docs: updated docs for retriever using param obj

refactor: Use parameter object for retrieve function of Retriever
fix: restore missing exports (#610 )
2026-07-19 18:43:34 -04:00 · 2024-03-06 14:30:20 +07:00 · 2024-03-06 14:19:42 +07:00 · 2024-03-05 14:56:25 -06:00 · 2024-03-05 14:37:54 -06:00 · 2024-03-05 14:20:58 -06:00
129 changed files with 1427 additions and 409 deletions
@@ -0,0 +1,5 @@
+---
+"create-llama": patch
+---
+
+Add LlamaParse option when selecting a pdf file or a folder
@@ -0,0 +1,12 @@
+---
+"llamaindex": patch
+"@llamaindex/core-test": patch
+---
+
+- Add missing exports:
+  - `IndexStructType`,
+  - `IndexDict`,
+  - `jsonToIndexStruct`,
+  - `IndexList`,
+  - `IndexStruct`
+- Fix `IndexDict.toJson()` method
@@ -0,0 +1,5 @@
+---
+"llamaindex": patch
+---
+
+Add streaming to agents
@@ -0,0 +1,5 @@
+---
+"create-llama": patch
+---
+
+Add embedding model option to create-llama
@@ -0,0 +1,5 @@
+---
+"llamaindex": minor
+---
+
+Use parameter object for retrieve function of Retriever (to align usage with query function of QueryEngine)
@@ -11,5 +11,13 @@ module.exports = {
    "max-params": ["error", 4],
    "prefer-const": "error",
  },
+  overrides: [
+    {
+      files: ["examples/**/*.ts"],
+      rules: {
+        "turbo/no-undeclared-env-vars": "off",
+      },
+    },
+  ],
  ignorePatterns: ["dist/", "lib/"],
 };
@@ -0,0 +1,24 @@
+name: Publish
+on:
+  push:
+    branches:
+      - main
+
+jobs:
+  publish:
+    runs-on: ubuntu-latest
+
+    permissions:
+      contents: read
+      id-token: write
+
+    steps:
+      - uses: actions/checkout@v4
+
+      - name: Publish @llamaindex/env
+        run: npx jsr publish
+        working-directory: packages/env
+
+      - name: Publish @llamaindex/core
+        run: npx jsr publish --allow-slow-types
+        working-directory: packages/core
@@ -44,6 +44,7 @@ test-results/
 playwright-report/
 blob-report/
 playwright/.cache/
+.tsbuildinfo

 # intellij
 **/.idea
@@ -1,5 +1,12 @@
 # docs

+## 0.0.4
+
+### Patch Changes
+
+- Updated dependencies [5116ad8]
+  - @llamaindex/env@0.0.5
+
 ## 0.0.3

 ### Patch Changes
@@ -23,3 +23,15 @@ const results = await queryEngine.query({
  query,
 });
 ```
+
+Per default, `HuggingFaceEmbedding` is using the `Xenova/all-MiniLM-L6-v2` model. You can change the model by passing the `modelType` parameter to the constructor.
+If you're not using a quantized model, set the `quantized` parameter to `false`.
+
+For example, to use the not quantized `BAAI/bge-small-en-v1.5` model, you can use the following code:
+
+```
+const embedModel = new HuggingFaceEmbedding({
+  modelType: "BAAI/bge-small-en-v1.5",
+  quantized: false,
+});
+```
@@ -100,7 +100,7 @@ const response = await queryEngine.query("<user_query>");
 ```ts
 import { SimilarityPostprocessor } from "llamaindex";

-nodes = await index.asRetriever().retrieve("test query str");
+nodes = await index.asRetriever().retrieve({ query: "test query str" });

 const processor = new SimilarityPostprocessor({
  similarityCutoff: 0.7,
@@ -11,7 +11,7 @@ const retriever = vector_index.asRetriever();
 retriever.similarityTopK = 3;

 // Fetch nodes!
-const nodesWithScore = await retriever.retrieve("query string");
+const nodesWithScore = await retriever.retrieve({ query: "query string" });
 ```

 ## API Reference
@@ -13,7 +13,7 @@ const retriever = vector_index.asRetriever();
 retriever.similarityTopK = 3;

 // جلب العقد!
-const nodesWithScore = await retriever.retrieve("سلسلة الاستعلام");
+const nodesWithScore = await retriever.retrieve({ query: "سلسلة الاستعلام" });
 ```

 ## مرجع الواجهة البرمجية (API Reference)
@@ -13,7 +13,7 @@ const retriever = vector_index.asRetriever();
 retriever.similarityTopK = 3;

 // Извличане на върхове!
-const nodesWithScore = await retriever.retrieve("query string");
+const nodesWithScore = await retriever.retrieve({ query: "query string" });
 ```

 ## API Reference (API справка)
@@ -13,7 +13,7 @@ const recuperador = vector_index.asRetriever();
 recuperador.similarityTopK = 3;

 // Obteniu els nodes!
-const nodesAmbPuntuació = await recuperador.retrieve("cadena de consulta");
+const nodesAmbPuntuació = await recuperador.retrieve({ query: "cadena de consulta" });
 ```

 ## Referència de l'API
@@ -13,7 +13,7 @@ const retriever = vector_index.asRetriever();
 retriever.similarityTopK = 3;

 // Získání uzlů!
-const nodesWithScore = await retriever.retrieve("dotazovací řetězec");
+const nodesWithScore = await retriever.retrieve({ query: "dotazovací řetězec" });
 ```

 ## API Reference (Odkazy na rozhraní)
@@ -13,7 +13,7 @@ const retriever = vector_index.asRetriever();
 retriever.similarityTopK = 3;

 // Hent noder!
-const nodesWithScore = await retriever.retrieve("forespørgselsstreng");
+const nodesWithScore = await retriever.retrieve({ query: "forespørgselsstreng" });
 ```

 ## API Reference
@@ -13,7 +13,7 @@ const retriever = vector_index.asRetriever();
 retriever.similarityTopK = 3;

 // Knoten abrufen!
-const nodesWithScore = await retriever.retrieve("Abfragezeichenfolge");
+const nodesWithScore = await retriever.retrieve({ query: "Abfragezeichenfolge" });
 ```

 ## API-Referenz
@@ -13,7 +13,7 @@ const retriever = vector_index.asRetriever();
 retriever.similarityTopK = 3;

 // Ανάκτηση κόμβων!
-const nodesWithScore = await retriever.retrieve("συμβολοσειρά ερωτήματος");
+const nodesWithScore = await retriever.retrieve({ query: "συμβολοσειρά ερωτήματος" });
 ```

 ## Αναφορά API
@@ -13,7 +13,7 @@ const recuperador = vector_index.asRetriever();
 recuperador.similarityTopK = 3;

 // ¡Obtener nodos!
-const nodosConPuntuación = await recuperador.retrieve("cadena de consulta");
+const nodosConPuntuación = await recuperador.retrieve({ query: "cadena de consulta" });
 ```

 ## Referencia de la API
@@ -13,7 +13,7 @@ const retriever = vector_index.asRetriever();
 retriever.similarityTopK = 3;

 // Too sõlmed!
-const nodesWithScore = await retriever.retrieve("päringu string");
+const nodesWithScore = await retriever.retrieve({ query: "päringu string" });
 ```

 ## API viide
@@ -13,7 +13,7 @@ const retriever = vector_index.asRetriever();
 retriever.similarityTopK = 3;

 // بازیابی گره ها!
-const nodesWithScore = await retriever.retrieve("رشته پرس و جو");
+const nodesWithScore = await retriever.retrieve({ query: "رشته پرس و جو" });
 ```

 ## مرجع API
@@ -13,7 +13,7 @@ const retriever = vector_index.asRetriever();
 retriever.similarityTopK = 3;

 // Hae solmut!
-const nodesWithScore = await retriever.retrieve("kyselymerkkijono");
+const nodesWithScore = await retriever.retrieve({ query: "kyselymerkkijono" });
 ```

 ## API-viite
@@ -11,7 +11,7 @@ const retriever = vector_index.asRetriever();
 retriever.similarityTopK = 3;

 // Récupérer les nœuds !
-const nodesWithScore = await retriever.retrieve("chaîne de requête");
+const nodesWithScore = await retriever.retrieve({ query: "chaîne de requête" });
 ```

 ## Référence de l'API
@@ -13,7 +13,7 @@ const retriever = vector_index.asRetriever();
 retriever.similarityTopK = 3;

 // אחזור צמתים!
-const nodesWithScore = await retriever.retrieve("מחרוזת שאילתה");
+const nodesWithScore = await retriever.retrieve({ query: "מחרוזת שאילתה" });
 ```

 ## מדריך לממשק API
@@ -13,7 +13,7 @@ const retriever = vector_index.asRetriever();
 retriever.similarityTopK = 3;

 // नोड्स प्राप्त करें!
-const nodesWithScore = await retriever.retrieve("क्वेरी स्ट्रिंग");
+const nodesWithScore = await retriever.retrieve({ query: "क्वेरी स्ट्रिंग" });
 ```

 ## एपीआई संदर्भ (API Reference)
@@ -13,7 +13,7 @@ const dohvatnik = vector_index.asRetriever();
 dohvatnik.similarityTopK = 3;

 // Dohvati čvorove!
-const čvoroviSaRezultatom = await dohvatnik.retrieve("upitni niz");
+const čvoroviSaRezultatom = await dohvatnik.retrieve({ query: "upitni niz" });
 ```

 ## API Referenca
@@ -13,7 +13,7 @@ const retriever = vector_index.asRetriever();
 retriever.similarityTopK = 3;

 // Node-ok lekérése!
-const nodesWithScore = await retriever.retrieve("lekérdezési karakterlánc");
+const nodesWithScore = await retriever.retrieve({ query: "lekérdezési karakterlánc" });
 ```

 ## API Referencia
@@ -13,7 +13,7 @@ const retriever = vector_index.asRetriever();
 retriever.similarityTopK = 3;

 // Mengambil node!
-const nodesWithScore = await retriever.retrieve("string query");
+const nodesWithScore = await retriever.retrieve({ query: "string query" });
 ```

 ## Referensi API
@@ -13,7 +13,7 @@ const retriever = vector_index.asRetriever();
 retriever.similarityTopK = 3;

 // Recupera i nodi!
-const nodesWithScore = await retriever.retrieve("stringa di query");
+const nodesWithScore = await retriever.retrieve({ query: "stringa di query" });
 ```

 ## Riferimento API
@@ -13,7 +13,7 @@ const retriever = vector_index.asRetriever();
 retriever.similarityTopK = 3;

 // ノードを取得します！
-const nodesWithScore = await retriever.retrieve("クエリ文字列");
+const nodesWithScore = await retriever.retrieve({ query: "クエリ文字列" });
 ```

 ## API リファレンス
@@ -13,7 +13,7 @@ const retriever = vector_index.asRetriever();
 retriever.similarityTopK = 3;

 // 노드를 가져옵니다!
-const nodesWithScore = await retriever.retrieve("쿼리 문자열");
+const nodesWithScore = await retriever.retrieve({ query: "쿼리 문자열" });
 ```

 ## API 참조
@@ -13,7 +13,7 @@ const gavėjas = vector_index.asRetriever();
 gavėjas.similarityTopK = 3;

 // Išgaunami mazgai!
-const mazgaiSuRezultatu = await gavėjas.retrieve("užklausos eilutė");
+const mazgaiSuRezultatu = await gavėjas.retrieve({ query: "užklausos eilutė" });
 ```

 ## API nuorodos (API Reference)
@@ -13,7 +13,7 @@ const retriever = vector_index.asRetriever();
 retriever.similarityTopK = 3;

 // Haal knooppunten op!
-const nodesWithScore = await retriever.retrieve("zoekopdracht");
+const nodesWithScore = await retriever.retrieve({ query: "zoekopdracht" });
 ```

 ## API Referentie
@@ -13,7 +13,7 @@ const retriever = vector_index.asRetriever();
 retriever.similarityTopK = 3;

 // Hent noder!
-const nodesWithScore = await retriever.retrieve("spørringsstreng");
+const nodesWithScore = await retriever.retrieve({ query: "spørringsstreng" });
 ```

 ## API-referanse
@@ -13,7 +13,7 @@ const retriever = vector_index.asRetriever();
 retriever.similarityTopK = 3;

 // Pobierz węzły!
-const nodesWithScore = await retriever.retrieve("ciąg zapytania");
+const nodesWithScore = await retriever.retrieve({ query: "ciąg zapytania" });
 ```

 ## Dokumentacja interfejsu API
@@ -13,7 +13,7 @@ const recuperador = vector_index.asRetriever();
 recuperador.similarityTopK = 3;

 // Buscar nós!
-const nósComPontuação = await recuperador.retrieve("string de consulta");
+const nósComPontuação = await recuperador.retrieve({ query: "string de consulta" });
 ```

 ## Referência da API
@@ -13,7 +13,7 @@ const recuperator = vector_index.asRetriever();
 recuperator.similarityTopK = 3;

 // Preia nodurile!
-const noduriCuScor = await recuperator.retrieve("șir de interogare");
+const noduriCuScor = await recuperator.retrieve({ query: "șir de interogare" });
 ```

 ## Referință API
@@ -13,7 +13,7 @@ const retriever = vector_index.asRetriever();
 retriever.similarityTopK = 3;

 // Получение узлов!
-const nodesWithScore = await retriever.retrieve("строка запроса");
+const nodesWithScore = await retriever.retrieve({ query: "строка запроса" });
 ```

 ## Справочник по API
@@ -13,7 +13,7 @@ const retriever = vector_index.asRetriever();
 retriever.similarityTopK = 3;

 // Dohvati čvorove!
-const nodesWithScore = await retriever.retrieve("upitni niz");
+const nodesWithScore = await retriever.retrieve({ query: "upitni niz" });
 ```

 ## API Referenca
@@ -13,7 +13,7 @@ const pridobitelj = vector_index.asRetriever();
 pridobitelj.similarityTopK = 3;

 // Pridobivanje vozlišč!
-const vozliščaZRezultatom = await pridobitelj.retrieve("poizvedbeni niz");
+const vozliščaZRezultatom = await pridobitelj.retrieve({ query: "poizvedbeni niz" });
 ```

 ## API Sklic
@@ -13,7 +13,7 @@ const retriever = vector_index.asRetriever();
 retriever.similarityTopK = 3;

 // Získajte uzly!
-const nodesWithScore = await retriever.retrieve("reťazec dotazu");
+const nodesWithScore = await retriever.retrieve({ query: "reťazec dotazu" });
 ```

 ## API Referencia
@@ -13,7 +13,7 @@ const retriever = vector_index.asRetriever();
 retriever.similarityTopK = 3;

 // Hämta noder!
-const nodesWithScore = await retriever.retrieve("frågesträng");
+const nodesWithScore = await retriever.retrieve({ query: "frågesträng" });
 ```

 ## API-referens
@@ -13,7 +13,7 @@ const retriever = vector_index.asRetriever();
 retriever.similarityTopK = 3;

 // เรียกคืนโหนด!
-const nodesWithScore = await retriever.retrieve("query string");
+const nodesWithScore = await retriever.retrieve({ query: "query string" });
 ```

 ## API Reference (การอ้างอิง API)
@@ -13,7 +13,7 @@ const retriever = vector_index.asRetriever();
 retriever.similarityTopK = 3;

 // Düğümleri getir!
-const nodesWithScore = await retriever.retrieve("sorgu dizesi");
+const nodesWithScore = await retriever.retrieve({ query: "sorgu dizesi" });
 ```

 ## API Referansı
@@ -13,7 +13,7 @@ const retriever = vector_index.asRetriever();
 retriever.similarityTopK = 3;

 // Отримати вузли!
-const nodesWithScore = await retriever.retrieve("рядок запиту");
+const nodesWithScore = await retriever.retrieve({ query: "рядок запиту" });
 ```

 ## Довідник API
@@ -13,7 +13,7 @@ const retriever = vector_index.asRetriever();
 retriever.similarityTopK = 3;

 // Lấy các node!
-const nodesWithScore = await retriever.retrieve("chuỗi truy vấn");
+const nodesWithScore = await retriever.retrieve({ query: "chuỗi truy vấn" });
 ```

 ## Tài liệu tham khảo API
@@ -11,7 +11,7 @@ const retriever = vector_index.asRetriever();
 retriever.similarityTopK = 3;

 // 获取节点！
-const nodesWithScore = await retriever.retrieve("查询字符串");
+const nodesWithScore = await retriever.retrieve({ query: "查询字符串" });
 ```

 ## API 参考
@@ -13,7 +13,7 @@ const retriever = vector_index.asRetriever();
 retriever.similarityTopK = 3;

 // 提取節點！
-const nodesWithScore = await retriever.retrieve("查詢字串");
+const nodesWithScore = await retriever.retrieve({ query: "查詢字串" });
 ```

 ## API 參考
@@ -1,6 +1,6 @@
 {
  "name": "docs",
-  "version": "0.0.3",
+  "version": "0.0.4",
  "private": true,
  "scripts": {
    "docusaurus": "docusaurus",
@@ -0,0 +1,95 @@
+import { FunctionTool, OpenAIAgent } from "llamaindex";
+
+// Define a function to sum two numbers
+function sumNumbers({ a, b }: { a: number; b: number }): number {
+  return a + b;
+}
+
+// Define a function to divide two numbers
+function divideNumbers({ a, b }: { a: number; b: number }): number {
+  return a / b;
+}
+
+// Define the parameters of the sum function as a JSON schema
+const sumJSON = {
+  type: "object",
+  properties: {
+    a: {
+      type: "number",
+      description: "The first number",
+    },
+    b: {
+      type: "number",
+      description: "The second number",
+    },
+  },
+  required: ["a", "b"],
+};
+
+const divideJSON = {
+  type: "object",
+  properties: {
+    a: {
+      type: "number",
+      description: "The dividend a to divide",
+    },
+    b: {
+      type: "number",
+      description: "The divisor b to divide by",
+    },
+  },
+  required: ["a", "b"],
+};
+
+async function main() {
+  // Create a function tool from the sum function
+  const functionTool = new FunctionTool(sumNumbers, {
+    name: "sumNumbers",
+    description: "Use this function to sum two numbers",
+    parameters: sumJSON,
+  });
+
+  // Create a function tool from the divide function
+  const functionTool2 = new FunctionTool(divideNumbers, {
+    name: "divideNumbers",
+    description: "Use this function to divide two numbers",
+    parameters: divideJSON,
+  });
+
+  // Create an OpenAIAgent with the function tools
+  const agent = new OpenAIAgent({
+    tools: [functionTool, functionTool2],
+    verbose: true,
+  });
+
+  // Create a task to sum and divide numbers
+  const task = agent.createTask("How much is 5 + 5? then divide by 2");
+
+  let count = 0;
+
+  while (true) {
+    const stepOutput = await agent.runStep(task.taskId);
+
+    console.log(`Runnning step ${count++}`);
+    console.log(`======== OUTPUT ==========`);
+    if (stepOutput.output.response) {
+      console.log(stepOutput.output.response);
+    } else {
+      console.log(stepOutput.output.sources);
+    }
+    console.log(`==========================`);
+
+    if (stepOutput.isLast) {
+      const finalResponse = await agent.finalizeResponse(
+        task.taskId,
+        stepOutput,
+      );
+      console.log({ finalResponse });
+      break;
+    }
+  }
+}
+
+main().then(() => {
+  console.log("Done");
+});
@@ -0,0 +1,64 @@
+import {
+  OpenAIAgent,
+  QueryEngineTool,
+  SimpleDirectoryReader,
+  VectorStoreIndex,
+} from "llamaindex";
+
+async function main() {
+  // Load the documents
+  const documents = await new SimpleDirectoryReader().loadData({
+    directoryPath: "node_modules/llamaindex/examples",
+  });
+
+  // Create a vector index from the documents
+  const vectorIndex = await VectorStoreIndex.fromDocuments(documents);
+
+  // Create a query engine from the vector index
+  const abramovQueryEngine = vectorIndex.asQueryEngine();
+
+  // Create a QueryEngineTool with the query engine
+  const queryEngineTool = new QueryEngineTool({
+    queryEngine: abramovQueryEngine,
+    metadata: {
+      name: "abramov_query_engine",
+      description: "A query engine for the Abramov documents",
+    },
+  });
+
+  // Create an OpenAIAgent with the function tools
+  const agent = new OpenAIAgent({
+    tools: [queryEngineTool],
+    verbose: true,
+  });
+
+  const task = agent.createTask("What was his salary?");
+
+  let count = 0;
+
+  while (true) {
+    const stepOutput = await agent.runStep(task.taskId);
+
+    console.log(`Runnning step ${count++}`);
+    console.log(`======== OUTPUT ==========`);
+    if (stepOutput.output.response) {
+      console.log(stepOutput.output.response);
+    } else {
+      console.log(stepOutput.output.sources);
+    }
+    console.log(`==========================`);
+
+    if (stepOutput.isLast) {
+      const finalResponse = await agent.finalizeResponse(
+        task.taskId,
+        stepOutput,
+      );
+      console.log({ finalResponse });
+      break;
+    }
+  }
+}
+
+main().then(() => {
+  console.log("Done");
+});
@@ -0,0 +1,90 @@
+import { FunctionTool, ReActAgent } from "llamaindex";
+
+// Define a function to sum two numbers
+function sumNumbers({ a, b }: { a: number; b: number }): number {
+  return a + b;
+}
+
+// Define a function to divide two numbers
+function divideNumbers({ a, b }: { a: number; b: number }): number {
+  return a / b;
+}
+
+// Define the parameters of the sum function as a JSON schema
+const sumJSON = {
+  type: "object",
+  properties: {
+    a: {
+      type: "number",
+      description: "The first number",
+    },
+    b: {
+      type: "number",
+      description: "The second number",
+    },
+  },
+  required: ["a", "b"],
+};
+
+const divideJSON = {
+  type: "object",
+  properties: {
+    a: {
+      type: "number",
+      description: "The dividend",
+    },
+    b: {
+      type: "number",
+      description: "The divisor",
+    },
+  },
+  required: ["a", "b"],
+};
+
+async function main() {
+  // Create a function tool from the sum function
+  const functionTool = new FunctionTool(sumNumbers, {
+    name: "sumNumbers",
+    description: "Use this function to sum two numbers",
+    parameters: sumJSON,
+  });
+
+  // Create a function tool from the divide function
+  const functionTool2 = new FunctionTool(divideNumbers, {
+    name: "divideNumbers",
+    description: "Use this function to divide two numbers",
+    parameters: divideJSON,
+  });
+
+  // Create an OpenAIAgent with the function tools
+  const agent = new ReActAgent({
+    tools: [functionTool, functionTool2],
+    verbose: true,
+  });
+
+  const task = agent.createTask("Divide 16 by 2 then add 20");
+
+  let count = 0;
+
+  while (true) {
+    const stepOutput = await agent.runStep(task.taskId);
+
+    console.log(`Runnning step ${count++}`);
+    console.log(`======== OUTPUT ==========`);
+    console.log(stepOutput.output);
+    console.log(`==========================`);
+
+    if (stepOutput.isLast) {
+      const finalResponse = await agent.finalizeResponse(
+        task.taskId,
+        stepOutput,
+      );
+      console.log({ finalResponse });
+      break;
+    }
+  }
+}
+
+main().then(() => {
+  console.log("Done");
+});
@@ -0,0 +1,77 @@
+import { FunctionTool, OpenAIAgent } from "llamaindex";
+
+// Define a function to sum two numbers
+function sumNumbers({ a, b }: { a: number; b: number }): number {
+  return a + b;
+}
+
+// Define a function to divide two numbers
+function divideNumbers({ a, b }: { a: number; b: number }): number {
+  return a / b;
+}
+
+// Define the parameters of the sum function as a JSON schema
+const sumJSON = {
+  type: "object",
+  properties: {
+    a: {
+      type: "number",
+      description: "The first number",
+    },
+    b: {
+      type: "number",
+      description: "The second number",
+    },
+  },
+  required: ["a", "b"],
+};
+
+const divideJSON = {
+  type: "object",
+  properties: {
+    a: {
+      type: "number",
+      description: "The dividend",
+    },
+    b: {
+      type: "number",
+      description: "The divisor",
+    },
+  },
+  required: ["a", "b"],
+};
+
+async function main() {
+  // Create a function tool from the sum function
+  const functionTool = new FunctionTool(sumNumbers, {
+    name: "sumNumbers",
+    description: "Use this function to sum two numbers",
+    parameters: sumJSON,
+  });
+
+  // Create a function tool from the divide function
+  const functionTool2 = new FunctionTool(divideNumbers, {
+    name: "divideNumbers",
+    description: "Use this function to divide two numbers",
+    parameters: divideJSON,
+  });
+
+  // Create an OpenAIAgent with the function tools
+  const agent = new OpenAIAgent({
+    tools: [functionTool, functionTool2],
+    verbose: false,
+  });
+
+  const stream = await agent.chat({
+    message: "Divide 16 by 2 then add 20",
+    stream: true,
+  });
+
+  for await (const chunk of stream.response) {
+    process.stdout.write(chunk.response);
+  }
+}
+
+main().then(() => {
+  console.log("\nDone");
+});
@@ -3,6 +3,7 @@ import { Anthropic } from "llamaindex";
 (async () => {
  const anthropic = new Anthropic({
    apiKey: process.env.ANTHROPIC_API_KEY,
+    model: "claude-3-opus",
  });
  const result = await anthropic.chat({
    messages: [
@@ -0,0 +1,34 @@
+import { Anthropic, SimpleChatEngine, SimpleChatHistory } from "llamaindex";
+import { stdin as input, stdout as output } from "node:process";
+import readline from "node:readline/promises";
+
+(async () => {
+  const llm = new Anthropic({
+    apiKey: process.env.ANTHROPIC_API_KEY,
+    model: "claude-3-opus",
+  });
+  // chatHistory will store all the messages in the conversation
+  const chatHistory = new SimpleChatHistory({
+    messages: [
+      {
+        content: "You want to talk in rhymes.",
+        role: "system",
+      },
+    ],
+  });
+  const chatEngine = new SimpleChatEngine({
+    llm,
+    chatHistory,
+  });
+  const rl = readline.createInterface({ input, output });
+
+  while (true) {
+    const query = await rl.question("User: ");
+    process.stdout.write("Assistant: ");
+    const stream = await chatEngine.chat({ message: query, stream: true });
+    for await (const chunk of stream) {
+      process.stdout.write(chunk.response);
+    }
+    process.stdout.write("\n");
+  }
+})();
@@ -0,0 +1,23 @@
+import { Anthropic } from "llamaindex";
+
+(async () => {
+  const anthropic = new Anthropic({
+    apiKey: process.env.ANTHROPIC_API_KEY,
+    model: "claude-instant-1.2",
+  });
+  const stream = await anthropic.chat({
+    messages: [
+      { content: "You want to talk in rhymes.", role: "system" },
+      {
+        content:
+          "How much wood would a woodchuck chuck if a woodchuck could chuck wood?",
+        role: "user",
+      },
+    ],
+    stream: true,
+  });
+
+  for await (const chunk of stream) {
+    process.stdout.write(chunk.delta);
+  }
+})();
@@ -27,9 +27,9 @@ async function main() {
  // retrieve documents using the index
  const index = await createIndex();
  const retriever = index.asRetriever({ similarityTopK: 3 });
-  const results = await retriever.retrieve(
-    "what are Vincent van Gogh's famous paintings",
-  );
+  const results = await retriever.retrieve({
+    query: "what are Vincent van Gogh's famous paintings",
+  });
  for (const result of results) {
    const node = result.node;
    if (!node) {
@@ -1,5 +1,23 @@
 # llamaindex

+## 0.1.21
+
+### Patch Changes
+
+- 552a61a: Add quantized parameter to HuggingFaceEmbedding
+- d824876: Add support for Claude 3
+
+## 0.1.20
+
+### Patch Changes
+
+- 64683a5: fix: prefix messages always true
+- 698cd9c: fix: step wise agent + examples
+- 7257751: fixed removeRefDocNode and persist store on delete
+- 5116ad8: fix: compatibility issue with Deno
+- Updated dependencies [5116ad8]
+  - @llamaindex/env@0.0.5
+
 ## 0.1.19

 ### Patch Changes
@@ -0,0 +1,8 @@
+{
+  "name": "@llamaindex/core",
+  "version": "0.1.21",
+  "exports": "./src/index.ts",
+  "imports": {
+    "@llamaindex/env": "jsr:@llamaindex/env@0.0.5"
+  }
+}
@@ -1,22 +1,22 @@
 {
  "name": "llamaindex",
-  "version": "0.1.19",
+  "version": "0.1.21",
  "license": "MIT",
  "type": "module",
  "dependencies": {
-    "@anthropic-ai/sdk": "^0.13.0",
+    "@anthropic-ai/sdk": "^0.15.0",
    "@aws-crypto/sha256-js": "^5.2.0",
    "@datastax/astra-db-ts": "^0.1.4",
-    "@types/lodash": "^4.14.202",
-    "@types/node": "^18.19.14",
-    "@types/papaparse": "^5.3.14",
-    "@types/pg": "^8.11.0",
-    "@llamaindex/cloud": "^0.0.1",
+    "@llamaindex/cloud": "0.0.4",
    "@llamaindex/env": "workspace:*",
    "@mistralai/mistralai": "^0.0.10",
    "@notionhq/client": "^2.2.14",
    "@pinecone-database/pinecone": "^2.0.1",
    "@qdrant/js-client-rest": "^1.7.0",
+    "@types/lodash": "^4.14.202",
+    "@types/node": "^18.19.14",
+    "@types/papaparse": "^5.3.14",
+    "@types/pg": "^8.11.0",
    "@xenova/transformers": "^2.15.0",
    "assemblyai": "^4.2.2",
    "chromadb": "~1.7.3",
@@ -2,14 +2,16 @@ import type { Event } from "./callbacks/CallbackManager.js";
 import type { NodeWithScore } from "./Node.js";
 import type { ServiceContext } from "./ServiceContext.js";

+export type RetrieveParams = {
+  query: string;
+  parentEvent?: Event;
+  preFilters?: unknown;
+};
+
 /**
 * Retrievers retrieve the nodes that most closely match our query in similarity.
 */
 export interface BaseRetriever {
-  retrieve(
-    query: string,
-    parentEvent?: Event,
-    preFilters?: unknown,
-  ): Promise<NodeWithScore[]>;
+  retrieve(params: RetrieveParams): Promise<NodeWithScore[]>;
  getServiceContext(): ServiceContext;
 }
@@ -37,8 +37,6 @@ export class OpenAIAgent extends AgentRunner {
    toolRetriever,
    systemPrompt,
  }: OpenAIAgentParams) {
-    prefixMessages = prefixMessages || [];
-
    llm = llm ?? new OpenAI({ model: "gpt-3.5-turbo-0613" });

    if (systemPrompt) {
@@ -1,10 +1,12 @@
 // Assuming that the necessary interfaces and classes (like BaseTool, OpenAI, ChatMessage, CallbackManager, etc.) are defined elsewhere

 import { randomUUID } from "@llamaindex/env";
+import { Response } from "../../Response.js";
 import type { CallbackManager } from "../../callbacks/CallbackManager.js";
 import {
  AgentChatResponse,
  ChatResponseMode,
+  StreamingAgentChatResponse,
 } from "../../engines/chat/types.js";
 import type {
  ChatMessage,
@@ -12,6 +14,7 @@ import type {
  ChatResponseChunk,
 } from "../../llm/index.js";
 import { OpenAI } from "../../llm/index.js";
+import { streamConverter, streamReducer } from "../../llm/utils.js";
 import { ChatMemoryBuffer } from "../../memory/ChatMemoryBuffer.js";
 import type { ObjectRetriever } from "../../objects/base.js";
 import type { ToolOutput } from "../../tools/types.js";
@@ -192,13 +195,40 @@ export class OpenAIAgentWorker implements AgentWorker {
  private _processMessage(
    task: Task,
    chatResponse: ChatResponse,
-  ): AgentChatResponse | AsyncIterable<ChatResponseChunk> {
+  ): AgentChatResponse {
    const aiMessage = chatResponse.message;
    task.extraState.newMemory.put(aiMessage);

    return new AgentChatResponse(aiMessage.content, task.extraState.sources);
  }

+  private async _getStreamAiResponse(
+    task: Task,
+    llmChatKwargs: any,
+  ): Promise<StreamingAgentChatResponse> {
+    const stream = await this.llm.chat({
+      stream: true,
+      ...llmChatKwargs,
+    });
+
+    const iterator = streamConverter(
+      streamReducer({
+        stream,
+        initialValue: "",
+        reducer: (accumulator, part) => (accumulator += part.delta),
+        finished: (accumulator) => {
+          task.extraState.newMemory.put({
+            content: accumulator,
+            role: "assistant",
+          });
+        },
+      }),
+      (r: ChatResponseChunk) => new Response(r.delta),
+    );
+
+    return new StreamingAgentChatResponse(iterator, task.extraState.sources);
+  }
+
  /**
   * Get agent response.
   * @param task: task
@@ -210,7 +240,7 @@ export class OpenAIAgentWorker implements AgentWorker {
    task: Task,
    mode: ChatResponseMode,
    llmChatKwargs: any,
-  ): Promise<AgentChatResponse> {
+  ): Promise<AgentChatResponse | StreamingAgentChatResponse> {
    if (mode === ChatResponseMode.WAIT) {
      const chatResponse = (await this.llm.chat({
        stream: false,
@@ -218,9 +248,11 @@ export class OpenAIAgentWorker implements AgentWorker {
      })) as unknown as ChatResponse;

      return this._processMessage(task, chatResponse) as AgentChatResponse;
-    } else {
-      throw new Error("Not implemented");
+    } else if (mode === ChatResponseMode.STREAM) {
+      return this._getStreamAiResponse(task, llmChatKwargs);
    }
+
+    throw new Error("Invalid mode");
  }

  /**
@@ -4,6 +4,7 @@ import type { ChatEngineAgentParams } from "../../engines/chat/index.js";
 import {
  AgentChatResponse,
  ChatResponseMode,
+  StreamingAgentChatResponse,
 } from "../../engines/chat/index.js";
 import type { ChatMessage, LLM } from "../../llm/index.js";
 import { ChatMemoryBuffer } from "../../memory/ChatMemoryBuffer.js";
@@ -14,7 +15,7 @@ import { AgentState, BaseAgentRunner, TaskState } from "./types.js";

 const validateStepFromArgs = (
  taskId: string,
-  input: string,
+  input?: string | null,
  step?: any,
  kwargs?: any,
 ): TaskStep | undefined => {
@@ -24,6 +25,7 @@ const validateStepFromArgs = (
    }
    return step;
  } else {
+    if (!input) return;
    return new TaskStep(taskId, step, input, kwargs);
  }
 };
@@ -194,7 +196,7 @@ export class AgentRunner extends BaseAgentRunner {
   */
  async runStep(
    taskId: string,
-    input: string,
+    input?: string | null,
    step?: TaskStep,
    kwargs: any = {},
  ): Promise<TaskStepOutput> {
@@ -230,23 +232,26 @@ export class AgentRunner extends BaseAgentRunner {
    taskId: string,
    stepOutput: TaskStepOutput,
    kwargs?: any,
-  ): Promise<AgentChatResponse> {
+  ): Promise<AgentChatResponse | StreamingAgentChatResponse> {
    if (!stepOutput) {
      stepOutput =
        this.getCompletedSteps(taskId)[
          this.getCompletedSteps(taskId).length - 1
        ];
    }
+
    if (!stepOutput.isLast) {
      throw new Error(
        "finalizeResponse can only be called on the last step output",
      );
    }

-    if (!(stepOutput.output instanceof AgentChatResponse)) {
-      throw new Error(
-        `When \`isLast\` is True, cur_step_output.output must be AGENT_CHAT_RESPONSE_TYPE: ${stepOutput.output}`,
-      );
+    if (!(stepOutput.output instanceof StreamingAgentChatResponse)) {
+      if (!(stepOutput.output instanceof AgentChatResponse)) {
+        throw new Error(
+          `When \`isLast\` is True, cur_step_output.output must be AGENT_CHAT_RESPONSE_TYPE: ${stepOutput.output}`,
+        );
+      }
    }

    this.agentWorker.finalizeTask(this.getTask(taskId), kwargs);
@@ -261,20 +266,32 @@ export class AgentRunner extends BaseAgentRunner {
  protected async _chat({
    message,
    toolChoice,
-  }: ChatEngineAgentParams & { mode: ChatResponseMode }) {
+    stream,
+  }: ChatEngineAgentParams): Promise<AgentChatResponse>;
+  protected async _chat({
+    message,
+    toolChoice,
+    stream,
+  }: ChatEngineAgentParams & {
+    stream: true;
+  }): Promise<StreamingAgentChatResponse>;
+  protected async _chat({
+    message,
+    toolChoice,
+    stream,
+  }: ChatEngineAgentParams): Promise<
+    AgentChatResponse | StreamingAgentChatResponse
+  > {
    const task = this.createTask(message as string);

    let resultOutput;

+    const mode = stream ? ChatResponseMode.STREAM : ChatResponseMode.WAIT;
+
    while (true) {
-      const curStepOutput = await this._runStep(
-        task.taskId,
-        undefined,
-        ChatResponseMode.WAIT,
-        {
-          toolChoice,
-        },
-      );
+      const curStepOutput = await this._runStep(task.taskId, undefined, mode, {
+        toolChoice,
+      });

      if (curStepOutput.isLast) {
        resultOutput = curStepOutput;
@@ -298,7 +315,26 @@ export class AgentRunner extends BaseAgentRunner {
    message,
    chatHistory,
    toolChoice,
-  }: ChatEngineAgentParams): Promise<AgentChatResponse> {
+    stream,
+  }: ChatEngineAgentParams & {
+    stream?: false;
+  }): Promise<AgentChatResponse>;
+  public async chat({
+    message,
+    chatHistory,
+    toolChoice,
+    stream,
+  }: ChatEngineAgentParams & {
+    stream: true;
+  }): Promise<StreamingAgentChatResponse>;
+  public async chat({
+    message,
+    chatHistory,
+    toolChoice,
+    stream,
+  }: ChatEngineAgentParams): Promise<
+    AgentChatResponse | StreamingAgentChatResponse
+  > {
    if (!toolChoice) {
      toolChoice = this.defaultToolChoice;
    }
@@ -307,7 +343,7 @@ export class AgentRunner extends BaseAgentRunner {
      message,
      chatHistory,
      toolChoice,
-      mode: ChatResponseMode.WAIT,
+      stream,
    });

    return chatResponse;
@@ -1,4 +1,7 @@
-import type { AgentChatResponse } from "../../engines/chat/index.js";
+import type {
+  AgentChatResponse,
+  StreamingAgentChatResponse,
+} from "../../engines/chat/index.js";
 import type { Task, TaskStep, TaskStepOutput } from "../types.js";
 import { BaseAgent } from "../types.js";

@@ -57,7 +60,7 @@ export abstract class BaseAgentRunner extends BaseAgent {
    taskId: string,
    stepOutput: TaskStepOutput,
    kwargs?: any,
-  ): Promise<AgentChatResponse>;
+  ): Promise<AgentChatResponse | StreamingAgentChatResponse>;

  abstract undoStep(taskId: string): void;
 }
@@ -1,6 +1,7 @@
 import type {
  AgentChatResponse,
  ChatEngineAgentParams,
+  StreamingAgentChatResponse,
 } from "../engines/chat/index.js";
 import type { QueryEngineParamsNonStreaming } from "../types.js";

@@ -12,11 +13,15 @@ export interface AgentWorker {
 }

 interface BaseChatEngine {
-  chat(params: ChatEngineAgentParams): Promise<AgentChatResponse>;
+  chat(
+    params: ChatEngineAgentParams,
+  ): Promise<AgentChatResponse | StreamingAgentChatResponse>;
 }

 interface BaseQueryEngine {
-  query(params: QueryEngineParamsNonStreaming): Promise<AgentChatResponse>;
+  query(
+    params: QueryEngineParamsNonStreaming,
+  ): Promise<AgentChatResponse | StreamingAgentChatResponse>;
 }

 /**
@@ -31,7 +36,10 @@ export abstract class BaseAgent implements BaseChatEngine, BaseQueryEngine {
    return [];
  }

-  abstract chat(params: ChatEngineAgentParams): Promise<AgentChatResponse>;
+  abstract chat(
+    params: ChatEngineAgentParams,
+  ): Promise<AgentChatResponse | StreamingAgentChatResponse>;
+
  abstract reset(): void;

  /**
@@ -41,7 +49,7 @@ export abstract class BaseAgent implements BaseChatEngine, BaseQueryEngine {
   */
  async query(
    params: QueryEngineParamsNonStreaming,
-  ): Promise<AgentChatResponse> {
+  ): Promise<AgentChatResponse | StreamingAgentChatResponse> {
    // Handle non-streaming query
    const agentResponse = await this.chat({
      message: params.query,
@@ -161,13 +169,13 @@ export class TaskStep implements ITaskStep {
 * @param isLast: isLast
 */
 export class TaskStepOutput {
-  output: unknown;
+  output: any;
  taskStep: TaskStep;
  nextSteps: TaskStep[];
  isLast: boolean;

  constructor(
-    output: unknown,
+    output: any,
    taskStep: TaskStep,
    nextSteps: TaskStep[],
    isLast: boolean = false,
@@ -3,7 +3,7 @@ import { RetrieverQueryEngine } from "../engines/query/RetrieverQueryEngine.js";
 import type { BaseNodePostprocessor } from "../postprocessors/types.js";
 import type { BaseSynthesizer } from "../synthesizers/types.js";
 import type { BaseQueryEngine } from "../types.js";
-import type { RetrieveParams } from "./LlamaCloudRetriever.js";
+import type { CloudRetrieveParams } from "./LlamaCloudRetriever.js";
 import { LlamaCloudRetriever } from "./LlamaCloudRetriever.js";
 import type { CloudConstructorParams } from "./types.js";

@@ -14,7 +14,7 @@ export class LlamaCloudIndex {
    this.params = params;
  }

-  asRetriever(params: RetrieveParams = {}): BaseRetriever {
+  asRetriever(params: CloudRetrieveParams = {}): BaseRetriever {
    return new LlamaCloudRetriever({ ...this.params, ...params });
  }

@@ -23,7 +23,7 @@ export class LlamaCloudIndex {
      responseSynthesizer?: BaseSynthesizer;
      preFilters?: unknown;
      nodePostprocessors?: BaseNodePostprocessor[];
-    } & RetrieveParams,
+    } & CloudRetrieveParams,
  ): BaseQueryEngine {
    const retriever = new LlamaCloudRetriever({
      ...this.params,
@@ -2,15 +2,14 @@ import type { PlatformApi, PlatformApiClient } from "@llamaindex/cloud";
 import { globalsHelper } from "../GlobalsHelper.js";
 import type { NodeWithScore } from "../Node.js";
 import { ObjectType, jsonToNode } from "../Node.js";
-import type { BaseRetriever } from "../Retriever.js";
+import type { BaseRetriever, RetrieveParams } from "../Retriever.js";
 import type { ServiceContext } from "../ServiceContext.js";
 import { serviceContextFromDefaults } from "../ServiceContext.js";
-import type { Event } from "../callbacks/CallbackManager.js";
 import type { ClientParams, CloudConstructorParams } from "./types.js";
 import { DEFAULT_PROJECT_NAME } from "./types.js";
 import { getClient } from "./utils.js";

-export type RetrieveParams = Omit<
+export type CloudRetrieveParams = Omit<
  PlatformApi.RetrievalParams,
  "query" | "searchFilters" | "pipelineId" | "className"
 > & { similarityTopK?: number };
@@ -18,7 +17,7 @@ export type RetrieveParams = Omit<
 export class LlamaCloudRetriever implements BaseRetriever {
  client?: PlatformApiClient;
  clientParams: ClientParams;
-  retrieveParams: RetrieveParams;
+  retrieveParams: CloudRetrieveParams;
  projectName: string = DEFAULT_PROJECT_NAME;
  pipelineName: string;
  serviceContext: ServiceContext;
@@ -35,7 +34,7 @@ export class LlamaCloudRetriever implements BaseRetriever {
    });
  }

-  constructor(params: CloudConstructorParams & RetrieveParams) {
+  constructor(params: CloudConstructorParams & CloudRetrieveParams) {
    this.clientParams = { apiKey: params.apiKey, baseUrl: params.baseUrl };
    if (params.similarityTopK) {
      params.denseSimilarityTopK = params.similarityTopK;
@@ -55,11 +54,11 @@ export class LlamaCloudRetriever implements BaseRetriever {
    return this.client;
  }

-  async retrieve(
-    query: string,
-    parentEvent?: Event | undefined,
-    preFilters?: unknown,
-  ): Promise<NodeWithScore[]> {
+  async retrieve({
+    query,
+    parentEvent,
+    preFilters,
+  }: RetrieveParams): Promise<NodeWithScore[]> {
    const pipelines = await (
      await this.getClient()
    ).pipeline.searchPipelines({
@@ -1,4 +1,5 @@
 import type { PlatformApiClient } from "@llamaindex/cloud";
+import { getEnv } from "@llamaindex/env";
 import type { ClientParams } from "./types.js";
 import { DEFAULT_BASE_URL } from "./types.js";

@@ -7,8 +8,8 @@ export async function getClient({
  baseUrl,
 }: ClientParams = {}): Promise<PlatformApiClient> {
  // Get the environment variables or use defaults
-  baseUrl = baseUrl ?? process.env.LLAMA_CLOUD_BASE_URL ?? DEFAULT_BASE_URL;
-  apiKey = apiKey ?? process.env.LLAMA_CLOUD_API_KEY;
+  baseUrl = baseUrl ?? getEnv("LLAMA_CLOUD_BASE_URL") ?? DEFAULT_BASE_URL;
+  apiKey = apiKey ?? getEnv("LLAMA_CLOUD_API_KEY");

  const { PlatformApiClient } = await import("@llamaindex/cloud");

@@ -20,6 +20,7 @@ export enum HuggingFaceEmbeddingModelType {
 */
 export class HuggingFaceEmbedding extends BaseEmbedding {
  modelType: string = HuggingFaceEmbeddingModelType.XENOVA_ALL_MINILM_L6_V2;
+  quantized: boolean = true;

  private extractor: any;

@@ -31,7 +32,9 @@ export class HuggingFaceEmbedding extends BaseEmbedding {
  async getExtractor() {
    if (!this.extractor) {
      const { pipeline } = await import("@xenova/transformers");
-      this.extractor = await pipeline("feature-extraction", this.modelType);
+      this.extractor = await pipeline("feature-extraction", this.modelType, {
+        quantized: this.quantized,
+      });
    }
    return this.extractor;
  }
@@ -1,9 +1,10 @@
+import { getEnv } from "@llamaindex/env";
 import { OpenAIEmbedding } from "./OpenAIEmbedding.js";

 export class FireworksEmbedding extends OpenAIEmbedding {
  constructor(init?: Partial<OpenAIEmbedding>) {
    const {
-      apiKey = process.env.FIREWORKS_API_KEY,
+      apiKey = getEnv("FIREWORKS_API_KEY"),
      additionalSessionOptions = {},
      model = "nomic-ai/nomic-embed-text-v1.5",
      ...rest
@@ -1,9 +1,10 @@
+import { getEnv } from "@llamaindex/env";
 import { OpenAIEmbedding } from "./OpenAIEmbedding.js";

 export class TogetherEmbedding extends OpenAIEmbedding {
  constructor(init?: Partial<OpenAIEmbedding>) {
    const {
-      apiKey = process.env.TOGETHER_API_KEY,
+      apiKey = getEnv("TOGETHER_API_KEY"),
      additionalSessionOptions = {},
      model = "togethercomputer/m2-bert-80M-32k-retrieval",
      ...rest
@@ -64,10 +64,10 @@ export class DefaultContextGenerator
        tags: ["final"],
      };
    }
-    const sourceNodesWithScore = await this.retriever.retrieve(
-      message,
+    const sourceNodesWithScore = await this.retriever.retrieve({
+      query: message,
      parentEvent,
-    );
+    });

    const nodes = await this.applyNodePostprocessors(
      sourceNodesWithScore,
@@ -27,6 +27,7 @@ export interface ChatEngineParamsNonStreaming extends ChatEngineParamsBase {

 export interface ChatEngineAgentParams extends ChatEngineParamsBase {
  toolChoice?: string | Record<string, any>;
+  stream?: boolean;
 }

 /**
@@ -86,3 +87,20 @@ export class AgentChatResponse {
    return this.response ?? "";
  }
 }
+
+export class StreamingAgentChatResponse {
+  response: AsyncIterable<Response>;
+
+  sources: ToolOutput[];
+  sourceNodes?: BaseNode[];
+
+  constructor(
+    response: AsyncIterable<Response>,
+    sources?: ToolOutput[],
+    sourceNodes?: BaseNode[],
+  ) {
+    this.response = response;
+    this.sources = sources ?? [];
+    this.sourceNodes = sourceNodes ?? [];
+  }
+}
@@ -63,11 +63,11 @@ export class RetrieverQueryEngine
  }

  private async retrieve(query: string, parentEvent: Event) {
-    const nodes = await this.retriever.retrieve(
+    const nodes = await this.retriever.retrieve({
      query,
      parentEvent,
-      this.preFilters,
-    );
+      preFilters: this.preFilters,
+    });

    return await this.applyNodePostprocessors(nodes, query);
  }
@@ -1,4 +1,6 @@
 export * from "./BaseIndex.js";
+export * from "./IndexStruct.js";
+export * from "./json-to-index-struct.js";
 export * from "./keyword/index.js";
 export * from "./summary/index.js";
 export * from "./vectorStore/index.js";
@@ -24,9 +24,15 @@ export class IndexDict extends IndexStruct {
  }

  toJson(): Record<string, unknown> {
+    const nodesDict: Record<string, unknown> = {};
+
+    for (const [key, node] of Object.entries(this.nodesDict)) {
+      nodesDict[key] = node.toJSON();
+    }
+
    return {
      ...super.toJson(),
-      nodesDict: this.nodesDict,
+      nodesDict,
      type: this.type,
    };
  }
@@ -8,7 +8,7 @@ import {
  defaultKeywordExtractPrompt,
  defaultQueryKeywordExtractPrompt,
 } from "../../Prompt.js";
-import type { BaseRetriever } from "../../Retriever.js";
+import type { BaseRetriever, RetrieveParams } from "../../Retriever.js";
 import type { ServiceContext } from "../../ServiceContext.js";
 import { serviceContextFromDefaults } from "../../ServiceContext.js";
 import { RetrieverQueryEngine } from "../../engines/query/index.js";
@@ -79,7 +79,7 @@ abstract class BaseKeywordTableRetriever implements BaseRetriever {

  abstract getKeywords(query: string): Promise<string[]>;

-  async retrieve(query: string): Promise<NodeWithScore[]> {
+  async retrieve({ query }: RetrieveParams): Promise<NodeWithScore[]> {
    const keywords = await this.getKeywords(query);
    const chunkIndicesCount: { [key: string]: number } = {};
    const filteredKeywords = keywords.filter((keyword) =>
@@ -3,10 +3,9 @@ import { globalsHelper } from "../../GlobalsHelper.js";
 import type { BaseNode, Document, NodeWithScore } from "../../Node.js";
 import type { ChoiceSelectPrompt } from "../../Prompt.js";
 import { defaultChoiceSelectPrompt } from "../../Prompt.js";
-import type { BaseRetriever } from "../../Retriever.js";
+import type { BaseRetriever, RetrieveParams } from "../../Retriever.js";
 import type { ServiceContext } from "../../ServiceContext.js";
 import { serviceContextFromDefaults } from "../../ServiceContext.js";
-import type { Event } from "../../callbacks/CallbackManager.js";
 import { RetrieverQueryEngine } from "../../engines/query/index.js";
 import type { BaseNodePostprocessor } from "../../postprocessors/index.js";
 import type {
@@ -281,7 +280,10 @@ export class SummaryIndexRetriever implements BaseRetriever {
    this.index = index;
  }

-  async retrieve(query: string, parentEvent?: Event): Promise<NodeWithScore[]> {
+  async retrieve({
+    query,
+    parentEvent,
+  }: RetrieveParams): Promise<NodeWithScore[]> {
    const nodeIds = this.index.indexStruct.nodes;
    const nodes = await this.index.docStore.getNodes(nodeIds);
    const result = nodes.map((node) => ({
@@ -337,7 +339,10 @@ export class SummaryIndexLLMRetriever implements BaseRetriever {
    this.serviceContext = serviceContext || index.serviceContext;
  }

-  async retrieve(query: string, parentEvent?: Event): Promise<NodeWithScore[]> {
+  async retrieve({
+    query,
+    parentEvent,
+  }: RetrieveParams): Promise<NodeWithScore[]> {
    const nodeIds = this.index.indexStruct.nodes;
    const results: NodeWithScore[] = [];

@@ -11,7 +11,7 @@ import {
  ObjectType,
  splitNodesByType,
 } from "../../Node.js";
-import type { BaseRetriever } from "../../Retriever.js";
+import type { BaseRetriever, RetrieveParams } from "../../Retriever.js";
 import type { ServiceContext } from "../../ServiceContext.js";
 import { serviceContextFromDefaults } from "../../ServiceContext.js";
 import type { Event } from "../../callbacks/CallbackManager.js";
@@ -426,14 +426,17 @@ export class VectorIndexRetriever implements BaseRetriever {
    this.imageSimilarityTopK = imageSimilarityTopK ?? DEFAULT_SIMILARITY_TOP_K;
  }

-  async retrieve(
-    query: string,
-    parentEvent?: Event,
-    preFilters?: MetadataFilters,
-  ): Promise<NodeWithScore[]> {
-    let nodesWithScores = await this.textRetrieve(query, preFilters);
+  async retrieve({
+    query,
+    parentEvent,
+    preFilters,
+  }: RetrieveParams): Promise<NodeWithScore[]> {
+    let nodesWithScores = await this.textRetrieve(
+      query,
+      preFilters as MetadataFilters,
+    );
    nodesWithScores = nodesWithScores.concat(
-      await this.textToImageRetrieve(query, preFilters),
+      await this.textToImageRetrieve(query, preFilters as MetadataFilters),
    );
    this.sendEvent(query, nodesWithScores, parentEvent);
    return nodesWithScores;
@@ -1,7 +1,6 @@
 import type OpenAILLM from "openai";
 import type { ClientOptions as OpenAIClientOptions } from "openai";
 import type {
-  AnthropicStreamToken,
  CallbackManager,
  Event,
  EventType,
@@ -13,11 +12,7 @@ import type { ChatCompletionMessageParam } from "openai/resources/index.js";
 import type { LLMOptions } from "portkey-ai";
 import { Tokenizers, globalsHelper } from "../GlobalsHelper.js";
 import type { AnthropicSession } from "./anthropic.js";
-import {
-  ANTHROPIC_AI_PROMPT,
-  ANTHROPIC_HUMAN_PROMPT,
-  getAnthropicSession,
-} from "./anthropic.js";
+import { getAnthropicSession } from "./anthropic.js";
 import type { AzureOpenAIConfig } from "./azure.js";
 import {
  getAzureBaseUrl,
@@ -260,7 +255,7 @@ export class OpenAI extends BaseLLM {
      stream: false,
    });

-    const content = response.choices[0].message?.content ?? "";
+    const content = response.choices[0].message?.content ?? null;

    const kwargsOutput: Record<string, any> = {};

@@ -613,12 +608,30 @@ If a question does not make any sense, or is not factually coherent, explain why
  }
 }

-export const ALL_AVAILABLE_ANTHROPIC_MODELS = {
-  // both models have 100k context window, see https://docs.anthropic.com/claude/reference/selecting-a-model
-  "claude-2": { contextWindow: 200000 },
-  "claude-instant-1": { contextWindow: 100000 },
+export const ALL_AVAILABLE_ANTHROPIC_LEGACY_MODELS = {
+  "claude-2.1": {
+    contextWindow: 200000,
+  },
+  "claude-instant-1.2": {
+    contextWindow: 100000,
+  },
 };

+export const ALL_AVAILABLE_V3_MODELS = {
+  "claude-3-opus": { contextWindow: 200000 },
+  "claude-3-sonnet": { contextWindow: 200000 },
+};
+
+export const ALL_AVAILABLE_ANTHROPIC_MODELS = {
+  ...ALL_AVAILABLE_ANTHROPIC_LEGACY_MODELS,
+  ...ALL_AVAILABLE_V3_MODELS,
+};
+
+const AVAILABLE_ANTHROPIC_MODELS_WITHOUT_DATE: { [key: string]: string } = {
+  "claude-3-opus": "claude-3-opus-20240229",
+  "claude-3-sonnet": "claude-3-sonnet-20240229",
+} as { [key in keyof typeof ALL_AVAILABLE_ANTHROPIC_MODELS]: string };
+
 /**
 * Anthropic LLM implementation
 */
@@ -640,7 +653,7 @@ export class Anthropic extends BaseLLM {

  constructor(init?: Partial<Anthropic>) {
    super();
-    this.model = init?.model ?? "claude-2";
+    this.model = init?.model ?? "claude-3-opus";
    this.temperature = init?.temperature ?? 0.1;
    this.topP = init?.topP ?? 0.999; // Per Ben Mann
    this.maxTokens = init?.maxTokens ?? undefined;
@@ -674,21 +687,24 @@ export class Anthropic extends BaseLLM {
    };
  }

-  mapMessagesToPrompt(messages: ChatMessage[]) {
-    return (
-      messages.reduce((acc, message) => {
-        return (
-          acc +
-          `${
-            message.role === "system"
-              ? ""
-              : message.role === "assistant"
-                ? ANTHROPIC_AI_PROMPT + " "
-                : ANTHROPIC_HUMAN_PROMPT + " "
-          }${message.content.trim()}`
-        );
-      }, "") + ANTHROPIC_AI_PROMPT
-    );
+  getModelName = (model: string): string => {
+    if (Object.keys(AVAILABLE_ANTHROPIC_MODELS_WITHOUT_DATE).includes(model)) {
+      return AVAILABLE_ANTHROPIC_MODELS_WITHOUT_DATE[model];
+    }
+    return model;
+  };
+
+  formatMessages(messages: ChatMessage[]) {
+    return messages.map((message) => {
+      if (message.role !== "user" && message.role !== "assistant") {
+        throw new Error("Unsupported Anthropic role");
+      }
+
+      return {
+        content: message.content,
+        role: message.role,
+      };
+    });
  }

  chat(
@@ -698,49 +714,67 @@ export class Anthropic extends BaseLLM {
  async chat(
    params: LLMChatParamsNonStreaming | LLMChatParamsStreaming,
  ): Promise<ChatResponse | AsyncIterable<ChatResponseChunk>> {
-    const { messages, parentEvent, stream } = params;
+    let { messages } = params;
+
+    const { parentEvent, stream } = params;
+
+    let systemPrompt: string | null = null;
+
+    const systemMessages = messages.filter(
+      (message) => message.role === "system",
+    );
+
+    if (systemMessages.length > 0) {
+      systemPrompt = systemMessages
+        .map((message) => message.content)
+        .join("\n");
+      messages = messages.filter((message) => message.role !== "system");
+    }
+
    //Streaming
    if (stream) {
-      return this.streamChat(messages, parentEvent);
+      return this.streamChat(messages, parentEvent, systemPrompt);
    }

    //Non-streaming
-    const response = await this.session.anthropic.completions.create({
-      model: this.model,
-      prompt: this.mapMessagesToPrompt(messages),
-      max_tokens_to_sample: this.maxTokens ?? 100000,
+    const response = await this.session.anthropic.messages.create({
+      model: this.getModelName(this.model),
+      messages: this.formatMessages(messages),
+      max_tokens: this.maxTokens ?? 4096,
      temperature: this.temperature,
      top_p: this.topP,
+      ...(systemPrompt && { system: systemPrompt }),
    });

    return {
-      message: { content: response.completion.trimStart(), role: "assistant" },
-      //^ We're trimming the start because Anthropic often starts with a space in the response
-      // That space will be re-added when we generate the next prompt.
+      message: { content: response.content[0].text, role: "assistant" },
    };
  }

  protected async *streamChat(
    messages: ChatMessage[],
    parentEvent?: Event | undefined,
+    systemPrompt?: string | null,
  ): AsyncIterable<ChatResponseChunk> {
-    // AsyncIterable<AnthropicStreamToken>
-    const stream: AsyncIterable<AnthropicStreamToken> =
-      await this.session.anthropic.completions.create({
-        model: this.model,
-        prompt: this.mapMessagesToPrompt(messages),
-        max_tokens_to_sample: this.maxTokens ?? 100000,
-        temperature: this.temperature,
-        top_p: this.topP,
-        stream: true,
-      });
+    const stream = await this.session.anthropic.messages.create({
+      model: this.getModelName(this.model),
+      messages: this.formatMessages(messages),
+      max_tokens: this.maxTokens ?? 4096,
+      temperature: this.temperature,
+      top_p: this.topP,
+      stream: true,
+      ...(systemPrompt && { system: systemPrompt }),
+    });

    let idx_counter: number = 0;
    for await (const part of stream) {
-      //TODO: LLM Stream Callback, pending re-work.
+      const content =
+        part.type === "content_block_delta" ? part.delta.text : null;
+
+      if (typeof content !== "string") continue;

      idx_counter++;
-      yield { delta: part.completion };
+      yield { delta: content };
    }
    return;
  }
@@ -1,5 +1,6 @@
 import type { ClientOptions } from "@anthropic-ai/sdk";
 import Anthropic, { AI_PROMPT, HUMAN_PROMPT } from "@anthropic-ai/sdk";
+import { getEnv } from "@llamaindex/env";
 import _ from "lodash";

 export class AnthropicSession {
@@ -7,9 +8,7 @@ export class AnthropicSession {

  constructor(options: ClientOptions = {}) {
    if (!options.apiKey) {
-      if (typeof process !== undefined) {
-        options.apiKey = process.env.ANTHROPIC_API_KEY;
-      }
+      options.apiKey = getEnv("ANTHROPIC_API_KEY");
    }

    if (!options.apiKey) {
@@ -1,3 +1,5 @@
+import { getEnv } from "@llamaindex/env";
+
 export interface AzureOpenAIConfig {
  apiKey?: string;
  endpoint?: string;
@@ -67,24 +69,24 @@ export function getAzureConfigFromEnv(
  return {
    apiKey:
      init?.apiKey ??
-      process.env.AZURE_OPENAI_KEY ?? // From Azure docs
-      process.env.OPENAI_API_KEY ?? // Python compatible
-      process.env.AZURE_OPENAI_API_KEY, // LCJS compatible
+      getEnv("AZURE_OPENAI_KEY") ?? // From Azure docs
+      getEnv("OPENAI_API_KEY") ?? // Python compatible
+      getEnv("AZURE_OPENAI_API_KEY"), // LCJS compatible
    endpoint:
      init?.endpoint ??
-      process.env.AZURE_OPENAI_ENDPOINT ?? // From Azure docs
-      process.env.OPENAI_API_BASE ?? // Python compatible
-      process.env.AZURE_OPENAI_API_INSTANCE_NAME, // LCJS compatible
+      getEnv("AZURE_OPENAI_ENDPOINT") ?? // From Azure docs
+      getEnv("OPENAI_API_BASE") ?? // Python compatible
+      getEnv("AZURE_OPENAI_API_INSTANCE_NAME"), // LCJS compatible
    apiVersion:
      init?.apiVersion ??
-      process.env.AZURE_OPENAI_API_VERSION ?? // From Azure docs
-      process.env.OPENAI_API_VERSION ?? // Python compatible
-      process.env.AZURE_OPENAI_API_VERSION ?? // LCJS compatible
+      getEnv("AZURE_OPENAI_API_VERSION") ?? // From Azure docs
+      getEnv("OPENAI_API_VERSION") ?? // Python compatible
+      getEnv("AZURE_OPENAI_API_VERSION") ?? // LCJS compatible
      DEFAULT_API_VERSION,
    deploymentName:
      init?.deploymentName ??
-      process.env.AZURE_OPENAI_DEPLOYMENT ?? // From Azure docs
-      process.env.AZURE_OPENAI_API_DEPLOYMENT_NAME ?? // LCJS compatible
+      getEnv("AZURE_OPENAI_DEPLOYMENT") ?? // From Azure docs
+      getEnv("AZURE_OPENAI_API_DEPLOYMENT_NAME") ?? // LCJS compatible
      init?.model, // Fall back to model name, Python compatible
  };
 }
@@ -113,8 +115,8 @@ export function getAzureModel(openAIModel: string) {

 export function shouldUseAzure() {
  return (
-    process.env.AZURE_OPENAI_ENDPOINT ||
-    process.env.AZURE_OPENAI_API_INSTANCE_NAME ||
-    process.env.OPENAI_API_TYPE === "azure"
+    getEnv("AZURE_OPENAI_ENDPOINT") ||
+    getEnv("AZURE_OPENAI_API_INSTANCE_NAME") ||
+    getEnv("OPENAI_API_TYPE") === "azure"
  );
 }
@@ -1,9 +1,10 @@
+import { getEnv } from "@llamaindex/env";
 import { OpenAI } from "./LLM.js";

 export class FireworksLLM extends OpenAI {
  constructor(init?: Partial<OpenAI>) {
    const {
-      apiKey = process.env.FIREWORKS_API_KEY,
+      apiKey = getEnv("FIREWORKS_API_KEY"),
      additionalSessionOptions = {},
      model = "accounts/fireworks/models/mixtral-8x7b-instruct",
      ...rest
@@ -1,9 +1,10 @@
+import { getEnv } from "@llamaindex/env";
 import { OpenAI } from "./LLM.js";

 export class Groq extends OpenAI {
  constructor(init?: Partial<OpenAI>) {
    const {
-      apiKey = process.env.GROQ_API_KEY,
+      apiKey = getEnv("GROQ_API_KEY"),
      additionalSessionOptions = {},
      model = "mixtral-8x7b-32768",
      ...rest
@@ -1,3 +1,4 @@
+import { getEnv } from "@llamaindex/env";
 import type {
  CallbackManager,
  Event,
@@ -27,9 +28,7 @@ export class MistralAISession {
    if (init?.apiKey) {
      this.apiKey = init?.apiKey;
    } else {
-      if (typeof process !== undefined) {
-        this.apiKey = process.env.MISTRAL_API_KEY;
-      }
+      this.apiKey = getEnv("MISTRAL_API_KEY");
    }
    if (!this.apiKey) {
      throw new Error("Set Mistral API key in MISTRAL_API_KEY env variable"); // Overriding MistralAI package's error message
@@ -1,3 +1,4 @@
+import { getEnv } from "@llamaindex/env";
 import _ from "lodash";
 import type { ClientOptions } from "openai";
 import OpenAI from "openai";
@@ -13,9 +14,7 @@ export class OpenAISession {

  constructor(options: ClientOptions & { azure?: boolean } = {}) {
    if (!options.apiKey) {
-      if (typeof process !== undefined) {
-        options.apiKey = process.env.OPENAI_API_KEY;
-      }
+      options.apiKey = getEnv("OPENAI_API_KEY");
    }

    if (!options.apiKey) {
@@ -1,17 +1,8 @@
+import { getEnv } from "@llamaindex/env";
 import _ from "lodash";
 import type { LLMOptions } from "portkey-ai";
 import { Portkey } from "portkey-ai";

-export const readEnv = (
-  env: string,
-  default_val?: string,
-): string | undefined => {
-  if (typeof process !== "undefined") {
-    return process.env?.[env] ?? default_val;
-  }
-  return default_val;
-};
-
 interface PortkeyOptions {
  apiKey?: string;
  baseURL?: string;
@@ -24,11 +15,11 @@ export class PortkeySession {

  constructor(options: PortkeyOptions = {}) {
    if (!options.apiKey) {
-      options.apiKey = readEnv("PORTKEY_API_KEY");
+      options.apiKey = getEnv("PORTKEY_API_KEY");
    }

    if (!options.baseURL) {
-      options.baseURL = readEnv("PORTKEY_BASE_URL", "https://api.portkey.ai");
+      options.baseURL = getEnv("PORTKEY_BASE_URL") ?? "https://api.portkey.ai";
    }

    this.portkey = new Portkey({});
@@ -1,3 +1,4 @@
+import { getEnv } from "@llamaindex/env";
 import Replicate from "replicate";

 export class ReplicateSession {
@@ -7,8 +8,8 @@ export class ReplicateSession {
  constructor(replicateKey: string | null = null) {
    if (replicateKey) {
      this.replicateKey = replicateKey;
-    } else if (process.env.REPLICATE_API_TOKEN) {
-      this.replicateKey = process.env.REPLICATE_API_TOKEN;
+    } else if (getEnv("REPLICATE_API_TOKEN")) {
+      this.replicateKey = getEnv("REPLICATE_API_TOKEN") as string;
    } else {
      throw new Error(
        "Set Replicate token in REPLICATE_API_TOKEN env variable",
@@ -1,9 +1,10 @@
+import { getEnv } from "@llamaindex/env";
 import { OpenAI } from "./LLM.js";

 export class TogetherLLM extends OpenAI {
  constructor(init?: Partial<OpenAI>) {
    const {
-      apiKey = process.env.TOGETHER_API_KEY,
+      apiKey = getEnv("TOGETHER_API_KEY"),
      additionalSessionOptions = {},
      model = "togethercomputer/llama-2-7b-chat",
      ...rest
@@ -69,7 +69,7 @@ export class ObjectRetriever {

  // Translating the retrieve method
  async retrieve(strOrQueryBundle: QueryType): Promise<any> {
-    const nodes = await this.retriever.retrieve(strOrQueryBundle);
+    const nodes = await this.retriever.retrieve({ query: strOrQueryBundle });
    const objs = nodes.map((n) => this._objectNodeMapping.fromNode(n.node));
    return objs;
  }
@@ -1,3 +1,4 @@
+import { getEnv } from "@llamaindex/env";
 import type {
  BaseServiceParams,
  SubtitleFormat,
@@ -28,7 +29,7 @@ abstract class AssemblyAIReader implements BaseReader {
      options = {};
    }
    if (!options.apiKey) {
-      options.apiKey = process.env.ASSEMBLYAI_API_KEY;
+      options.apiKey = getEnv("ASSEMBLYAI_API_KEY");
    }
    if (!options.apiKey) {
      throw new Error(
@@ -1,5 +1,4 @@
-import type { GenericFileSystem } from "@llamaindex/env";
-import { defaultFS } from "@llamaindex/env";
+import { defaultFS, getEnv, type GenericFileSystem } from "@llamaindex/env";
 import { Document } from "../Node.js";
 import type { FileReader } from "./type.js";

@@ -24,7 +23,7 @@ export class LlamaParseReader implements FileReader {

  constructor(params: Partial<LlamaParseReader> = {}) {
    Object.assign(this, params);
-    params.apiKey = params.apiKey ?? process.env.LLAMA_CLOUD_API_KEY;
+    params.apiKey = params.apiKey ?? getEnv("LLAMA_CLOUD_API_KEY");
    if (!params.apiKey) {
      throw new Error(
        "API Key is required for LlamaParseReader. Please pass the apiKey parameter or set the LLAMA_CLOUD_API_KEY environment variable.",
@@ -1,4 +1,4 @@
-import _, * as lodash from "lodash";
+import _ from "lodash";
 import type { BaseNode } from "../../Node.js";
 import { ObjectType } from "../../Node.js";
 import { DEFAULT_NAMESPACE } from "../constants.js";
@@ -123,10 +123,10 @@ export class KVDocumentStore extends BaseDocumentStore {

    const refDocInfo = await this.kvstore.get(refDocId, this.refDocCollection);
    if (!_.isNil(refDocInfo)) {
-      lodash.pull(refDocInfo.docIds, docId);
+      !_.pull(refDocInfo.nodeIds, docId);

-      if (refDocInfo.docIds.length > 0) {
-        this.kvstore.put(refDocId, refDocInfo.toDict(), this.refDocCollection);
+      if (refDocInfo.nodeIds.length > 0) {
+        this.kvstore.put(refDocId, refDocInfo, this.refDocCollection);
      }
      this.kvstore.delete(refDocId, this.metadataCollection);
    }
@@ -54,6 +54,9 @@ export class SimpleKVStore extends BaseKVStore {
  ): Promise<boolean> {
    if (key in this.data[collection]) {
      delete this.data[collection][key];
+      if (this.persistPath) {
+        await this.persist(this.persistPath, this.fs);
+      }
      return true;
    }
    return false;
@@ -1,5 +1,6 @@
 import { AstraDB } from "@datastax/astra-db-ts";
 import type { Collection } from "@datastax/astra-db-ts/dist/collections";
+import { getEnv } from "@llamaindex/env";
 import type { BaseNode } from "../../Node.js";
 import { MetadataMode } from "../../Node.js";
 import type {
@@ -34,9 +35,8 @@ export class AstraDBVectorStore implements VectorStore {
    if (init?.astraDBClient) {
      this.astraDBClient = init.astraDBClient;
    } else {
-      const token =
-        init?.params?.token ?? process.env.ASTRA_DB_APPLICATION_TOKEN;
-      const endpoint = init?.params?.endpoint ?? process.env.ASTRA_DB_ENDPOINT;
+      const token = init?.params?.token ?? getEnv("ASTRA_DB_APPLICATION_TOKEN");
+      const endpoint = init?.params?.endpoint ?? getEnv("ASTRA_DB_ENDPOINT");

      if (!token) {
        throw new Error(
@@ -48,7 +48,7 @@ export class AstraDBVectorStore implements VectorStore {
      }
      const namespace =
        init?.params?.namespace ??
-        process.env.ASTRA_DB_NAMESPACE ??
+        getEnv("ASTRA_DB_NAMESPACE") ??
        "default_keyspace";
      this.astraDBClient = new AstraDB(token, endpoint, namespace);
    }
@@ -1,3 +1,4 @@
+import { getEnv } from "@llamaindex/env";
 import type { BulkWriteOptions, Collection } from "mongodb";
 import { MongoClient } from "mongodb";
 import type { BaseNode } from "../../Node.js";
@@ -44,7 +45,7 @@ export class MongoDBAtlasVectorSearch implements VectorStore {
    if (init.mongodbClient) {
      this.mongodbClient = init.mongodbClient;
    } else {
-      const mongoUri = process.env.MONGODB_URI;
+      const mongoUri = getEnv("MONGODB_URI");
      if (!mongoUri) {
        throw new Error(
          "Must specify MONGODB_URI via env variable if not directly passing in client.",
@@ -6,7 +6,7 @@ import type {
  VectorStoreQueryResult,
 } from "./types.js";

-import type { GenericFileSystem } from "@llamaindex/env";
+import { getEnv, type GenericFileSystem } from "@llamaindex/env";
 import type {
  FetchResponse,
  Index,
@@ -45,11 +45,11 @@ export class PineconeVectorStore implements VectorStore {

  constructor(params?: PineconeParams) {
    this.indexName =
-      params?.indexName ?? process.env.PINECONE_INDEX_NAME ?? "llama";
-    this.namespace = params?.namespace ?? process.env.PINECONE_NAMESPACE ?? "";
+      params?.indexName ?? getEnv("PINECONE_INDEX_NAME") ?? "llama";
+    this.namespace = params?.namespace ?? getEnv("PINECONE_NAMESPACE") ?? "";
    this.chunkSize =
      params?.chunkSize ??
-      Number.parseInt(process.env.PINECONE_CHUNK_SIZE ?? "100");
+      Number.parseInt(getEnv("PINECONE_CHUNK_SIZE") ?? "100");
    this.textKey = params?.textKey ?? "text";
  }

@@ -82,6 +82,9 @@ export class SimpleVectorStore implements VectorStore {
      delete this.data.embeddingDict[textId];
      delete this.data.textIdToRefDocId[textId];
    }
+    if (this.persistPath) {
+      await this.persist(this.persistPath, this.fs);
+    }
    return Promise.resolve();
  }

--- a/Show More
+++ b/Show More
Author	SHA1	Message	Date
Marcus Schiesser	4a36de7397	docs: updated docs for retriever using param obj	2024-03-06 14:30:20 +07:00
Marcus Schiesser	1f1ee1eb8e	refactor: Use parameter object for retrieve function of Retriever	2024-03-06 14:19:42 +07:00
Wojciech Grzebieniowski	484a7105a9	fix: restore missing exports (#610 )	2024-03-05 14:56:25 -06:00
Alex Yang	8d18ea167b	fix: publish.yml	2024-03-05 14:37:54 -06:00
Alex Yang	a2ca89bfe0	fix: config (#611 )	2024-03-05 14:20:58 -06:00
Alex Yang	edeea40898	ci: add publish.yml	2024-03-05 13:49:49 -06:00
Alex Yang	2a7080b094	build: fix version	2024-03-05 12:26:47 -06:00
Huu Le (Lee)	b354f2386b	feat: add embedding model option to create-llama (#608 )	2024-03-05 16:59:51 +07:00
Emanuel Ferreira	d766bd03d2	feat: OpenAI Agent Stream (#597 )	2024-03-05 15:46:44 +07:00
Huu Le (Lee)	6a69148356	fix: add --no-llama-parse and improve e2e test (#607 )	2024-03-05 14:57:38 +07:00
Marcus Schiesser	e1e1b0b522	RELEASING: Releasing 1 package(s) Releases: llamaindex@0.1.21 [skip ci]	2024-03-05 13:00:16 +07:00
Marcus Schiesser	d824876653	docs(changeset): Add support for Claude 3	2024-03-05 12:59:51 +07:00
Marcus Schiesser	2048698f77	docs: add interactive chat for anthropic	2024-03-05 11:15:50 +07:00
Emanuel Ferreira	9942979aa7	feat: Claude 3 (#604 )	2024-03-04 15:02:18 -08:00
Alex Yang	3c2655a1f9	fix: .tsbuildinfo	2024-03-04 16:05:45 -06:00
Marcus Schiesser	552a61a66f	Add quantized parameter to HuggingFaceEmbedding (#601 )	2024-03-04 12:10:40 +07:00
Alex Yang	d13143e322	RELEASING: Releasing 3 package(s) Releases: llamaindex@0.1.20 @llamaindex/env@0.0.5 docs@0.0.4 [skip ci]	2024-03-02 18:42:24 -06:00
Alex Yang	5116ad8d08	fix: compatibility issue with Deno (#598 )	2024-03-02 18:40:01 -06:00
Emanuel Ferreira	64683a55f3	fix: prefix messages always true (#596 )	2024-03-01 21:45:02 -03:00
Emanuel Ferreira	698cd9c631	fix: step wise agent + examples (#594 )	2024-03-01 21:28:02 -03:00
Alex Yang	c744a99102	chore: bump `@llamaindex/cloud` (#595 )	2024-03-01 17:22:50 -06:00
Huu Le (Lee)	2d2935085e	feat: Add use LlamaParse option to create-llama (#591 )	2024-03-01 16:54:06 +07:00
Marcus Schiesser	1b31e2c8cd	chore: update @llamaindex/cloud to 0.0.2	2024-03-01 15:59:10 +07:00
Thuc Pham	7257751993	fix: empty store bugs (#592 ) Co-authored-by: Marcus Schiesser <mail@marcusschiesser.de>	2024-03-01 15:04:10 +07:00