+      expect(source).toHaveProperty("nodeId");
+      expect((source as { nodeId: string }).nodeId).toEqual(docId);
+    });
+  });
 });

packages/experimental/CHANGELOG.md

@@ -1,5 +1,82 @@
 # @llamaindex/experimental
 
+## 0.0.101
+
+### Patch Changes
+
+- Updated dependencies [1364e8e]
+- Updated dependencies [3b7736f]
+- Updated dependencies [96fc69c]
+  - llamaindex@0.7.0
+
+## 0.0.100
+
+### Patch Changes
+
+- Updated dependencies [5729bd9]
+  - llamaindex@0.6.22
+
+## 0.0.99
+
+### Patch Changes
+
+- Updated dependencies [6f75306]
+- Updated dependencies [94cb4ad]
+  - llamaindex@0.6.21
+
+## 0.0.98
+
+### Patch Changes
+
+- Updated dependencies [6a9a7b1]
+  - llamaindex@0.6.20
+
+## 0.0.97
+
+### Patch Changes
+
+- Updated dependencies [62cba52]
+- Updated dependencies [d265e96]
+- Updated dependencies [d30bbf7]
+- Updated dependencies [53fd00a]
+  - llamaindex@0.6.19
+
+## 0.0.96
+
+### Patch Changes
+
+- Updated dependencies [5f67820]
+- Updated dependencies [fe08d04]
+  - llamaindex@0.6.18
+
+## 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": "0.0.101",
   "type": "module",
   "types": "dist/type/index.d.ts",
   "main": "dist/cjs/index.js",

packages/llamaindex/CHANGELOG.md

@@ -1,5 +1,114 @@
 # llamaindex
 
+## 0.7.0
+
+### Minor Changes
+
+- 1364e8e: update metadata extractors to use PromptTemplate
+- 96fc69c: Correct initialization of QuestionsAnsweredExtractor so that it uses the promptTemplate arg when passed in
+
+### Patch Changes
+
+- 3b7736f: feat: added gemini 002 support
+- Updated dependencies [1364e8e]
+- Updated dependencies [96fc69c]
+  - @llamaindex/core@0.3.0
+  - @llamaindex/cloud@1.0.0
+  - @llamaindex/ollama@0.0.8
+  - @llamaindex/openai@0.1.17
+  - @llamaindex/groq@0.0.16
+
+## 0.6.22
+
+### Patch Changes
+
+- 5729bd9: Fix LlamaCloud API calls for ensuring an index and for file uploads
+
+## 0.6.21
+
+### Patch Changes
+
+- 6f75306: feat: support metadata filters for AstraDB
+- 94cb4ad: feat: Add metadata filters to ChromaDb and update to 1.9.2
+
+## 0.6.20
+
+### Patch Changes
+
+- 6a9a7b1: fix: take init api key into account
+- Updated dependencies [6a9a7b1]
+  - @llamaindex/openai@0.1.16
+  - @llamaindex/groq@0.0.15
+
+## 0.6.19
+
+### Patch Changes
+
+- 62cba52: Add ensureIndex function to LlamaCloudIndex
+- d265e96: fix: ignore resolving unpdf for nextjs
+- d30bbf7: Convert undefined values to null in LlamaCloud filters
+- 53fd00a: Fix getPipelineId in LlamaCloudIndex
+
+## 0.6.18
+
+### Patch Changes
+
+- 5f67820: Fix that node parsers generate nodes with UUIDs
+- fe08d04: Fix LlamaCloud retrieval with multiple pipelines
+- Updated dependencies [5f67820]
+  - @llamaindex/core@0.2.12
+  - @llamaindex/cloud@0.2.14
+  - @llamaindex/ollama@0.0.7
+  - @llamaindex/openai@0.1.15
+  - @llamaindex/groq@0.0.14
+
+## 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 +1,82 @@
 # @llamaindex/cloudflare-worker-agent-test
 
+## 0.0.85
+
+### Patch Changes
+
+- Updated dependencies [1364e8e]
+- Updated dependencies [3b7736f]
+- Updated dependencies [96fc69c]
+  - llamaindex@0.7.0
+
+## 0.0.84
+
+### Patch Changes
+
+- Updated dependencies [5729bd9]
+  - llamaindex@0.6.22
+
+## 0.0.83
+
+### Patch Changes
+
+- Updated dependencies [6f75306]
+- Updated dependencies [94cb4ad]
+  - llamaindex@0.6.21
+
+## 0.0.82
+
+### Patch Changes
+
+- Updated dependencies [6a9a7b1]
+  - llamaindex@0.6.20
+
+## 0.0.81
+
+### Patch Changes
+
+- Updated dependencies [62cba52]
+- Updated dependencies [d265e96]
+- Updated dependencies [d30bbf7]
+- Updated dependencies [53fd00a]
+  - llamaindex@0.6.19
+
+## 0.0.80
+
+### Patch Changes
+
+- Updated dependencies [5f67820]
+- Updated dependencies [fe08d04]
+  - llamaindex@0.6.18
+
+## 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": "0.0.85",
   "type": "module",
   "private": true,
   "scripts": {

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

@@ -1,5 +1,36 @@
 # @llamaindex/llama-parse-browser-test
 
+## 0.0.11
+
+### Patch Changes
+
+- @llamaindex/cloud@1.0.0
+
+## 0.0.10
+
+### Patch Changes
+
+- @llamaindex/cloud@0.2.14
+
+## 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": "0.0.11",
   "type": "module",
   "scripts": {
     "dev": "vite",

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

@@ -1,5 +1,82 @@
 # @llamaindex/next-agent-test
 
+## 0.1.85
+
+### Patch Changes
+
+- Updated dependencies [1364e8e]
+- Updated dependencies [3b7736f]
+- Updated dependencies [96fc69c]
+  - llamaindex@0.7.0
+
+## 0.1.84
+
+### Patch Changes
+
+- Updated dependencies [5729bd9]
+  - llamaindex@0.6.22
+
+## 0.1.83
+
+### Patch Changes
+
+- Updated dependencies [6f75306]
+- Updated dependencies [94cb4ad]
+  - llamaindex@0.6.21
+
+## 0.1.82
+
+### Patch Changes
+
+- Updated dependencies [6a9a7b1]
+  - llamaindex@0.6.20
+
+## 0.1.81
+
+### Patch Changes
+
+- Updated dependencies [62cba52]
+- Updated dependencies [d265e96]
+- Updated dependencies [d30bbf7]
+- Updated dependencies [53fd00a]
+  - llamaindex@0.6.19
+
+## 0.1.80
+
+### Patch Changes
+
+- Updated dependencies [5f67820]
+- Updated dependencies [fe08d04]
+  - llamaindex@0.6.18
+
+## 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": "0.1.85",
   "private": true,
   "scripts": {
     "dev": "next dev",

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

@@ -1,5 +1,82 @@
 # test-edge-runtime
 
+## 0.1.84
+
+### Patch Changes
+
+- Updated dependencies [1364e8e]
+- Updated dependencies [3b7736f]
+- Updated dependencies [96fc69c]
+  - llamaindex@0.7.0
+
+## 0.1.83
+
+### Patch Changes
+
+- Updated dependencies [5729bd9]
+  - llamaindex@0.6.22
+
+## 0.1.82
+
+### Patch Changes
+
+- Updated dependencies [6f75306]
+- Updated dependencies [94cb4ad]
+  - llamaindex@0.6.21
+
+## 0.1.81
+
+### Patch Changes
+
+- Updated dependencies [6a9a7b1]
+  - llamaindex@0.6.20
+
+## 0.1.80
+
+### Patch Changes
+
+- Updated dependencies [62cba52]
+- Updated dependencies [d265e96]
+- Updated dependencies [d30bbf7]
+- Updated dependencies [53fd00a]
+  - llamaindex@0.6.19
+
+## 0.1.79
+
+### Patch Changes
+
+- Updated dependencies [5f67820]
+- Updated dependencies [fe08d04]
+  - llamaindex@0.6.18
+
+## 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": "0.1.84",
   "private": true,
   "scripts": {
     "dev": "next dev",

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

@@ -1,5 +1,82 @@
 # @llamaindex/next-node-runtime
 
+## 0.0.66
+
+### Patch Changes
+
+- Updated dependencies [1364e8e]
+- Updated dependencies [3b7736f]
+- Updated dependencies [96fc69c]
+  - llamaindex@0.7.0
+
+## 0.0.65
+
+### Patch Changes
+
+- Updated dependencies [5729bd9]
+  - llamaindex@0.6.22
+
+## 0.0.64
+
+### Patch Changes
+
+- Updated dependencies [6f75306]
+- Updated dependencies [94cb4ad]
+  - llamaindex@0.6.21
+
+## 0.0.63
+
+### Patch Changes
+
+- Updated dependencies [6a9a7b1]
+  - llamaindex@0.6.20
+
+## 0.0.62
+
+### Patch Changes
+
+- Updated dependencies [62cba52]
+- Updated dependencies [d265e96]
+- Updated dependencies [d30bbf7]
+- Updated dependencies [53fd00a]
+  - llamaindex@0.6.19
+
+## 0.0.61
+
+### Patch Changes
+
+- Updated dependencies [5f67820]
+- Updated dependencies [fe08d04]
+  - llamaindex@0.6.18
+
+## 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": "0.0.66",
   "private": true,
   "scripts": {
     "dev": "next dev",

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

@@ -1,5 +1,82 @@
 # @llamaindex/waku-query-engine-test
 
+## 0.0.85
+
+### Patch Changes
+
+- Updated dependencies [1364e8e]
+- Updated dependencies [3b7736f]
+- Updated dependencies [96fc69c]
+  - llamaindex@0.7.0
+
+## 0.0.84
+
+### Patch Changes
+
+- Updated dependencies [5729bd9]
+  - llamaindex@0.6.22
+
+## 0.0.83
+
+### Patch Changes
+
+- Updated dependencies [6f75306]
+- Updated dependencies [94cb4ad]
+  - llamaindex@0.6.21
+
+## 0.0.82
+
+### Patch Changes
+
+- Updated dependencies [6a9a7b1]
+  - llamaindex@0.6.20
+
+## 0.0.81
+
+### Patch Changes
+
+- Updated dependencies [62cba52]
+- Updated dependencies [d265e96]
+- Updated dependencies [d30bbf7]
+- Updated dependencies [53fd00a]
+  - llamaindex@0.6.19
+
+## 0.0.80
+
+### Patch Changes
+
+- Updated dependencies [5f67820]
+- Updated dependencies [fe08d04]
+  - llamaindex@0.6.18
+
+## 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": "0.0.85",
   "type": "module",
   "private": true,
   "scripts": {

packages/llamaindex/package.json

@@ -1,6 +1,6 @@
 {
   "name": "llamaindex",
-  "version": "0.6.13",
+  "version": "0.7.0",
   "license": "MIT",
   "type": "module",
   "keywords": [
@@ -48,7 +48,7 @@
     "@zilliz/milvus2-sdk-node": "^2.4.6",
     "ajv": "^8.17.1",
     "assemblyai": "^4.7.0",
-    "chromadb": "1.8.1",
+    "chromadb": "1.9.2",
     "cohere-ai": "7.13.0",
     "discord-api-types": "^0.37.98",
     "groq-sdk": "^0.6.1",

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/LLamaCloudFileService.ts

@@ -41,7 +41,7 @@ export class LLamaCloudFileService {
   ) {
     initService();
     const { data: file } = await FilesService.uploadFileApiV1FilesPost({
-      path: { project_id: projectId },
+      query: { project_id: projectId },
       body: {
         upload_file: uploadFile,
       },
@@ -85,7 +85,7 @@ export class LLamaCloudFileService {
       await new Promise((resolve) => setTimeout(resolve, 100)); // Sleep for 100ms
     }
     throw new Error(
-      `File processing did not complete after ${maxAttempts} attempts.`,
+      `File processing did not complete after ${maxAttempts} attempts. Check your LlamaCloud index at https://cloud.llamaindex.ai/project/${projectId}/deploy/${pipelineId} for more details.`,
     );
   }
 

packages/llamaindex/src/cloud/LlamaCloudIndex.ts

@@ -1,19 +1,21 @@
 import type { BaseQueryEngine } from "@llamaindex/core/query-engine";
 import type { BaseSynthesizer } from "@llamaindex/core/response-synthesizers";
-import type { Document, TransformComponent } from "@llamaindex/core/schema";
+import type { Document } from "@llamaindex/core/schema";
 import { RetrieverQueryEngine } from "../engines/query/RetrieverQueryEngine.js";
 import type { BaseNodePostprocessor } from "../postprocessors/types.js";
 import type { CloudRetrieveParams } from "./LlamaCloudRetriever.js";
 import { LlamaCloudRetriever } from "./LlamaCloudRetriever.js";
-import { getPipelineCreate } from "./config.js";
 import type { CloudConstructorParams } from "./type.js";
-import { getAppBaseUrl, getProjectId, initService } from "./utils.js";
+import {
+  getAppBaseUrl,
+  getPipelineId,
+  getProjectId,
+  initService,
+} from "./utils.js";
 
-import { PipelinesService, ProjectsService } from "@llamaindex/cloud/api";
-import { SentenceSplitter } from "@llamaindex/core/node-parser";
+import { PipelinesService, type PipelineCreate } from "@llamaindex/cloud/api";
 import type { BaseRetriever } from "@llamaindex/core/retriever";
 import { getEnv } from "@llamaindex/env";
-import { OpenAIEmbedding } from "@llamaindex/openai";
 import { Settings } from "../Settings.js";
 
 export class LlamaCloudIndex {
@@ -28,10 +30,7 @@ export class LlamaCloudIndex {
     verbose = Settings.debug,
     raiseOnError = false,
   ): Promise<void> {
-    const pipelineId = await this.getPipelineId(
-      this.params.name,
-      this.params.projectName,
-    );
+    const pipelineId = await this.getPipelineId();
 
     if (verbose) {
       console.log("Waiting for pipeline ingestion: ");
@@ -78,10 +77,7 @@ export class LlamaCloudIndex {
     verbose = Settings.debug,
     raiseOnError = false,
   ): Promise<void> {
-    const pipelineId = await this.getPipelineId(
-      this.params.name,
-      this.params.projectName,
-    );
+    const pipelineId = await this.getPipelineId();
 
     if (verbose) {
       console.log("Loading data: ");
@@ -143,17 +139,13 @@ export class LlamaCloudIndex {
   public async getPipelineId(
     name?: string,
     projectName?: string,
+    organizationId?: string,
   ): Promise<string> {
-    const { data: pipelines } =
-      await PipelinesService.searchPipelinesApiV1PipelinesGet({
-        path: {
-          project_id: await this.getProjectId(projectName),
-          project_name: name ?? this.params.name,
-        },
-        throwOnError: true,
-      });
-
-    return pipelines[0]!.id;
+    return await getPipelineId(
+      name ?? this.params.name,
+      projectName ?? this.params.projectName,
+      organizationId ?? this.params.organizationId,
+    );
   }
 
   public async getProjectId(
@@ -166,75 +158,42 @@ export class LlamaCloudIndex {
     );
   }
 
+  /**
+   * Adds documents to the given index parameters. If the index does not exist, it will be created.
+   *
+   * @param params - An object containing the following properties:
+   *   - documents: An array of Document objects to be added to the index.
+   *   - verbose: Optional boolean to enable verbose logging.
+   *   - Additional properties from CloudConstructorParams.
+   * @returns A Promise that resolves to a new LlamaCloudIndex instance.
+   */
   static async fromDocuments(
     params: {
       documents: Document[];
-      transformations?: TransformComponent[];
       verbose?: boolean;
     } & CloudConstructorParams,
+    config?: {
+      embedding: PipelineCreate["embedding_config"];
+      transform: PipelineCreate["transform_config"];
+    },
   ): Promise<LlamaCloudIndex> {
-    initService(params);
-    const defaultTransformations: TransformComponent[] = [
-      new SentenceSplitter(),
-      new OpenAIEmbedding({
-        apiKey: getEnv("OPENAI_API_KEY"),
-      }),
-    ];
+    const index = new LlamaCloudIndex({ ...params });
+    await index.ensureIndex({ ...config, verbose: params.verbose ?? false });
+    await index.addDocuments(params.documents, params.verbose);
+    return index;
+  }
+
+  async addDocuments(documents: Document[], verbose?: boolean): Promise<void> {
     const apiUrl = getAppBaseUrl();
-
-    const pipelineCreateParams = await getPipelineCreate({
-      pipelineName: params.name,
-      pipelineType: "MANAGED",
-      inputNodes: params.documents,
-      transformations: params.transformations ?? defaultTransformations,
-    });
-
-    const { data: project } =
-      await ProjectsService.upsertProjectApiV1ProjectsPut({
-        path: {
-          organization_id: params.organizationId,
-        },
-        body: {
-          name: params.projectName ?? "default",
-        },
-        throwOnError: true,
-      });
-
-    if (!project.id) {
-      throw new Error("Project ID should be defined");
-    }
-
-    const { data: pipeline } =
-      await PipelinesService.upsertPipelineApiV1PipelinesPut({
-        path: {
-          project_id: project.id,
-        },
-        body: pipelineCreateParams.configured_transformations
-          ? {
-              name: params.name,
-              configured_transformations:
-                pipelineCreateParams.configured_transformations,
-            }
-          : {
-              name: params.name,
-            },
-        throwOnError: true,
-      });
-
-    if (!pipeline.id) {
-      throw new Error("Pipeline ID must be defined");
-    }
-
-    if (params.verbose) {
-      console.log(`Created pipeline ${pipeline.id} with name ${params.name}`);
-    }
+    const projectId = await this.getProjectId();
+    const pipelineId = await this.getPipelineId();
 
     await PipelinesService.upsertBatchPipelineDocumentsApiV1PipelinesPipelineIdDocumentsPut(
       {
         path: {
-          pipeline_id: pipeline.id,
+          pipeline_id: pipelineId,
         },
-        body: params.documents.map((doc) => ({
+        body: documents.map((doc) => ({
           metadata: doc.metadata,
           text: doc.text,
           excluded_embed_metadata_keys: doc.excludedEmbedMetadataKeys,
@@ -248,7 +207,7 @@ export class LlamaCloudIndex {
       const { data: pipelineStatus } =
         await PipelinesService.getPipelineStatusApiV1PipelinesPipelineIdStatusGet(
           {
-            path: { pipeline_id: pipeline.id },
+            path: { pipeline_id: pipelineId },
             throwOnError: true,
           },
         );
@@ -262,32 +221,30 @@ export class LlamaCloudIndex {
 
       if (pipelineStatus.status === "ERROR") {
         console.error(
-          `Some documents failed to ingest, check your pipeline logs at ${apiUrl}/project/${project.id}/deploy/${pipeline.id}`,
+          `Some documents failed to ingest, check your pipeline logs at ${apiUrl}/project/${projectId}/deploy/${pipelineId}`,
         );
         throw new Error("Some documents failed to ingest");
       }
 
       if (pipelineStatus.status === "PARTIAL_SUCCESS") {
         console.info(
-          `Documents ingestion partially succeeded, to check a more complete status check your pipeline at ${apiUrl}/project/${project.id}/deploy/${pipeline.id}`,
+          `Documents ingestion partially succeeded, to check a more complete status check your pipeline at ${apiUrl}/project/${projectId}/deploy/${pipelineId}`,
         );
         break;
       }
 
-      if (params.verbose) {
+      if (verbose) {
         process.stdout.write(".");
       }
 
       await new Promise((resolve) => setTimeout(resolve, 1000));
     }
 
-    if (params.verbose) {
+    if (verbose) {
       console.info(
-        `Ingestion completed, find your index at ${apiUrl}/project/${project.id}/deploy/${pipeline.id}`,
+        `Ingestion completed, find your index at ${apiUrl}/project/${projectId}/deploy/${pipelineId}`,
       );
     }
-
-    return new LlamaCloudIndex({ ...params });
   }
 
   asRetriever(params: CloudRetrieveParams = {}): BaseRetriever {
@@ -308,20 +265,12 @@ export class LlamaCloudIndex {
     return new RetrieverQueryEngine(
       retriever,
       params?.responseSynthesizer,
-      params?.preFilters,
       params?.nodePostprocessors,
     );
   }
 
   async insert(document: Document) {
-    const pipelineId = await this.getPipelineId(
-      this.params.name,
-      this.params.projectName,
-    );
-
-    if (!pipelineId) {
-      throw new Error("We couldn't find the pipeline ID for the given name");
-    }
+    const pipelineId = await this.getPipelineId();
 
     await PipelinesService.createBatchPipelineDocumentsApiV1PipelinesPipelineIdDocumentsPost(
       {
@@ -344,14 +293,7 @@ export class LlamaCloudIndex {
   }
 
   async delete(document: Document) {
-    const pipelineId = await this.getPipelineId(
-      this.params.name,
-      this.params.projectName,
-    );
-
-    if (!pipelineId) {
-      throw new Error("We couldn't find the pipeline ID for the given name");
-    }
+    const pipelineId = await this.getPipelineId();
 
     await PipelinesService.deletePipelineDocumentApiV1PipelinesPipelineIdDocumentsDocumentIdDelete(
       {
@@ -366,14 +308,7 @@ export class LlamaCloudIndex {
   }
 
   async refreshDoc(document: Document) {
-    const pipelineId = await this.getPipelineId(
-      this.params.name,
-      this.params.projectName,
-    );
-
-    if (!pipelineId) {
-      throw new Error("We couldn't find the pipeline ID for the given name");
-    }
+    const pipelineId = await this.getPipelineId();
 
     await PipelinesService.upsertBatchPipelineDocumentsApiV1PipelinesPipelineIdDocumentsPut(
       {
@@ -394,4 +329,71 @@ export class LlamaCloudIndex {
 
     await this.waitForDocumentIngestion([document.id_]);
   }
+
+  public async ensureIndex(config?: {
+    embedding?: PipelineCreate["embedding_config"];
+    transform?: PipelineCreate["transform_config"];
+    verbose?: boolean;
+  }): Promise<void> {
+    const projectId = await this.getProjectId();
+
+    const { data: pipelines } =
+      await PipelinesService.searchPipelinesApiV1PipelinesGet({
+        query: {
+          project_id: projectId,
+          pipeline_name: this.params.name,
+        },
+        throwOnError: true,
+      });
+
+    if (pipelines.length === 0) {
+      // no pipeline found, create a new one
+      let embeddingConfig = config?.embedding;
+      if (!embeddingConfig) {
+        // no embedding config provided, use OpenAI as default
+        const openAIApiKey = getEnv("OPENAI_API_KEY");
+        const embeddingModel = getEnv("EMBEDDING_MODEL");
+        if (!openAIApiKey || !embeddingModel) {
+          throw new Error(
+            "No embedding configuration provided. Fallback to OpenAI embedding model. OPENAI_API_KEY and EMBEDDING_MODEL environment variables must be set.",
+          );
+        }
+        embeddingConfig = {
+          type: "OPENAI_EMBEDDING",
+          component: {
+            api_key: openAIApiKey,
+            model_name: embeddingModel,
+          },
+        };
+      }
+
+      let transformConfig = config?.transform;
+      if (!transformConfig) {
+        transformConfig = {
+          mode: "auto",
+          chunk_size: 1024,
+          chunk_overlap: 200,
+        };
+      }
+
+      const { data: pipeline } =
+        await PipelinesService.upsertPipelineApiV1PipelinesPut({
+          query: {
+            project_id: projectId,
+          },
+          body: {
+            name: this.params.name,
+            embedding_config: embeddingConfig,
+            transform_config: transformConfig,
+          },
+          throwOnError: true,
+        });
+
+      if (config?.verbose) {
+        console.log(
+          `Created pipeline ${pipeline.id} with name ${pipeline.name}`,
+        );
+      }
+    }
+  }
 }

packages/llamaindex/src/cloud/LlamaCloudRetriever.ts

@@ -1,4 +1,5 @@
 import {
+  type MetadataFilter,
   type MetadataFilters,
   PipelinesService,
   type RetrievalParams,
@@ -11,7 +12,7 @@ import type { NodeWithScore } from "@llamaindex/core/schema";
 import { jsonToNode, ObjectType } from "@llamaindex/core/schema";
 import { extractText } from "@llamaindex/core/utils";
 import type { ClientParams, CloudConstructorParams } from "./type.js";
-import { getProjectId, initService } from "./utils.js";
+import { getPipelineId, initService } from "./utils.js";
 
 export type CloudRetrieveParams = Omit<
   RetrievalParams,
@@ -42,6 +43,24 @@ export class LlamaCloudRetriever extends BaseRetriever {
     });
   }
 
+  // LlamaCloud expects null values for filters, but LlamaIndexTS uses undefined for empty values
+  // This function converts the undefined values to null
+  private convertFilter(filters?: MetadataFilters): MetadataFilters | null {
+    if (!filters) return null;
+
+    const processFilter = (
+      filter: MetadataFilter | MetadataFilters,
+    ): MetadataFilter | MetadataFilters => {
+      if ("filters" in filter) {
+        // type MetadataFilters
+        return { ...filter, filters: filter.filters.map(processFilter) };
+      }
+      return { ...filter, value: filter.value ?? null };
+    };
+
+    return { ...filters, filters: filters.filters.map(processFilter) };
+  }
+
   constructor(params: CloudConstructorParams & CloudRetrieveParams) {
     super();
     this.clientParams = { apiKey: params.apiKey, baseUrl: params.baseUrl };
@@ -57,45 +76,24 @@ export class LlamaCloudRetriever extends BaseRetriever {
   }
 
   async _retrieve(query: QueryBundle): Promise<NodeWithScore[]> {
-    const { data: pipelines } =
-      await PipelinesService.searchPipelinesApiV1PipelinesGet({
-        query: {
-          project_id: await getProjectId(this.projectName, this.organizationId),
-          project_name: this.pipelineName,
-        },
-        throwOnError: true,
-      });
+    const pipelineId = await getPipelineId(
+      this.pipelineName,
+      this.projectName,
+      this.organizationId,
+    );
 
-    if (pipelines.length === 0 || !pipelines[0]!.id) {
-      throw new Error(
-        `No pipeline found with name ${this.pipelineName} in project ${this.projectName}`,
-      );
-    }
-
-    const { data: pipeline } =
-      await PipelinesService.getPipelineApiV1PipelinesPipelineIdGet({
-        path: {
-          pipeline_id: pipelines[0]!.id,
-        },
-        throwOnError: true,
-      });
-
-    if (!pipeline) {
-      throw new Error(
-        `No pipeline found with name ${this.pipelineName} in project ${this.projectName}`,
-      );
-    }
+    const filters = this.convertFilter(this.retrieveParams.filters);
 
     const { data: results } =
       await PipelinesService.runSearchApiV1PipelinesPipelineIdRetrievePost({
         throwOnError: true,
         path: {
-          pipeline_id: pipeline.id,
+          pipeline_id: pipelineId,
         },
         body: {
           ...this.retrieveParams,
           query: extractText(query),
-          search_filters: this.retrieveParams.filters as MetadataFilters,
+          search_filters: filters,
           dense_similarity_top_k: this.retrieveParams.similarityTopK!,
         },
       });

packages/llamaindex/src/cloud/config.ts

@@ -1,55 +0,0 @@
-import type {
-  ConfiguredTransformationItem,
-  PipelineCreate,
-  PipelineType,
-} from "@llamaindex/cloud/api";
-import { SentenceSplitter } from "@llamaindex/core/node-parser";
-import { BaseNode, type TransformComponent } from "@llamaindex/core/schema";
-import { OpenAIEmbedding } from "@llamaindex/openai";
-
-export type GetPipelineCreateParams = {
-  pipelineName: string;
-  pipelineType: PipelineType;
-  transformations?: TransformComponent[];
-  inputNodes?: BaseNode[];
-};
-
-function getTransformationConfig(
-  transformation: TransformComponent,
-): ConfiguredTransformationItem {
-  if (transformation instanceof SentenceSplitter) {
-    return {
-      configurable_transformation_type: "SENTENCE_AWARE_NODE_PARSER",
-      component: {
-        chunk_size: transformation.chunkSize, // TODO: set to public in SentenceSplitter
-        chunk_overlap: transformation.chunkOverlap, // TODO: set to public in SentenceSplitter
-        include_metadata: transformation.includeMetadata,
-        include_prev_next_rel: transformation.includePrevNextRel,
-      },
-    };
-  }
-  if (transformation instanceof OpenAIEmbedding) {
-    return {
-      configurable_transformation_type: "OPENAI_EMBEDDING",
-      component: {
-        model: transformation.model,
-        api_key: transformation.apiKey,
-        embed_batch_size: transformation.embedBatchSize,
-        dimensions: transformation.dimensions,
-      },
-    };
-  }
-  throw new Error(`Unsupported transformation: ${typeof transformation}`);
-}
-
-export async function getPipelineCreate(
-  params: GetPipelineCreateParams,
-): Promise<PipelineCreate> {
-  const { pipelineName, pipelineType, transformations = [] } = params;
-
-  return {
-    name: pipelineName,
-    configured_transformations: transformations.map(getTransformationConfig),
-    pipeline_type: pipelineType,
-  };
-}

packages/llamaindex/src/cloud/utils.ts

@@ -1,4 +1,8 @@
-import { client, ProjectsService } from "@llamaindex/cloud/api";
+import {
+  client,
+  PipelinesService,
+  ProjectsService,
+} from "@llamaindex/cloud/api";
 import { DEFAULT_BASE_URL } from "@llamaindex/core/global";
 import { getEnv } from "@llamaindex/env";
 import type { ClientParams } from "./type.js";
@@ -40,9 +44,9 @@ export async function getProjectId(
 ): Promise<string> {
   const { data: projects } = await ProjectsService.listProjectsApiV1ProjectsGet(
     {
-      path: {
+      query: {
         project_name: projectName,
-        organization_id: organizationId,
+        organization_id: organizationId ?? null,
       },
       throwOnError: true,
     },
@@ -66,3 +70,26 @@ export async function getProjectId(
 
   return project.id;
 }
+
+export async function getPipelineId(
+  name: string,
+  projectName: string,
+  organizationId?: string,
+): Promise<string> {
+  const { data: pipelines } =
+    await PipelinesService.searchPipelinesApiV1PipelinesGet({
+      query: {
+        project_id: await getProjectId(projectName, organizationId),
+        pipeline_name: name,
+      },
+      throwOnError: true,
+    });
+
+  if (pipelines.length === 0 || !pipelines[0]!.id) {
+    throw new Error(
+      `No pipeline found with name ${name} in project ${projectName}`,
+    );
+  }
+
+  return pipelines[0]!.id;
+}

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/extractors/MetadataExtractors.ts

@@ -1,14 +1,20 @@
 import type { LLM } from "@llamaindex/core/llms";
+import {
+  PromptTemplate,
+  defaultKeywordExtractPrompt,
+  defaultQuestionExtractPrompt,
+  defaultSummaryPrompt,
+  defaultTitleCombinePromptTemplate,
+  defaultTitleExtractorPromptTemplate,
+  type KeywordExtractPrompt,
+  type QuestionExtractPrompt,
+  type SummaryPrompt,
+  type TitleCombinePrompt,
+  type TitleExtractorPrompt,
+} from "@llamaindex/core/prompts";
 import type { BaseNode } from "@llamaindex/core/schema";
 import { MetadataMode, TextNode } from "@llamaindex/core/schema";
 import { OpenAI } from "@llamaindex/openai";
-import {
-  defaultKeywordExtractorPromptTemplate,
-  defaultQuestionAnswerPromptTemplate,
-  defaultSummaryExtractorPromptTemplate,
-  defaultTitleCombinePromptTemplate,
-  defaultTitleExtractorPromptTemplate,
-} from "./prompts.js";
 import { BaseExtractor } from "./types.js";
 
 const STRIP_REGEX = /(\r\n|\n|\r)/gm;
@@ -16,6 +22,7 @@ const STRIP_REGEX = /(\r\n|\n|\r)/gm;
 type KeywordExtractArgs = {
   llm?: LLM;
   keywords?: number;
+  promptTemplate?: KeywordExtractPrompt["template"];
 };
 
 type ExtractKeyword = {
@@ -39,6 +46,12 @@ export class KeywordExtractor extends BaseExtractor {
    */
   keywords: number = 5;
 
+  /**
+   * The prompt template to use for the question extractor.
+   * @type {string}
+   */
+  promptTemplate: KeywordExtractPrompt;
+
   /**
    * Constructor for the KeywordExtractor class.
    * @param {LLM} llm LLM instance.
@@ -53,6 +66,12 @@ export class KeywordExtractor extends BaseExtractor {
 
     this.llm = options?.llm ?? new OpenAI();
     this.keywords = options?.keywords ?? 5;
+    this.promptTemplate = options?.promptTemplate
+      ? new PromptTemplate({
+          templateVars: ["context", "maxKeywords"],
+          template: options.promptTemplate,
+        })
+      : defaultKeywordExtractPrompt;
   }
 
   /**
@@ -66,9 +85,9 @@ export class KeywordExtractor extends BaseExtractor {
     }
 
     const completion = await this.llm.complete({
-      prompt: defaultKeywordExtractorPromptTemplate({
-        contextStr: node.getContent(MetadataMode.ALL),
-        keywords: this.keywords,
+      prompt: this.promptTemplate.format({
+        context: node.getContent(MetadataMode.ALL),
+        maxKeywords: this.keywords.toString(),
       }),
     });
 
@@ -93,8 +112,8 @@ export class KeywordExtractor extends BaseExtractor {
 type TitleExtractorsArgs = {
   llm?: LLM;
   nodes?: number;
-  nodeTemplate?: string;
-  combineTemplate?: string;
+  nodeTemplate?: TitleExtractorPrompt["template"];
+  combineTemplate?: TitleCombinePrompt["template"];
 };
 
 type ExtractTitle = {
@@ -129,19 +148,19 @@ export class TitleExtractor extends BaseExtractor {
    * The prompt template to use for the title extractor.
    * @type {string}
    */
-  nodeTemplate: string;
+  nodeTemplate: TitleExtractorPrompt;
 
   /**
    * The prompt template to merge title with..
    * @type {string}
    */
-  combineTemplate: string;
+  combineTemplate: TitleCombinePrompt;
 
   /**
    * Constructor for the TitleExtractor class.
    * @param {LLM} llm LLM instance.
    * @param {number} nodes Number of nodes to extract titles from.
-   * @param {string} nodeTemplate The prompt template to use for the title extractor.
+   * @param {TitleExtractorPrompt} nodeTemplate The prompt template to use for the title extractor.
    * @param {string} combineTemplate The prompt template to merge title with..
    */
   constructor(options?: TitleExtractorsArgs) {
@@ -150,10 +169,19 @@ export class TitleExtractor extends BaseExtractor {
     this.llm = options?.llm ?? new OpenAI();
     this.nodes = options?.nodes ?? 5;
 
-    this.nodeTemplate =
-      options?.nodeTemplate ?? defaultTitleExtractorPromptTemplate();
-    this.combineTemplate =
-      options?.combineTemplate ?? defaultTitleCombinePromptTemplate();
+    this.nodeTemplate = options?.nodeTemplate
+      ? new PromptTemplate({
+          templateVars: ["context"],
+          template: options.nodeTemplate,
+        })
+      : defaultTitleExtractorPromptTemplate;
+
+    this.combineTemplate = options?.combineTemplate
+      ? new PromptTemplate({
+          templateVars: ["context"],
+          template: options.combineTemplate,
+        })
+      : defaultTitleCombinePromptTemplate;
   }
 
   /**
@@ -218,8 +246,8 @@ export class TitleExtractor extends BaseExtractor {
       const titleCandidates = await this.getTitlesCandidates(nodes);
       const combinedTitles = titleCandidates.join(", ");
       const completion = await this.llm.complete({
-        prompt: defaultTitleCombinePromptTemplate({
-          contextStr: combinedTitles,
+        prompt: this.combineTemplate.format({
+          context: combinedTitles,
         }),
       });
 
@@ -232,8 +260,8 @@ export class TitleExtractor extends BaseExtractor {
   private async getTitlesCandidates(nodes: BaseNode[]): Promise<string[]> {
     const titleJobs = nodes.map(async (node) => {
       const completion = await this.llm.complete({
-        prompt: defaultTitleExtractorPromptTemplate({
-          contextStr: node.getContent(MetadataMode.ALL),
+        prompt: this.nodeTemplate.format({
+          context: node.getContent(MetadataMode.ALL),
         }),
       });
 
@@ -247,7 +275,7 @@ export class TitleExtractor extends BaseExtractor {
 type QuestionAnswerExtractArgs = {
   llm?: LLM;
   questions?: number;
-  promptTemplate?: string;
+  promptTemplate?: QuestionExtractPrompt["template"];
   embeddingOnly?: boolean;
 };
 
@@ -276,7 +304,7 @@ export class QuestionsAnsweredExtractor extends BaseExtractor {
    * The prompt template to use for the question extractor.
    * @type {string}
    */
-  promptTemplate: string;
+  promptTemplate: QuestionExtractPrompt;
 
   /**
    * Wheter to use metadata for embeddings only
@@ -289,7 +317,7 @@ export class QuestionsAnsweredExtractor extends BaseExtractor {
    * Constructor for the QuestionsAnsweredExtractor class.
    * @param {LLM} llm LLM instance.
    * @param {number} questions Number of questions to generate.
-   * @param {string} promptTemplate The prompt template to use for the question extractor.
+   * @param {TextQAPrompt} promptTemplate The prompt template to use for the question extractor.
    * @param {boolean} embeddingOnly Wheter to use metadata for embeddings only.
    */
   constructor(options?: QuestionAnswerExtractArgs) {
@@ -300,12 +328,14 @@ export class QuestionsAnsweredExtractor extends BaseExtractor {
 
     this.llm = options?.llm ?? new OpenAI();
     this.questions = options?.questions ?? 5;
-    this.promptTemplate =
-      options?.promptTemplate ??
-      defaultQuestionAnswerPromptTemplate({
-        numQuestions: this.questions,
-        contextStr: "",
-      });
+    this.promptTemplate = options?.promptTemplate
+      ? new PromptTemplate({
+          templateVars: ["numQuestions", "context"],
+          template: options.promptTemplate,
+        }).partialFormat({
+          numQuestions: "5",
+        })
+      : defaultQuestionExtractPrompt;
     this.embeddingOnly = options?.embeddingOnly ?? false;
   }
 
@@ -323,9 +353,9 @@ export class QuestionsAnsweredExtractor extends BaseExtractor {
 
     const contextStr = node.getContent(this.metadataMode);
 
-    const prompt = defaultQuestionAnswerPromptTemplate({
-      contextStr,
-      numQuestions: this.questions,
+    const prompt = this.promptTemplate.format({
+      context: contextStr,
+      numQuestions: this.questions.toString(),
     });
 
     const questions = await this.llm.complete({
@@ -356,7 +386,7 @@ export class QuestionsAnsweredExtractor extends BaseExtractor {
 type SummaryExtractArgs = {
   llm?: LLM;
   summaries?: string[];
-  promptTemplate?: string;
+  promptTemplate?: SummaryPrompt["template"];
 };
 
 type ExtractSummary = {
@@ -385,7 +415,7 @@ export class SummaryExtractor extends BaseExtractor {
    * The prompt template to use for the summary extractor.
    * @type {string}
    */
-  promptTemplate: string;
+  promptTemplate: SummaryPrompt;
 
   private selfSummary: boolean;
   private prevSummary: boolean;
@@ -404,8 +434,12 @@ export class SummaryExtractor extends BaseExtractor {
 
     this.llm = options?.llm ?? new OpenAI();
     this.summaries = summaries;
-    this.promptTemplate =
-      options?.promptTemplate ?? defaultSummaryExtractorPromptTemplate();
+    this.promptTemplate = options?.promptTemplate
+      ? new PromptTemplate({
+          templateVars: ["context"],
+          template: options.promptTemplate,
+        })
+      : defaultSummaryPrompt;
 
     this.selfSummary = summaries?.includes("self") ?? false;
     this.prevSummary = summaries?.includes("prev") ?? false;
@@ -422,10 +456,10 @@ export class SummaryExtractor extends BaseExtractor {
       return "";
     }
 
-    const contextStr = node.getContent(this.metadataMode);
+    const context = node.getContent(this.metadataMode);
 
-    const prompt = defaultSummaryExtractorPromptTemplate({
-      contextStr,
+    const prompt = this.promptTemplate.format({
+      context,
     });
 
     const summary = await this.llm.complete({

packages/llamaindex/src/extractors/prompts.ts

@@ -1,74 +0,0 @@
-export interface DefaultPromptTemplate {
-  contextStr: string;
-}
-
-export interface DefaultKeywordExtractorPromptTemplate
-  extends DefaultPromptTemplate {
-  keywords: number;
-}
-
-export interface DefaultQuestionAnswerPromptTemplate
-  extends DefaultPromptTemplate {
-  numQuestions: number;
-}
-
-export interface DefaultNodeTextTemplate {
-  metadataStr: string;
-  content: string;
-}
-
-export const defaultKeywordExtractorPromptTemplate = ({
-  contextStr = "",
-  keywords = 5,
-}: DefaultKeywordExtractorPromptTemplate) => `${contextStr} 
-Give ${keywords} unique keywords for this document. 
-Format as comma separated. 
-Keywords: `;
-
-export const defaultTitleExtractorPromptTemplate = (
-  { contextStr = "" }: DefaultPromptTemplate = {
-    contextStr: "",
-  },
-) => `${contextStr}
-Give a title that summarizes all of the unique entities, titles or themes found in the context. 
-Title: `;
-
-export const defaultTitleCombinePromptTemplate = (
-  { contextStr = "" }: DefaultPromptTemplate = {
-    contextStr: "",
-  },
-) => `${contextStr} 
-Based on the above candidate titles and contents, what is the comprehensive title for this document? 
-Title: `;
-
-export const defaultQuestionAnswerPromptTemplate = (
-  { contextStr = "", numQuestions = 5 }: DefaultQuestionAnswerPromptTemplate = {
-    contextStr: "",
-    numQuestions: 5,
-  },
-) => `${contextStr}
-Given the contextual informations, generate ${numQuestions} questions this context can provides specific answers to which are unlikely to be found else where. Higher-level summaries of surrounding context may be provideds as well. 
-Try using these summaries to generate better questions that this context can answer.
-`;
-
-export const defaultSummaryExtractorPromptTemplate = (
-  { contextStr = "" }: DefaultPromptTemplate = {
-    contextStr: "",
-  },
-) => `${contextStr}
-Summarize the key topics and entities of the sections.
-Summary: `;
-
-export const defaultNodeTextTemplate = ({
-  metadataStr = "",
-  content = "",
-}: {
-  metadataStr?: string;
-  content?: string;
-} = {}) => `[Excerpt from document]
-${metadataStr}
-Excerpt:
------
-${content}
------
-`;

packages/llamaindex/src/extractors/types.ts

@@ -1,10 +1,10 @@
+import { defaultNodeTextTemplate } from "@llamaindex/core/prompts";
 import {
   BaseNode,
   MetadataMode,
   TextNode,
   TransformComponent,
 } from "@llamaindex/core/schema";
-import { defaultNodeTextTemplate } from "./prompts.js";
 
 /*
  * Abstract class for all extractors.
@@ -71,7 +71,7 @@ export abstract class BaseExtractor extends TransformComponent {
         if (newNodes[idx] instanceof TextNode) {
           newNodes[idx] = new TextNode({
             ...newNodes[idx],
-            textTemplate: defaultNodeTextTemplate(),
+            textTemplate: defaultNodeTextTemplate.format(),
           });
         }
       }

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

@@ -31,8 +31,8 @@ import type { StorageContext } from "../../storage/StorageContext.js";
 import { storageContextFromDefaults } from "../../storage/StorageContext.js";
 import type { BaseIndexStore } from "../../storage/indexStore/types.js";
 import type {
+  BaseVectorStore,
   MetadataFilters,
-  VectorStore,
   VectorStoreByType,
   VectorStoreQueryResult,
 } from "../../vector-store/index.js";
@@ -264,7 +264,7 @@ export class VectorStoreIndex extends BaseIndex<IndexDict> {
   }
 
   static async fromVectorStore(
-    vectorStore: VectorStore,
+    vectorStore: BaseVectorStore,
     serviceContext?: ServiceContext,
   ) {
     return this.fromVectorStores(
@@ -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,
     );
   }
@@ -308,7 +307,7 @@ export class VectorStoreIndex extends BaseIndex<IndexDict> {
   protected async insertNodesToStore(
     newIds: string[],
     nodes: BaseNode[],
-    vectorStore: VectorStore,
+    vectorStore: BaseVectorStore,
   ): Promise<void> {
     // NOTE: if the vector store doesn't store text,
     // we need to add the nodes to the index struct and document store
@@ -358,7 +357,7 @@ export class VectorStoreIndex extends BaseIndex<IndexDict> {
   }
 
   protected async deleteRefDocFromStore(
-    vectorStore: VectorStore,
+    vectorStore: BaseVectorStore,
     refDocId: string,
   ): Promise<void> {
     await vectorStore.delete(refDocId);
@@ -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 {
@@ -426,7 +425,7 @@ export class VectorIndexRetriever extends BaseRetriever {
     let nodesWithScores: NodeWithScore[] = [];
 
     for (const type in vectorStores) {
-      const vectorStore: VectorStore = vectorStores[type as ModalityType]!;
+      const vectorStore: BaseVectorStore = vectorStores[type as ModalityType]!;
       nodesWithScores = nodesWithScores.concat(
         await this.retrieveQuery(query, type as ModalityType, vectorStore),
       );
@@ -437,7 +436,7 @@ export class VectorIndexRetriever extends BaseRetriever {
   protected async retrieveQuery(
     query: MessageContent,
     type: ModalityType,
-    vectorStore: VectorStore,
+    vectorStore: BaseVectorStore,
     filters?: MetadataFilters,
   ): Promise<NodeWithScore[]> {
     // convert string message to multi-modal format

packages/llamaindex/src/ingestion/IngestionPipeline.ts

@@ -7,7 +7,10 @@ import {
   type Metadata,
 } from "@llamaindex/core/schema";
 import type { BaseDocumentStore } from "../storage/docStore/types.js";
-import type { VectorStore, VectorStoreByType } from "../vector-store/types.js";
+import type {
+  BaseVectorStore,
+  VectorStoreByType,
+} from "../vector-store/types.js";
 import { IngestionCache, getTransformationHash } from "./IngestionCache.js";
 import {
   DocStoreStrategy,
@@ -59,7 +62,7 @@ export class IngestionPipeline {
   transformations: TransformComponent[] = [];
   documents?: Document[] | undefined;
   reader?: BaseReader | undefined;
-  vectorStore?: VectorStore | undefined;
+  vectorStore?: BaseVectorStore | undefined;
   vectorStores?: VectorStoreByType | undefined;
   docStore?: BaseDocumentStore;
   docStoreStrategy: DocStoreStrategy = DocStoreStrategy.UPSERTS;
@@ -133,7 +136,7 @@ export async function addNodesToVectorStores(
   nodesAdded?: (
     newIds: string[],
     nodes: BaseNode<Metadata>[],
-    vectorStore: VectorStore,
+    vectorStore: BaseVectorStore,
   ) => Promise<void>,
 ) {
   const nodeMap = splitNodesByType(nodes);

packages/llamaindex/src/ingestion/strategies/UpsertsAndDeleteStrategy.ts

@@ -1,6 +1,6 @@
 import { BaseNode, TransformComponent } from "@llamaindex/core/schema";
 import type { BaseDocumentStore } from "../../storage/docStore/types.js";
-import type { VectorStore } from "../../vector-store/types.js";
+import type { BaseVectorStore } from "../../vector-store/types.js";
 import { classify } from "./classify.js";
 
 /**
@@ -9,9 +9,9 @@ import { classify } from "./classify.js";
  */
 export class UpsertsAndDeleteStrategy extends TransformComponent {
   protected docStore: BaseDocumentStore;
-  protected vectorStores: VectorStore[] | undefined;
+  protected vectorStores: BaseVectorStore[] | undefined;
 
-  constructor(docStore: BaseDocumentStore, vectorStores?: VectorStore[]) {
+  constructor(docStore: BaseDocumentStore, vectorStores?: BaseVectorStore[]) {
     super(async (nodes: BaseNode[]): Promise<BaseNode[]> => {
       const { dedupedNodes, missingDocs, unusedDocs } = await classify(
         this.docStore,

packages/llamaindex/src/ingestion/strategies/UpsertsStrategy.ts

@@ -1,6 +1,6 @@
 import { BaseNode, TransformComponent } from "@llamaindex/core/schema";
 import type { BaseDocumentStore } from "../../storage/docStore/types.js";
-import type { VectorStore } from "../../vector-store/types.js";
+import type { BaseVectorStore } from "../../vector-store/types.js";
 import { classify } from "./classify.js";
 
 /**
@@ -8,9 +8,9 @@ import { classify } from "./classify.js";
  */
 export class UpsertsStrategy extends TransformComponent {
   protected docStore: BaseDocumentStore;
-  protected vectorStores: VectorStore[] | undefined;
+  protected vectorStores: BaseVectorStore[] | undefined;
 
-  constructor(docStore: BaseDocumentStore, vectorStores?: VectorStore[]) {
+  constructor(docStore: BaseDocumentStore, vectorStores?: BaseVectorStore[]) {
     super(async (nodes: BaseNode[]): Promise<BaseNode[]> => {
       const { dedupedNodes, unusedDocs } = await classify(this.docStore, nodes);
       // remove unused docs

packages/llamaindex/src/ingestion/strategies/index.ts

@@ -1,6 +1,6 @@
 import { TransformComponent } from "@llamaindex/core/schema";
 import type { BaseDocumentStore } from "../../storage/docStore/types.js";
-import type { VectorStore } from "../../vector-store/types.js";
+import type { BaseVectorStore } from "../../vector-store/types.js";
 import { DuplicatesStrategy } from "./DuplicatesStrategy.js";
 import { UpsertsAndDeleteStrategy } from "./UpsertsAndDeleteStrategy.js";
 import { UpsertsStrategy } from "./UpsertsStrategy.js";
@@ -28,7 +28,7 @@ class NoOpStrategy extends TransformComponent {
 export function createDocStoreStrategy(
   docStoreStrategy: DocStoreStrategy,
   docStore?: BaseDocumentStore,
-  vectorStores: VectorStore[] = [],
+  vectorStores: BaseVectorStore[] = [],
 ): TransformComponent {
   if (docStoreStrategy === DocStoreStrategy.NONE) {
     return new NoOpStrategy();

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/gemini/base.ts

@@ -44,12 +44,15 @@ import {
 export const GEMINI_MODEL_INFO_MAP: Record<GEMINI_MODEL, GeminiModelInfo> = {
   [GEMINI_MODEL.GEMINI_PRO]: { contextWindow: 30720 },
   [GEMINI_MODEL.GEMINI_PRO_VISION]: { contextWindow: 12288 },
-  [GEMINI_MODEL.GEMINI_PRO_LATEST]: { contextWindow: 10 ** 6 },
   // multi-modal/multi turn
+  [GEMINI_MODEL.GEMINI_PRO_LATEST]: { contextWindow: 10 ** 6 },
+  [GEMINI_MODEL.GEMINI_PRO_FLASH_LATEST]: { contextWindow: 10 ** 6 },
   [GEMINI_MODEL.GEMINI_PRO_1_5_PRO_PREVIEW]: { contextWindow: 10 ** 6 },
   [GEMINI_MODEL.GEMINI_PRO_1_5_FLASH_PREVIEW]: { contextWindow: 10 ** 6 },
   [GEMINI_MODEL.GEMINI_PRO_1_5]: { contextWindow: 2 * 10 ** 6 },
   [GEMINI_MODEL.GEMINI_PRO_1_5_FLASH]: { contextWindow: 10 ** 6 },
+  [GEMINI_MODEL.GEMINI_PRO_1_5_LATEST]: { contextWindow: 2 * 10 ** 6 },
+  [GEMINI_MODEL.GEMINI_PRO_1_5_FLASH_LATEST]: { contextWindow: 10 ** 6 },
 };
 
 const SUPPORT_TOOL_CALL_MODELS: GEMINI_MODEL[] = [
@@ -59,6 +62,10 @@ const SUPPORT_TOOL_CALL_MODELS: GEMINI_MODEL[] = [
   GEMINI_MODEL.GEMINI_PRO_1_5_FLASH_PREVIEW,
   GEMINI_MODEL.GEMINI_PRO_1_5,
   GEMINI_MODEL.GEMINI_PRO_1_5_FLASH,
+  GEMINI_MODEL.GEMINI_PRO_LATEST,
+  GEMINI_MODEL.GEMINI_PRO_FLASH_LATEST,
+  GEMINI_MODEL.GEMINI_PRO_1_5_LATEST,
+  GEMINI_MODEL.GEMINI_PRO_1_5_FLASH_LATEST,
 ];
 
 const DEFAULT_GEMINI_PARAMS = {

packages/llamaindex/src/llm/gemini/types.ts

@@ -56,10 +56,14 @@ export enum GEMINI_MODEL {
   GEMINI_PRO = "gemini-pro",
   GEMINI_PRO_VISION = "gemini-pro-vision",
   GEMINI_PRO_LATEST = "gemini-1.5-pro-latest",
+  GEMINI_PRO_FLASH_LATEST = "gemini-1.5-flash-latest",
   GEMINI_PRO_1_5_PRO_PREVIEW = "gemini-1.5-pro-preview-0514",
   GEMINI_PRO_1_5_FLASH_PREVIEW = "gemini-1.5-flash-preview-0514",
   GEMINI_PRO_1_5 = "gemini-1.5-pro-001",
   GEMINI_PRO_1_5_FLASH = "gemini-1.5-flash-001",
+  // Note: should be switched to -latest suffix when google supports it
+  GEMINI_PRO_1_5_LATEST = "gemini-1.5-pro-002",
+  GEMINI_PRO_1_5_FLASH_LATEST = "gemini-1.5-flash-002",
 }
 
 export interface GeminiModelInfo {

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/next.ts

@@ -37,12 +37,14 @@ export default function withLlamaIndex(config: any) {
     webpackConfig.resolve.alias = {
       ...webpackConfig.resolve.alias,
       "@google-cloud/vertexai": false,
+      unpdf: false,
     };
     // Following lines will fix issues with onnxruntime-node when using pnpm
     // See: https://github.com/vercel/next.js/issues/43433
     webpackConfig.externals.push({
       "onnxruntime-node": "commonjs onnxruntime-node",
       sharp: "commonjs sharp",
+      chromadb: "commonjs chromadb",
     });
     return webpackConfig;
   };

packages/llamaindex/src/storage/StorageContext.ts

@@ -7,7 +7,10 @@ import { path } from "@llamaindex/env";
 import { getImageEmbedModel } from "../internal/settings/image-embed-model.js";
 import type { ServiceContext } from "../ServiceContext.js";
 import { SimpleVectorStore } from "../vector-store/SimpleVectorStore.js";
-import type { VectorStore, VectorStoreByType } from "../vector-store/types.js";
+import type {
+  BaseVectorStore,
+  VectorStoreByType,
+} from "../vector-store/types.js";
 import { SimpleDocumentStore } from "./docStore/SimpleDocumentStore.js";
 import type { BaseDocumentStore } from "./docStore/types.js";
 import { SimpleIndexStore } from "./indexStore/SimpleIndexStore.js";
@@ -22,7 +25,7 @@ export interface StorageContext {
 type BuilderParams = {
   docStore: BaseDocumentStore;
   indexStore: BaseIndexStore;
-  vectorStore: VectorStore;
+  vectorStore: BaseVectorStore;
   vectorStores: VectorStoreByType;
   storeImages: boolean;
   persistDir: string;
@@ -50,7 +53,7 @@ export async function storageContextFromDefaults({
     }
     if (storeImages && !(ModalityType.IMAGE in vectorStores)) {
       vectorStores[ModalityType.IMAGE] = new SimpleVectorStore({
-        embedModel: new (await getImageEmbedModel())(),
+        embeddingModel: new (await getImageEmbedModel())(),
       });
     }
   } else {

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

@@ -2,24 +2,30 @@ import {
   Collection,
   DataAPIClient,
   Db,
+  type Filter,
   type FindOptions,
+  type SomeDoc,
 } from "@datastax/astra-db-ts";
 import type { BaseNode } from "@llamaindex/core/schema";
 import { MetadataMode } from "@llamaindex/core/schema";
 import { getEnv } from "@llamaindex/env";
 import {
-  VectorStoreBase,
-  type IEmbedModel,
-  type VectorStoreNoEmbedModel,
+  BaseVectorStore,
+  FilterCondition,
+  FilterOperator,
+  type MetadataFilter,
+  type MetadataFilters,
+  type VectorStoreBaseParams,
   type VectorStoreQuery,
   type VectorStoreQueryResult,
 } from "./types.js";
-import { metadataDictToNode, nodeToMetadata } from "./utils.js";
+import {
+  metadataDictToNode,
+  nodeToMetadata,
+  parseArrayValue,
+} from "./utils.js";
 
-export class AstraDBVectorStore
-  extends VectorStoreBase
-  implements VectorStoreNoEmbedModel
-{
+export class AstraDBVectorStore extends BaseVectorStore {
   storesText: boolean = true;
   flatMetadata: boolean = true;
 
@@ -37,9 +43,9 @@ export class AstraDBVectorStore
         endpoint: string;
         namespace?: string;
       };
-    } & Partial<IEmbedModel>,
+    } & VectorStoreBaseParams,
   ) {
-    super(init?.embedModel);
+    super(init);
     const token = init?.params?.token ?? getEnv("ASTRA_DB_APPLICATION_TOKEN");
     const endpoint = init?.params?.endpoint ?? getEnv("ASTRA_DB_API_ENDPOINT");
 
@@ -183,12 +189,8 @@ export class AstraDBVectorStore
     }
     const collection = this.collection;
 
-    const filters: Record<string, any> = {};
-    query.filters?.filters?.forEach((f) => {
-      filters[f.key] = f.value;
-    });
-
-    const cursor = await collection.find(filters, <FindOptions>{
+    const astraFilter = this.toAstraFilter(query.filters);
+    const cursor = await collection.find(astraFilter, <FindOptions>{
       ...options,
       sort: query.queryEmbedding
         ? { $vector: query.queryEmbedding }
@@ -230,4 +232,39 @@ export class AstraDBVectorStore
       nodes,
     };
   }
+
+  private toAstraFilter(filters?: MetadataFilters): Filter<SomeDoc> {
+    if (!filters || filters.filters?.length === 0) return {};
+    const condition = filters.condition ?? FilterCondition.AND;
+    const listFilter = filters.filters.map((f) => this.buildFilterItem(f));
+    if (condition === FilterCondition.OR) return { $or: listFilter };
+    if (condition === FilterCondition.AND) return { $and: listFilter };
+    throw new Error(`Not supported filter condition: ${condition}`);
+  }
+
+  private buildFilterItem(filter: MetadataFilter): Filter<SomeDoc> {
+    const { key, operator, value } = filter;
+    switch (operator) {
+      case FilterOperator.EQ:
+        return { [key]: value };
+      case FilterOperator.NE:
+        return { [key]: { $ne: value } };
+      case FilterOperator.GT:
+        return { [key]: { $gt: value } };
+      case FilterOperator.LT:
+        return { [key]: { $lt: value } };
+      case FilterOperator.GTE:
+        return { [key]: { $gte: value } };
+      case FilterOperator.LTE:
+        return { [key]: { $lte: value } };
+      case FilterOperator.IN:
+        return { [key]: { $in: parseArrayValue(value) } };
+      case FilterOperator.NIN:
+        return { [key]: { $nin: parseArrayValue(value) } };
+      case FilterOperator.IS_EMPTY:
+        return { [key]: { $size: 0 } };
+      default:
+        throw new Error(`Not supported filter operator: ${operator}`);
+    }
+  }
 }

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

@@ -2,21 +2,20 @@ import type { BaseNode } from "@llamaindex/core/schema";
 import { MetadataMode } from "@llamaindex/core/schema";
 import {
   ChromaClient,
-  IncludeEnum,
-  type AddParams,
   type ChromaClientParams,
-  type Collection,
   type DeleteParams,
-  type QueryParams,
+  type QueryRecordsParams,
   type QueryResponse,
   type Where,
   type WhereDocument,
 } from "chromadb";
 import {
-  VectorStoreBase,
+  BaseVectorStore,
+  FilterCondition,
+  FilterOperator,
   VectorStoreQueryMode,
-  type IEmbedModel,
-  type VectorStoreNoEmbedModel,
+  type MetadataFilters,
+  type VectorStoreBaseParams,
   type VectorStoreQuery,
   type VectorStoreQueryResult,
 } from "./types.js";
@@ -31,12 +30,22 @@ type ChromaQueryOptions = {
   whereDocument?: WhereDocument;
 };
 
+type Collection = Awaited<ReturnType<ChromaClient["getOrCreateCollection"]>>;
+
 const DEFAULT_TEXT_KEY = "text";
 
-export class ChromaVectorStore
-  extends VectorStoreBase
-  implements VectorStoreNoEmbedModel
-{
+type ChromaFilterCondition = "$and" | "$or";
+type ChromaFilterOperator =
+  | "$eq"
+  | "$ne"
+  | "$gt"
+  | "$lt"
+  | "$gte"
+  | "$lte"
+  | "$in"
+  | "$nin";
+
+export class ChromaVectorStore extends BaseVectorStore {
   storesText: boolean = true;
   flatMetadata: boolean = true;
   textKey: string;
@@ -49,9 +58,9 @@ export class ChromaVectorStore
       collectionName: string;
       textKey?: string;
       chromaClientParams?: ChromaClientParams;
-    } & Partial<IEmbedModel>,
+    } & VectorStoreBaseParams,
   ) {
-    super(init.embedModel);
+    super(init);
     this.collectionName = init.collectionName;
     this.chromaClient = new ChromaClient(init.chromaClientParams);
     this.textKey = init.textKey ?? DEFAULT_TEXT_KEY;
@@ -71,7 +80,7 @@ export class ChromaVectorStore
     return this.collection;
   }
 
-  private getDataToInsert(nodes: BaseNode[]): AddParams {
+  private getDataToInsert(nodes: BaseNode[]) {
     const metadatas = nodes.map((node) =>
       nodeToMetadata(node, true, this.textKey, this.flatMetadata),
     );
@@ -106,6 +115,79 @@ export class ChromaVectorStore
     });
   }
 
+  private transformChromaFilterCondition(
+    condition: FilterCondition,
+  ): ChromaFilterCondition {
+    switch (condition) {
+      case FilterCondition.AND:
+        return "$and";
+      case FilterCondition.OR:
+        return "$or";
+      default:
+        throw new Error(`Filter condition ${condition} not supported`);
+    }
+  }
+
+  private transformChromaFilterOperator(
+    operator: FilterOperator,
+  ): ChromaFilterOperator {
+    switch (operator) {
+      case FilterOperator.EQ:
+        return "$eq";
+      case FilterOperator.NE:
+        return "$ne";
+      case FilterOperator.GT:
+        return "$gt";
+      case FilterOperator.LT:
+        return "$lt";
+      case FilterOperator.GTE:
+        return "$gte";
+      case FilterOperator.LTE:
+        return "$lte";
+      case FilterOperator.IN:
+        return "$in";
+      case FilterOperator.NIN:
+        return "$nin";
+      default:
+        throw new Error(`Filter operator ${operator} not supported`);
+    }
+  }
+
+  private toChromaFilter(filters: MetadataFilters): Where {
+    const chromaFilter: Where = {};
+    const filtersList: Where[] = [];
+
+    const condition = filters.condition
+      ? this.transformChromaFilterCondition(
+          filters.condition as FilterCondition,
+        )
+      : "$and";
+
+    if (filters.filters) {
+      for (const filter of filters.filters) {
+        if (filter.operator) {
+          filtersList.push({
+            [filter.key]: {
+              [this.transformChromaFilterOperator(
+                filter.operator as FilterOperator,
+              )]: filter.value,
+            },
+          });
+        } else {
+          filtersList.push({ [filter.key]: filter.value });
+        }
+      }
+
+      if (filtersList.length === 1) {
+        return filtersList[0]!;
+      } else if (filtersList.length > 1) {
+        chromaFilter[condition] = filtersList;
+      }
+    }
+
+    return chromaFilter;
+  }
+
   async query(
     query: VectorStoreQuery,
     options?: ChromaQueryOptions,
@@ -117,49 +199,22 @@ export class ChromaVectorStore
       throw new Error("ChromaDB does not support querying by mode");
     }
 
-    // fixme: type is broken
-    let chromaWhere: any = {};
-    if (query.filters?.filters) {
-      query.filters.filters.map((filter) => {
-        const filterKey = filter.key;
-        const filterValue = filter.value;
-        if (filterKey in chromaWhere || "$or" in chromaWhere) {
-          if (!chromaWhere["$or"]) {
-            chromaWhere = {
-              $or: [
-                {
-                  ...chromaWhere,
-                },
-                {
-                  [filterKey]: filterValue,
-                },
-              ],
-            };
-          } else {
-            chromaWhere["$or"].push({
-              [filterKey]: filterValue,
-            });
-          }
-        } else {
-          chromaWhere[filterKey] = filterValue;
-        }
-      });
+    let chromaWhere: Where = {};
+    if (query.filters) {
+      chromaWhere = this.toChromaFilter(query.filters);
     }
 
     const collection = await this.getCollection();
-    const queryResponse: QueryResponse = await collection.query(<QueryParams>{
+    const queryResponse: QueryResponse = await collection.query(<
+      QueryRecordsParams
+    >{
       queryEmbeddings: query.queryEmbedding ?? undefined,
       queryTexts: query.queryStr ?? undefined,
       nResults: query.similarityTopK,
       where: Object.keys(chromaWhere).length ? chromaWhere : undefined,
       whereDocument: options?.whereDocument,
       //ChromaDB doesn't return the result embeddings by default so we need to include them
-      include: [
-        IncludeEnum.Distances,
-        IncludeEnum.Metadatas,
-        IncludeEnum.Documents,
-        IncludeEnum.Embeddings,
-      ],
+      include: ["distances", "metadatas", "documents", "embeddings"],
     });
 
     const vectorStoreQueryResult: VectorStoreQueryResult = {

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

@@ -10,10 +10,9 @@ import {
   type SearchSimpleReq,
 } from "@zilliz/milvus2-sdk-node";
 import {
-  VectorStoreBase,
-  type IEmbedModel,
+  BaseVectorStore,
   type MetadataFilters,
-  type VectorStoreNoEmbedModel,
+  type VectorStoreBaseParams,
   type VectorStoreQuery,
   type VectorStoreQueryResult,
 } from "./types.js";
@@ -72,12 +71,9 @@ function parseScalarFilters(scalarFilters: MetadataFilters): string {
   return filters.join(` ${condition} `);
 }
 
-export class MilvusVectorStore
-  extends VectorStoreBase
-  implements VectorStoreNoEmbedModel
-{
+export class MilvusVectorStore extends BaseVectorStore {
   public storesText: boolean = true;
-  public isEmbeddingQuery?: boolean;
+  public isEmbeddingQuery?: boolean = false;
   private flatMetadata: boolean = true;
 
   private milvusClient: MilvusClient;
@@ -91,7 +87,7 @@ export class MilvusVectorStore
 
   constructor(
     init?: Partial<{ milvusClient: MilvusClient }> &
-      Partial<IEmbedModel> & {
+      VectorStoreBaseParams & {
         params?: {
           configOrAddress: ClientConfig | string;
           ssl?: boolean;
@@ -106,7 +102,7 @@ export class MilvusVectorStore
         embeddingKey?: string;
       },
   ) {
-    super(init?.embedModel);
+    super(init);
     if (init?.milvusClient) {
       this.milvusClient = init.milvusClient;
     } else {

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

@@ -5,12 +5,12 @@ import { getEnv } from "@llamaindex/env";
 import type { BulkWriteOptions, Collection } from "mongodb";
 import { MongoClient } from "mongodb";
 import {
+  BaseVectorStore,
   FilterCondition,
-  VectorStoreBase,
   type FilterOperator,
   type MetadataFilter,
   type MetadataFilters,
-  type VectorStoreNoEmbedModel,
+  type VectorStoreBaseParams,
   type VectorStoreQuery,
   type VectorStoreQueryResult,
 } from "./types.js";
@@ -67,10 +67,7 @@ function toMongoDBFilter(filters?: MetadataFilters): Record<string, any> {
  * Vector store that uses MongoDB Atlas for storage and vector search.
  * This store uses the $vectorSearch aggregation stage to perform vector similarity search.
  */
-export class MongoDBAtlasVectorSearch
-  extends VectorStoreBase
-  implements VectorStoreNoEmbedModel
-{
+export class MongoDBAtlasVectorSearch extends BaseVectorStore {
   storesText: boolean = true;
   flatMetadata: boolean = true;
 
@@ -147,9 +144,9 @@ export class MongoDBAtlasVectorSearch
       autoCreateIndex?: boolean;
       indexedMetadataFields?: string[];
       embeddingDefinition?: Record<string, unknown>;
-    },
+    } & VectorStoreBaseParams,
   ) {
-    super(init.embedModel);
+    super(init);
     if (init.mongodbClient) {
       this.mongodbClient = init.mongodbClient;
     } else {
@@ -274,7 +271,7 @@ export class MongoDBAtlasVectorSearch
     );
   }
 
-  get client(): any {
+  client(): any {
     return this.mongodbClient;
   }
 

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

@@ -4,12 +4,12 @@ import type { IsomorphicDB } from "@llamaindex/core/vector-store";
 import type { VercelPool } from "@vercel/postgres";
 import type { Sql } from "postgres";
 import {
+  BaseVectorStore,
   FilterCondition,
   FilterOperator,
   type MetadataFilter,
   type MetadataFilterValue,
-  VectorStoreBase,
-  type VectorStoreNoEmbedModel,
+  type VectorStoreBaseParams,
   type VectorStoreQuery,
   type VectorStoreQueryResult,
 } from "./types.js";
@@ -116,7 +116,8 @@ type PGVectorStoreBaseConfig = {
   embedModel?: BaseEmbedding | undefined;
 };
 
-export type PGVectorStoreConfig = PGVectorStoreBaseConfig &
+export type PGVectorStoreConfig = VectorStoreBaseParams &
+  PGVectorStoreBaseConfig &
   (
     | {
         /**
@@ -147,10 +148,7 @@ export type PGVectorStoreConfig = PGVectorStoreBaseConfig &
  * Provides support for writing and querying vector data in Postgres.
  * Note: Can't be used with data created using the Python version of the vector store (https://docs.llamaindex.ai/en/stable/examples/vector_stores/postgres/)
  */
-export class PGVectorStore
-  extends VectorStoreBase
-  implements VectorStoreNoEmbedModel
-{
+export class PGVectorStore extends BaseVectorStore {
   storesText: boolean = true;
 
   private collection: string = DEFAULT_COLLECTION;
@@ -163,7 +161,7 @@ export class PGVectorStore
   private readonly clientConfig: pg.ClientConfig | null = null;
 
   constructor(config: PGVectorStoreConfig) {
-    super(config?.embedModel);
+    super(config);
     this.schemaName = config?.schemaName ?? PGVECTOR_SCHEMA;
     this.tableName = config?.tableName ?? PGVECTOR_TABLE;
     this.dimensions = config?.dimensions ?? DEFAULT_DIMENSIONS;

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

@@ -1,11 +1,10 @@
 import {
+  BaseVectorStore,
   FilterCondition,
   FilterOperator,
-  VectorStoreBase,
-  type IEmbedModel,
   type MetadataFilter,
   type MetadataFilters,
-  type VectorStoreNoEmbedModel,
+  type VectorStoreBaseParams,
   type VectorStoreQuery,
   type VectorStoreQueryResult,
 } from "./types.js";
@@ -25,15 +24,12 @@ type PineconeParams = {
   chunkSize?: number;
   namespace?: string;
   textKey?: string;
-} & IEmbedModel;
+} & VectorStoreBaseParams;
 
 /**
  * Provides support for writing and querying vector data in Pinecone.
  */
-export class PineconeVectorStore
-  extends VectorStoreBase
-  implements VectorStoreNoEmbedModel
-{
+export class PineconeVectorStore extends BaseVectorStore {
   storesText: boolean = true;
 
   /*
@@ -51,7 +47,7 @@ export class PineconeVectorStore
   textKey: string;
 
   constructor(params?: PineconeParams) {
-    super(params?.embedModel);
+    super(params);
     this.indexName =
       params?.indexName ?? getEnv("PINECONE_INDEX_NAME") ?? "llama";
     this.namespace = params?.namespace ?? getEnv("PINECONE_NAMESPACE") ?? "";

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

@@ -1,16 +1,21 @@
 import type { BaseNode } from "@llamaindex/core/schema";
 import {
-  VectorStoreBase,
-  type IEmbedModel,
-  type VectorStoreNoEmbedModel,
+  BaseVectorStore,
+  FilterCondition,
+  FilterOperator,
+  type MetadataFilters,
+  type VectorStoreBaseParams,
   type VectorStoreQuery,
   type VectorStoreQueryResult,
 } from "./types.js";
 
-import type { QdrantClientParams } from "@qdrant/js-client-rest";
+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>;
@@ -23,7 +28,7 @@ type QdrantParams = {
   url?: string;
   apiKey?: string;
   batchSize?: number;
-} & Partial<IEmbedModel>;
+} & VectorStoreBaseParams;
 
 type QuerySearchResult = {
   id: string;
@@ -36,10 +41,7 @@ type QuerySearchResult = {
 /**
  * Qdrant vector store.
  */
-export class QdrantVectorStore
-  extends VectorStoreBase
-  implements VectorStoreNoEmbedModel
-{
+export class QdrantVectorStore extends BaseVectorStore {
   storesText: boolean = true;
 
   batchSize: number;
@@ -63,9 +65,9 @@ export class QdrantVectorStore
     url,
     apiKey,
     batchSize,
-    embedModel,
+    ...init
   }: QdrantParams) {
-    super(embedModel);
+    super(init);
     if (!client && !url) {
       if (!url) {
         throw new Error("QdrantVectorStore requires url and collectionName");
@@ -272,7 +274,7 @@ 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 +283,7 @@ 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 +294,118 @@ 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/src/vector-store/SimpleVectorStore.ts

@@ -5,12 +5,12 @@ import { fs, path } from "@llamaindex/env";
 import { getTopKEmbeddings, getTopKMMREmbeddings } from "../internal/utils.js";
 import { exists } from "../storage/FileSystem.js";
 import {
+  BaseVectorStore,
   FilterOperator,
-  VectorStoreBase,
   VectorStoreQueryMode,
   type MetadataFilter,
   type MetadataFilters,
-  type VectorStoreNoEmbedModel,
+  type VectorStoreBaseParams,
   type VectorStoreQuery,
   type VectorStoreQueryResult,
 } from "./types.js";
@@ -120,19 +120,17 @@ class SimpleVectorStoreData {
   metadataDict: Record<string, MetadataValue> = {};
 }
 
-export class SimpleVectorStore
-  extends VectorStoreBase
-  implements VectorStoreNoEmbedModel
-{
+export class SimpleVectorStore extends BaseVectorStore {
   storesText: boolean = false;
   private data: SimpleVectorStoreData;
   private persistPath: string | undefined;
 
-  constructor(init?: {
-    data?: SimpleVectorStoreData | undefined;
-    embedModel?: BaseEmbedding | undefined;
-  }) {
-    super(init?.embedModel);
+  constructor(
+    init?: {
+      data?: SimpleVectorStoreData | undefined;
+    } & VectorStoreBaseParams,
+  ) {
+    super(init);
     this.data = init?.data || new SimpleVectorStoreData();
   }
 
@@ -144,7 +142,7 @@ export class SimpleVectorStore
     return await SimpleVectorStore.fromPersistPath(persistPath, embedModel);
   }
 
-  get client(): any {
+  client(): any {
     return null;
   }
 
@@ -272,7 +270,7 @@ export class SimpleVectorStore
 
   static async fromPersistPath(
     persistPath: string,
-    embedModel?: BaseEmbedding,
+    embeddingModel?: BaseEmbedding,
   ): Promise<SimpleVectorStore> {
     const dirPath = path.dirname(persistPath);
     if (!(await exists(dirPath))) {
@@ -298,20 +296,20 @@ export class SimpleVectorStore
     data.embeddingDict = dataDict.embeddingDict ?? {};
     data.textIdToRefDocId = dataDict.textIdToRefDocId ?? {};
     data.metadataDict = dataDict.metadataDict ?? {};
-    const store = new SimpleVectorStore({ data, embedModel });
+    const store = new SimpleVectorStore({ data, embeddingModel });
     store.persistPath = persistPath;
     return store;
   }
 
   static fromDict(
     saveDict: SimpleVectorStoreData,
-    embedModel?: BaseEmbedding,
+    embeddingModel?: BaseEmbedding,
   ): SimpleVectorStore {
     const data = new SimpleVectorStoreData();
     data.embeddingDict = saveDict.embeddingDict;
     data.textIdToRefDocId = saveDict.textIdToRefDocId;
     data.metadataDict = saveDict.metadataDict;
-    return new SimpleVectorStore({ data, embedModel });
+    return new SimpleVectorStore({ data, embeddingModel });
   }
 
   toDict(): SimpleVectorStoreData {

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

@@ -1,9 +1,8 @@
 import {
-  VectorStoreBase,
-  type IEmbedModel,
+  BaseVectorStore,
   type MetadataFilter,
   type MetadataFilters,
-  type VectorStoreNoEmbedModel,
+  type VectorStoreBaseParams,
   type VectorStoreQuery,
   type VectorStoreQueryResult,
 } from "./types.js";
@@ -18,15 +17,12 @@ type UpstashParams = {
   token?: string;
   endpoint?: string;
   maxBatchSize?: number;
-} & IEmbedModel;
+} & VectorStoreBaseParams;
 
 /**
  * Provides support for writing and querying vector data in Upstash.
  */
-export class UpstashVectorStore
-  extends VectorStoreBase
-  implements VectorStoreNoEmbedModel
-{
+export class UpstashVectorStore extends BaseVectorStore {
   storesText: boolean = true;
 
   private db: Index;
@@ -45,7 +41,7 @@ export class UpstashVectorStore
    * ```
    */
   constructor(params?: UpstashParams) {
-    super(params?.embedModel);
+    super(params);
     this.namespace = params?.namespace ?? "";
     this.maxBatchSize = params?.maxBatchSize ?? 1000;
     const token = params?.token ?? getEnv("UPSTASH_VECTOR_REST_TOKEN");

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

@@ -11,12 +11,11 @@ import weaviate, {
 import { getEnv } from "@llamaindex/env";
 import type { BaseHybridOptions } from "weaviate-client";
 import {
-  VectorStoreBase,
+  BaseVectorStore,
   VectorStoreQueryMode,
-  type IEmbedModel,
   type MetadataFilter,
   type MetadataFilters,
-  type VectorStoreNoEmbedModel,
+  type VectorStoreBaseParams,
   type VectorStoreQuery,
   type VectorStoreQueryResult,
 } from "./types.js";
@@ -119,10 +118,7 @@ const toWeaviateFilter = (
   return Filters[condition](...filtersList);
 };
 
-export class WeaviateVectorStore
-  extends VectorStoreBase
-  implements VectorStoreNoEmbedModel
-{
+export class WeaviateVectorStore extends BaseVectorStore {
   public storesText: boolean = true;
   private flatMetadata: boolean = true;
 
@@ -137,7 +133,7 @@ export class WeaviateVectorStore
   private metadataKey: string;
 
   constructor(
-    init?: Partial<IEmbedModel> & {
+    init?: VectorStoreBaseParams & {
       weaviateClient?: WeaviateClient;
       cloudOptions?: {
         clusterURL?: string;
@@ -150,7 +146,7 @@ export class WeaviateVectorStore
       embeddingKey?: string;
     },
   ) {
-    super(init?.embedModel);
+    super(init);
 
     if (init?.weaviateClient) {
       // Use the provided client

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

@@ -76,34 +76,28 @@ export interface VectorStoreQuery {
   mmrThreshold?: number;
 }
 
-export interface VectorStoreNoEmbedModel {
-  storesText: boolean;
+// Supported types of vector stores (for each modality)
+export type VectorStoreByType = {
+  [P in ModalityType]?: BaseVectorStore;
+};
+
+export type VectorStoreBaseParams = {
+  embeddingModel?: BaseEmbedding | undefined;
+};
+
+export abstract class BaseVectorStore {
+  embedModel: BaseEmbedding;
+  abstract storesText: boolean;
   isEmbeddingQuery?: boolean;
-  client(): any;
-  add(embeddingResults: BaseNode[]): Promise<string[]>;
-  delete(refDocId: string, deleteOptions?: any): Promise<void>;
-  query(
+  abstract client(): any;
+  abstract add(embeddingResults: BaseNode[]): Promise<string[]>;
+  abstract delete(refDocId: string, deleteOptions?: any): Promise<void>;
+  abstract query(
     query: VectorStoreQuery,
     options?: any,
   ): Promise<VectorStoreQueryResult>;
-}
 
-export interface IEmbedModel {
-  embedModel: BaseEmbedding;
-}
-
-export interface VectorStore extends VectorStoreNoEmbedModel, IEmbedModel {}
-
-// Supported types of vector stores (for each modality)
-
-export type VectorStoreByType = {
-  [P in ModalityType]?: VectorStore;
-};
-
-export abstract class VectorStoreBase implements IEmbedModel {
-  embedModel: BaseEmbedding;
-
-  protected constructor(embedModel?: BaseEmbedding) {
-    this.embedModel = embedModel ?? getEmbeddedModel();
+  protected constructor(params?: VectorStoreBaseParams) {
+    this.embedModel = params?.embeddingModel ?? getEmbeddedModel();
   }
 }

packages/llamaindex/tests/MetadataExtractors.test.ts

@@ -102,6 +102,35 @@ describe("[MetadataExtractor]: Extractors should populate the metadata", () => {
     });
   });
 
+  test("[MetadataExtractor] QuestionsAnsweredExtractor uses custom prompt template", async () => {
+    const nodeParser = new SentenceSplitter();
+
+    const nodes = nodeParser.getNodesFromDocuments([
+      new Document({ text: DEFAULT_LLM_TEXT_OUTPUT }),
+    ]);
+
+    const llmCompleteSpy = vi.spyOn(serviceContext.llm, "complete");
+
+    const questionsAnsweredExtractor = new QuestionsAnsweredExtractor({
+      llm: serviceContext.llm,
+      questions: 5,
+      promptTemplate: `This is a custom prompt template for {context} with {numQuestions} questions`,
+    });
+
+    await questionsAnsweredExtractor.processNodes(nodes);
+
+    expect(llmCompleteSpy).toHaveBeenCalled();
+
+    // Build the expected prompt
+    const expectedPrompt = `This is a custom prompt template for ${DEFAULT_LLM_TEXT_OUTPUT} with 5 questions`;
+
+    // Get the actual prompt used in llm.complete
+    const actualPrompt = llmCompleteSpy.mock?.calls?.[0]?.[0];
+
+    // Assert that the prompts match
+    expect(actualPrompt).toEqual({ prompt: expectedPrompt });
+  });
+
   test("[MetadataExtractor] SumamryExtractor returns sectionSummary metadata", async () => {
     const nodeParser = new SentenceSplitter();
 
@@ -119,4 +148,33 @@ describe("[MetadataExtractor]: Extractors should populate the metadata", () => {
       sectionSummary: DEFAULT_LLM_TEXT_OUTPUT,
     });
   });
+
+  test("[KeywordExtractor] KeywordExtractor uses custom prompt template", async () => {
+    const nodeParser = new SentenceSplitter();
+
+    const nodes = nodeParser.getNodesFromDocuments([
+      new Document({ text: DEFAULT_LLM_TEXT_OUTPUT }),
+    ]);
+
+    const llmCompleteSpy = vi.spyOn(serviceContext.llm, "complete");
+
+    const keywordExtractor = new KeywordExtractor({
+      llm: serviceContext.llm,
+      keywords: 5,
+      promptTemplate: `This is a custom prompt template for {context} with {maxKeywords} keywords`,
+    });
+
+    await keywordExtractor.processNodes(nodes);
+
+    expect(llmCompleteSpy).toHaveBeenCalled();
+
+    // Build the expected prompt
+    const expectedPrompt = `This is a custom prompt template for ${DEFAULT_LLM_TEXT_OUTPUT} with 5 keywords`;
+
+    // Get the actual prompt used in llm.complete
+    const actualPrompt = llmCompleteSpy.mock?.calls?.[0]?.[0];
+
+    // Assert that the prompts match
+    expect(actualPrompt).toEqual({ prompt: expectedPrompt });
+  });
 });

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/llamaindex/tests/vector-stores/SimpleVectorStore.test.ts

@@ -59,7 +59,7 @@ describe("SimpleVectorStore", () => {
       }),
     ];
     store = new SimpleVectorStore({
-      embedModel: {} as BaseEmbedding, // Mocking the embedModel
+      embeddingModel: {} as BaseEmbedding, // Mocking the embedModel
       data: {
         embeddingDict: {},
         textIdToRefDocId: {},

packages/llm/groq/CHANGELOG.md

@@ -1,5 +1,50 @@
 # @llamaindex/groq
 
+## 0.0.16
+
+### Patch Changes
+
+- @llamaindex/openai@0.1.17
+
+## 0.0.15
+
+### Patch Changes
+
+- Updated dependencies [6a9a7b1]
+  - @llamaindex/openai@0.1.16
+
+## 0.0.14
+
+### Patch Changes
+
+- @llamaindex/openai@0.1.15
+
+## 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": "0.0.16",
   "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 +1,42 @@
 # @llamaindex/ollama
 
+## 0.0.8
+
+### Patch Changes
+
+- Updated dependencies [1364e8e]
+- Updated dependencies [96fc69c]
+  - @llamaindex/core@0.3.0
+
+## 0.0.7
+
+### Patch Changes
+
+- Updated dependencies [5f67820]
+  - @llamaindex/core@0.2.12
+
+## 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": "0.0.8",
   "type": "module",
   "main": "./dist/index.cjs",
   "module": "./dist/index.js",

packages/llm/openai/CHANGELOG.md

@@ -1,5 +1,54 @@
 # @llamaindex/openai
 
+## 0.1.17
+
+### Patch Changes
+
+- Updated dependencies [1364e8e]
+- Updated dependencies [96fc69c]
+  - @llamaindex/core@0.3.0
+
+## 0.1.16
+
+### Patch Changes
+
+- 6a9a7b1: fix: take init api key into account
+
+## 0.1.15
+
+### Patch Changes
+
+- Updated dependencies [5f67820]
+  - @llamaindex/core@0.2.12
+
+## 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": "0.1.17",
   "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 +100,33 @@ 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 ?? init?.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 +140,9 @@ 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.