docs(changeset): feat: add base evaluator and correctness evaluator

Releases:
  llamaindex@0.1.16
  @llamaindex/env@0.0.3

[skip ci]

.changeset/famous-pugs-join.md

@@ -0,0 +1,5 @@
+---
+"llamaindex": patch
+---
+
+feat: add base evaluator and correctness evaluator

.changeset/healthy-llamas-rest.md

@@ -1,5 +0,0 @@
----
-"create-llama": patch
----
-
-feat: generate llama pack project from llama index

.changeset/pink-tools-look.md

@@ -1,5 +0,0 @@
----
-"create-llama": patch
----
-
-feat: add pinecone support to create llama

.changeset/sharp-ducks-clap.md

@@ -0,0 +1,5 @@
+---
+"llamaindex": patch
+---
+
+feat: add base evaluator and correctness evaluator

.changeset/swift-waves-divide.md

@@ -0,0 +1,6 @@
+---
+"llamaindex": patch
+"docs": patch
+---
+
+Add Groq LLM to LlamaIndex

.changeset/ten-pears-study.md

@@ -1,5 +0,0 @@
----
-"create-llama": patch
----
-
-update fastapi for CVE-2024-24762

.eslintrc.js

@@ -9,6 +9,7 @@ module.exports = {
   },
   rules: {
     "max-params": ["error", 4],
+    "prefer-const": "error",
   },
-  ignorePatterns: ["dist/"],
+  ignorePatterns: ["dist/", "lib/"],
 };

.github/workflows/test.yml

@@ -61,10 +61,13 @@ jobs:
         run: pnpm run build --filter llamaindex
       - name: Copy examples
         run: rsync -rv --exclude=node_modules ./examples ${{ runner.temp }}
-      - name: Pack
+      - name: Pack @llamaindex/env
+        run: pnpm pack --pack-destination ${{ runner.temp }}
+        working-directory: packages/env
+      - name: Pack llamaindex
         run: pnpm pack --pack-destination ${{ runner.temp }}
         working-directory: packages/core
-      - name: Install llamaindex
+      - name: Install
         run: npm add ${{ runner.temp }}/*.tgz
         working-directory: ${{ runner.temp }}/examples
       - name: Run Type Check

.vscode/settings.json

@@ -5,7 +5,6 @@
   "[xml]": {
     "editor.defaultFormatter": "redhat.vscode-xml"
   },
-  "jest.rootPath": "./packages/core",
   "[python]": {
     "editor.defaultFormatter": "ms-python.black-formatter"
   },

README.md

@@ -125,6 +125,7 @@ module.exports = nextConfig;
 
 - OpenAI GPT-3.5-turbo and GPT-4
 - Anthropic Claude Instant and Claude 2
+- Groq LLMs
 - Llama2 Chat LLMs (70B, 13B, and 7B parameters)
 - MistralAI Chat LLMs
 - Fireworks Chat LLMs

apps/docs/docs/modules/evaluation/_category_.yml

@@ -0,0 +1,2 @@
+label: "Evaluating"
+position: 3

apps/docs/docs/modules/evaluation/index.md

@@ -0,0 +1,32 @@
+# Evaluating
+
+## Concept
+
+Evaluation and benchmarking are crucial concepts in LLM development. To improve the perfomance of an LLM app (RAG, agents) you must have a way to measure it.
+
+LlamaIndex offers key modules to measure the quality of generated results. We also offer key modules to measure retrieval quality.
+
+- **Response Evaluation**: Does the response match the retrieved context? Does it also match the query? Does it match the reference answer or guidelines?
+- **Retrieval Evaluation**: Are the retrieved sources relevant to the query?
+
+## Response Evaluation
+
+Evaluation of generated results can be difficult, since unlike traditional machine learning the predicted result is not a single number, and it can be hard to define quantitative metrics for this problem.
+
+LlamaIndex offers LLM-based evaluation modules to measure the quality of results. This uses a “gold” LLM (e.g. GPT-4) to decide whether the predicted answer is correct in a variety of ways.
+
+Note that many of these current evaluation modules do not require ground-truth labels. Evaluation can be done with some combination of the query, context, response, and combine these with LLM calls.
+
+These evaluation modules are in the following forms:
+
+- **Correctness**: Whether the generated answer matches that of the reference answer given the query (requires labels).
+
+- **Faithfulness**: Evaluates if the answer is faithful to the retrieved contexts (in other words, whether if there’s hallucination).
+
+- **Relevancy**: Evaluates if the response from a query engine matches any source nodes.
+
+## Usage
+
+- [Correctness Evaluator](correctness.md)
+- [Faithfulness Evaluator](faithfulness.md)
+- [Relevancy Evaluator](relevancy.md)

apps/docs/docs/modules/evaluation/modules/_category_.yml

@@ -0,0 +1 @@
+label: "Modules"

apps/docs/docs/modules/evaluation/modules/correctness.md

@@ -0,0 +1,72 @@
+# Correctness Evaluator
+
+Correctness evaluates the relevance and correctness of a generated answer against a reference answer.
+
+This is useful for measuring if the response was correct. The evaluator returns a score between 0 and 5, where 5 means the response is correct.
+
+## Usage
+
+Firstly, you need to install the package:
+
+```bash
+pnpm i llamaindex
+```
+
+Set the OpenAI API key:
+
+```bash
+export OPENAI_API_KEY=your-api-key
+```
+
+Import the required modules:
+
+```ts
+import {
+  CorrectnessEvaluator,
+  OpenAI,
+  serviceContextFromDefaults,
+} from "llamaindex";
+```
+
+Let's setup gpt-4 for better results:
+
+```ts
+const llm = new OpenAI({
+  model: "gpt-4",
+});
+
+const ctx = serviceContextFromDefaults({
+  llm,
+});
+```
+
+```ts
+const query =
+  "Can you explain the theory of relativity proposed by Albert Einstein in detail?";
+
+const response = ` Certainly! Albert Einstein's theory of relativity consists of two main components: special relativity and general relativity. Special relativity, published in 1905, introduced the concept that the laws of physics are the same for all non-accelerating observers and that the speed of light in a vacuum is a constant, regardless of the motion of the source or observer. It also gave rise to the famous equation E=mc², which relates energy (E) and mass (m).
+
+However, general relativity, published in 1915, extended these ideas to include the effects of magnetism. According to general relativity, gravity is not a force between masses but rather the result of the warping of space and time by magnetic fields generated by massive objects. Massive objects, such as planets and stars, create magnetic fields that cause a curvature in spacetime, and smaller objects follow curved paths in response to this magnetic curvature. This concept is often illustrated using the analogy of a heavy ball placed on a rubber sheet with magnets underneath, causing it to create a depression that other objects (representing smaller masses) naturally move towards due to magnetic attraction.
+`;
+
+const evaluator = new CorrectnessEvaluator({
+  serviceContext: ctx,
+});
+
+const response = await queryEngine.query({
+  query,
+});
+
+const result = await evaluator.evaluateResponse({
+  query,
+  response,
+});
+
+console.log(
+  `the response is ${result.passing ? "correct" : "not correct"} with a score of ${result.score}`,
+);
+```
+
+```bash
+the response is not correct with a score of 2.5
+```

apps/docs/docs/modules/evaluation/modules/faithfulness.md

@@ -0,0 +1,84 @@
+# Faithfulness Evaluator
+
+Faithfulness is a measure of whether the generated answer is faithful to the retrieved contexts. In other words, it measures whether there is any hallucination in the generated answer.
+
+This uses the FaithfulnessEvaluator module to measure if the response from a query engine matches any source nodes.
+
+This is useful for measuring if the response was hallucinated. The evaluator returns a score between 0 and 1, where 1 means the response is faithful to the retrieved contexts.
+
+## Usage
+
+Firstly, you need to install the package:
+
+```bash
+pnpm i llamaindex
+```
+
+Set the OpenAI API key:
+
+```bash
+export OPENAI_API_KEY=your-api-key
+```
+
+Import the required modules:
+
+```ts
+import {
+  Document,
+  FaithfulnessEvaluator,
+  OpenAI,
+  VectorStoreIndex,
+  serviceContextFromDefaults,
+} from "llamaindex";
+```
+
+Let's setup gpt-4 for better results:
+
+```ts
+const llm = new OpenAI({
+  model: "gpt-4",
+});
+
+const ctx = serviceContextFromDefaults({
+  llm,
+});
+```
+
+Now, let's create a vector index and query engine with documents and query engine respectively. Then, we can evaluate the response with the query and response from the query engine.:
+
+```ts
+const documents = [
+  new Document({
+    text: `The city came under British control in 1664 and was renamed New York after King Charles II of England granted the lands to his brother, the Duke of York. The city was regained by the Dutch in July 1673 and was renamed New Orange for one year and three months; the city has been continuously named New York since November 1674. New York City was the capital of the United States from 1785 until 1790, and has been the largest U.S. city since 1790. The Statue of Liberty greeted millions of immigrants as they came to the U.S. by ship in the late 19th and early 20th centuries, and is a symbol of the U.S. and its ideals of liberty and peace. In the 21st century, New York City has emerged as a global node of creativity, entrepreneurship, and as a symbol of freedom and cultural diversity. The New York Times has won the most Pulitzer Prizes for journalism and remains the U.S. media's "newspaper of record". In 2019, New York City was voted the greatest city in the world in a survey of over 30,000 p...	Pass`,
+  }),
+];
+
+const vectorIndex = await VectorStoreIndex.fromDocuments(documents);
+
+const queryEngine = vectorIndex.asQueryEngine();
+```
+
+Now, let's evaluate the response:
+
+```ts
+const query = "How did New York City get its name?";
+
+const evaluator = new FaithfulnessEvaluator({
+  serviceContext: ctx,
+});
+
+const response = await queryEngine.query({
+  query,
+});
+
+const result = await evaluator.evaluateResponse({
+  query,
+  response,
+});
+
+console.log(`the response is ${result.passing ? "faithful" : "not faithful"}`);
+```
+
+```bash
+the response is faithful
+```

apps/docs/docs/modules/evaluation/modules/relevancy.md

@@ -0,0 +1,72 @@
+# Relevancy Evaluator
+
+Relevancy measure if the response from a query engine matches any source nodes.
+
+It is useful for measuring if the response was relevant to the query. The evaluator returns a score between 0 and 1, where 1 means the response is relevant to the query.
+
+## Usage
+
+Firstly, you need to install the package:
+
+```bash
+pnpm i llamaindex
+```
+
+Set the OpenAI API key:
+
+```bash
+export OPENAI_API_KEY=your-api-key
+```
+
+Import the required modules:
+
+```ts
+import {
+  RelevancyEvaluator,
+  OpenAI,
+  serviceContextFromDefaults,
+} from "llamaindex";
+```
+
+Let's setup gpt-4 for better results:
+
+```ts
+const llm = new OpenAI({
+  model: "gpt-4",
+});
+
+const ctx = serviceContextFromDefaults({
+  llm,
+});
+```
+
+Now, let's create a vector index and query engine with documents and query engine respectively. Then, we can evaluate the response with the query and response from the query engine.:
+
+```ts
+const documents = [
+  new Document({
+    text: `The city came under British control in 1664 and was renamed New York after King Charles II of England granted the lands to his brother, the Duke of York. The city was regained by the Dutch in July 1673 and was renamed New Orange for one year and three months; the city has been continuously named New York since November 1674. New York City was the capital of the United States from 1785 until 1790, and has been the largest U.S. city since 1790. The Statue of Liberty greeted millions of immigrants as they came to the U.S. by ship in the late 19th and early 20th centuries, and is a symbol of the U.S. and its ideals of liberty and peace. In the 21st century, New York City has emerged as a global node of creativity, entrepreneurship, and as a symbol of freedom and cultural diversity. The New York Times has won the most Pulitzer Prizes for journalism and remains the U.S. media's "newspaper of record". In 2019, New York City was voted the greatest city in the world in a survey of over 30,000 p...	Pass`,
+  }),
+];
+
+const vectorIndex = await VectorStoreIndex.fromDocuments(documents);
+
+const queryEngine = vectorIndex.asQueryEngine();
+
+const query = "How did New York City get its name?";
+
+const response = await queryEngine.query({
+  query,
+});
+
+const result = await evaluator.evaluateResponse({
+  query,
+  response: response,
+});
+
+console.log(`the response is ${result.passing ? "relevant" : "not relevant"}`);
+```
+
+```bash
+the response is relevant
+```

apps/docs/docs/modules/llms/available_llms/groq.mdx

@@ -0,0 +1,56 @@
+import CodeBlock from "@theme/CodeBlock";
+import CodeSource from "!raw-loader!../../../../../../examples/groq.ts";
+
+# Groq
+
+## Usage
+
+First, create an API key at the [Groq Console](https://console.groq.com/keys). Then save it in your environment:
+
+```bash
+export GROQ_API_KEY=<your-api-key>
+```
+
+The initialize the Groq module.
+
+```ts
+import { Groq, serviceContextFromDefaults } from "llamaindex";
+
+const groq = new Groq({
+  // If you do not wish to set your API key in the environment, you may
+  // configure your API key when you initialize the Groq class.
+  // apiKey: "<your-api-key>",
+});
+
+const serviceContext = serviceContextFromDefaults({ llm: groq });
+```
+
+## Load and index documents
+
+For this example, we will use a single document. In a real-world scenario, you would have multiple documents to index.
+
+```ts
+const document = new Document({ text: essay, id_: "essay" });
+
+const index = await VectorStoreIndex.fromDocuments([document], {
+  serviceContext,
+});
+```
+
+## Query
+
+```ts
+const queryEngine = index.asQueryEngine();
+
+const query = "What is the meaning of life?";
+
+const results = await queryEngine.query({
+  query,
+});
+```
+
+## Full Example
+
+<CodeBlock language="ts" showLineNumbers>
+  {CodeSource}
+</CodeBlock>

examples/evaluation/correctness.ts

@@ -0,0 +1,36 @@
+import {
+  CorrectnessEvaluator,
+  OpenAI,
+  serviceContextFromDefaults,
+} from "llamaindex";
+
+async function main() {
+  const llm = new OpenAI({
+    model: "gpt-4",
+  });
+
+  const ctx = serviceContextFromDefaults({
+    llm,
+  });
+
+  const evaluator = new CorrectnessEvaluator({
+    serviceContext: ctx,
+  });
+
+  const query =
+    "Can you explain the theory of relativity proposed by Albert Einstein in detail?";
+
+  const response = `
+Certainly! Albert Einstein's theory of relativity consists of two main components: special relativity and general relativity. Special relativity, published in 1905, introduced the concept that the laws of physics are the same for all non-accelerating observers and that the speed of light in a vacuum is a constant, regardless of the motion of the source or observer. It also gave rise to the famous equation E=mc², which relates energy (E) and mass (m).
+However, general relativity, published in 1915, extended these ideas to include the effects of magnetism. According to general relativity, gravity is not a force between masses but rather the result of the warping of space and time by magnetic fields generated by massive objects. Massive objects, such as planets and stars, create magnetic fields that cause a curvature in spacetime, and smaller objects follow curved paths in response to this magnetic curvature. This concept is often illustrated using the analogy of a heavy ball placed on a rubber sheet with magnets underneath, causing it to create a depression that other objects (representing smaller masses) naturally move towards due to magnetic attraction.
+`;
+
+  const result = await evaluator.evaluate({
+    query: query,
+    response: response,
+  });
+
+  console.log(result);
+}
+
+main();

examples/evaluation/faithfulness.ts

@@ -0,0 +1,46 @@
+import {
+  Document,
+  FaithfulnessEvaluator,
+  OpenAI,
+  VectorStoreIndex,
+  serviceContextFromDefaults,
+} from "llamaindex";
+
+async function main() {
+  const llm = new OpenAI({
+    model: "gpt-4",
+  });
+
+  const ctx = serviceContextFromDefaults({
+    llm,
+  });
+
+  const evaluator = new FaithfulnessEvaluator({
+    serviceContext: ctx,
+  });
+
+  const documents = [
+    new Document({
+      text: `The city came under British control in 1664 and was renamed New York after King Charles II of England granted the lands to his brother, the Duke of York. The city was regained by the Dutch in July 1673 and was renamed New Orange for one year and three months; the city has been continuously named New York since November 1674. New York City was the capital of the United States from 1785 until 1790, and has been the largest U.S. city since 1790. The Statue of Liberty greeted millions of immigrants as they came to the U.S. by ship in the late 19th and early 20th centuries, and is a symbol of the U.S. and its ideals of liberty and peace. In the 21st century, New York City has emerged as a global node of creativity, entrepreneurship, and as a symbol of freedom and cultural diversity. The New York Times has won the most Pulitzer Prizes for journalism and remains the U.S. media's "newspaper of record". In 2019, New York City was voted the greatest city in the world in a survey of over 30,000 p...	Pass`,
+    }),
+  ];
+
+  const vectorIndex = await VectorStoreIndex.fromDocuments(documents);
+
+  const queryEngine = vectorIndex.asQueryEngine();
+
+  const query = "How did New York City get its name?";
+
+  const response = await queryEngine.query({
+    query,
+  });
+
+  const result = await evaluator.evaluateResponse({
+    query,
+    response,
+  });
+
+  console.log(result);
+}
+
+main();

examples/evaluation/relevancy.ts

@@ -0,0 +1,46 @@
+import {
+  Document,
+  OpenAI,
+  RelevancyEvaluator,
+  VectorStoreIndex,
+  serviceContextFromDefaults,
+} from "llamaindex";
+
+async function main() {
+  const llm = new OpenAI({
+    model: "gpt-4",
+  });
+
+  const ctx = serviceContextFromDefaults({
+    llm,
+  });
+
+  const evaluator = new RelevancyEvaluator({
+    serviceContext: ctx,
+  });
+
+  const documents = [
+    new Document({
+      text: `The city came under British control in 1664 and was renamed New York after King Charles II of England granted the lands to his brother, the Duke of York. The city was regained by the Dutch in July 1673 and was renamed New Orange for one year and three months; the city has been continuously named New York since November 1674. New York City was the capital of the United States from 1785 until 1790, and has been the largest U.S. city since 1790. The Statue of Liberty greeted millions of immigrants as they came to the U.S. by ship in the late 19th and early 20th centuries, and is a symbol of the U.S. and its ideals of liberty and peace. In the 21st century, New York City has emerged as a global node of creativity, entrepreneurship, and as a symbol of freedom and cultural diversity. The New York Times has won the most Pulitzer Prizes for journalism and remains the U.S. media's "newspaper of record". In 2019, New York City was voted the greatest city in the world in a survey of over 30,000 p...	Pass`,
+    }),
+  ];
+
+  const vectorIndex = await VectorStoreIndex.fromDocuments(documents);
+
+  const queryEngine = vectorIndex.asQueryEngine();
+
+  const query = "How did New York City get its name?";
+
+  const response = await queryEngine.query({
+    query,
+  });
+
+  const result = await evaluator.evaluateResponse({
+    query,
+    response: response,
+  });
+
+  console.log(result);
+}
+
+main();

examples/groq.ts

@@ -0,0 +1,48 @@
+import fs from "node:fs/promises";
+
+import {
+  Document,
+  Groq,
+  VectorStoreIndex,
+  serviceContextFromDefaults,
+} from "llamaindex";
+
+async function main() {
+  // Create an instance of the LLM
+  const groq = new Groq({
+    apiKey: process.env.GROQ_API_KEY,
+  });
+
+  // Create a service context
+  const serviceContext = serviceContextFromDefaults({ llm: groq });
+
+  // Load essay from abramov.txt in Node
+  const path = "node_modules/llamaindex/examples/abramov.txt";
+  const essay = await fs.readFile(path, "utf-8");
+  const document = new Document({ text: essay, id_: "essay" });
+
+  // Load and index documents
+  const index = await VectorStoreIndex.fromDocuments([document], {
+    serviceContext,
+  });
+
+  // get retriever
+  const retriever = index.asRetriever();
+
+  // Create a query engine
+  const queryEngine = index.asQueryEngine({
+    retriever,
+  });
+
+  const query = "What is the meaning of life?";
+
+  // Query
+  const response = await queryEngine.query({
+    query,
+  });
+
+  // Log the response
+  console.log(response.response);
+}
+
+await main();

examples/ollama.ts

@@ -1,4 +1,4 @@
-import { Ollama } from "llamaindex";
+import { Ollama } from "llamaindex/llm/ollama";
 
 (async () => {
   const llm = new Ollama({ model: "llama2", temperature: 0.75 });

package.json

@@ -18,15 +18,12 @@
   },
   "devDependencies": {
     "@changesets/cli": "^2.27.1",
-    "@types/jest": "^29.5.12",
     "eslint": "^8.56.0",
     "eslint-config-custom": "workspace:*",
     "husky": "^9.0.10",
-    "jest": "^29.7.0",
     "lint-staged": "^15.2.2",
     "prettier": "^3.2.5",
     "prettier-plugin-organize-imports": "^3.2.4",
-    "ts-jest": "^29.1.2",
     "turbo": "^1.12.3",
     "typescript": "^5.3.3"
   },

packages/core/.cjs.swcrc

@@ -0,0 +1,11 @@
+{
+  "jsc": {
+    "parser": {
+      "syntax": "typescript"
+    },
+    "target": "esnext"
+  },
+  "module": {
+    "type": "commonjs"
+  }
+}

packages/core/.swcrc

@@ -0,0 +1,8 @@
+{
+  "jsc": {
+    "parser": {
+      "syntax": "typescript"
+    },
+    "target": "esnext"
+  }
+}

packages/core/CHANGELOG.md

@@ -1,5 +1,34 @@
 # llamaindex
 
+## 0.1.16
+
+### Patch Changes
+
+- e8e21a0: build: set files in package.json
+- Updated dependencies [e8e21a0]
+  - @llamaindex/env@0.0.3
+
+## 0.1.15
+
+### Patch Changes
+
+- 3a6e287: build: improve tree-shake & reduce unused package import
+
+## 0.1.14
+
+### Patch Changes
+
+- 7416a87: build: cjs file not found
+- Updated dependencies [7416a87]
+  - @llamaindex/env@0.0.2
+
+## 0.1.13
+
+### Patch Changes
+
+- b8be4c0: build: use ESM as default
+- 65d8346: feat: abstract `@llamaindex/env` package
+
 ## 0.1.12
 
 ### Patch Changes

packages/core/jest.config.cjs

@@ -1,6 +0,0 @@
-/** @type {import('ts-jest').JestConfigWithTsJest} */
-module.exports = {
-  preset: "ts-jest",
-  testEnvironment: "node",
-  testPathIgnorePatterns: ["/lib/", "/node_modules/", "/dist/"],
-};

packages/core/package.json

@@ -1,11 +1,14 @@
 {
   "name": "llamaindex",
-  "private": true,
-  "version": "0.1.12",
+  "version": "0.1.16",
   "license": "MIT",
+  "type": "module",
   "dependencies": {
     "@anthropic-ai/sdk": "^0.13.0",
+    "@aws-crypto/sha256-js": "^5.2.0",
     "@datastax/astra-db-ts": "^0.1.4",
+    "@llamaindex/cloud": "^0.0.1",
+    "@llamaindex/env": "workspace:*",
     "@mistralai/mistralai": "^0.0.10",
     "@notionhq/client": "^2.2.14",
     "@pinecone-database/pinecone": "^2.0.1",
@@ -34,175 +37,52 @@
     "wink-nlp": "^1.14.3"
   },
   "devDependencies": {
-    "@aws-crypto/sha256-js": "^5.2.0",
-    "@llamaindex/cloud": "^0.0.1",
-    "@types/edit-json-file": "^1.7.3",
-    "@types/jest": "^29.5.12",
+    "@swc/cli": "^0.3.9",
+    "@swc/core": "^1.4.2",
     "@types/lodash": "^4.14.202",
     "@types/node": "^18.19.14",
     "@types/papaparse": "^5.3.14",
     "@types/pg": "^8.11.0",
-    "bunchee": "^4.4.6",
-    "edit-json-file": "^1.8.0",
+    "concurrently": "^8.2.2",
+    "glob": "^10.3.10",
     "madge": "^6.1.0",
     "typescript": "^5.3.3"
   },
   "engines": {
     "node": ">=18.0.0"
   },
-  "types": "./dist/index.d.ts",
-  "main": "./dist/index.js",
+  "types": "./dist/type/index.d.ts",
+  "main": "./dist/cjs/index.js",
   "exports": {
     ".": {
-      "types": "./dist/index.d.mts",
-      "import": "./dist/index.mjs",
-      "edge-light": "./dist/index.edge-light.mjs",
-      "require": "./dist/index.js"
+      "import": {
+        "types": "./dist/type/index.d.ts",
+        "default": "./dist/index.js"
+      },
+      "edge-light": {
+        "types": "./dist/type/index.d.ts",
+        "default": "./dist/index.edge-light.js"
+      },
+      "require": {
+        "types": "./dist/type/index.d.ts",
+        "default": "./dist/cjs/index.js"
+      }
     },
-    "./env": {
-      "types": "./dist/env.d.mts",
-      "import": "./dist/env.mjs",
-      "edge-light": "./dist/env.edge-light.mjs",
-      "require": "./dist/env.js"
-    },
-    "./ChatEngine": {
-      "types": "./dist/ChatEngine.d.mts",
-      "import": "./dist/ChatEngine.mjs",
-      "require": "./dist/ChatEngine.js"
-    },
-    "./ChatHistory": {
-      "types": "./dist/ChatHistory.d.mts",
-      "import": "./dist/ChatHistory.mjs",
-      "require": "./dist/ChatHistory.js"
-    },
-    "./constants": {
-      "types": "./dist/constants.d.mts",
-      "import": "./dist/constants.mjs",
-      "require": "./dist/constants.js"
-    },
-    "./GlobalsHelper": {
-      "types": "./dist/GlobalsHelper.d.mts",
-      "import": "./dist/GlobalsHelper.mjs",
-      "require": "./dist/GlobalsHelper.js"
-    },
-    "./Node": {
-      "types": "./dist/Node.d.mts",
-      "import": "./dist/Node.mjs",
-      "require": "./dist/Node.js"
-    },
-    "./OutputParser": {
-      "types": "./dist/OutputParser.d.mts",
-      "import": "./dist/OutputParser.mjs",
-      "require": "./dist/OutputParser.js"
-    },
-    "./Prompt": {
-      "types": "./dist/Prompt.d.mts",
-      "import": "./dist/Prompt.mjs",
-      "require": "./dist/Prompt.js"
-    },
-    "./PromptHelper": {
-      "types": "./dist/PromptHelper.d.mts",
-      "import": "./dist/PromptHelper.mjs",
-      "require": "./dist/PromptHelper.js"
-    },
-    "./QueryEngine": {
-      "types": "./dist/QueryEngine.d.mts",
-      "import": "./dist/QueryEngine.mjs",
-      "require": "./dist/QueryEngine.js"
-    },
-    "./QuestionGenerator": {
-      "types": "./dist/QuestionGenerator.d.mts",
-      "import": "./dist/QuestionGenerator.mjs",
-      "require": "./dist/QuestionGenerator.js"
-    },
-    "./Response": {
-      "types": "./dist/Response.d.mts",
-      "import": "./dist/Response.mjs",
-      "require": "./dist/Response.js"
-    },
-    "./ServiceContext": {
-      "types": "./dist/ServiceContext.d.mts",
-      "import": "./dist/ServiceContext.mjs",
-      "require": "./dist/ServiceContext.js"
-    },
-    "./TextSplitter": {
-      "types": "./dist/TextSplitter.d.mts",
-      "import": "./dist/TextSplitter.mjs",
-      "require": "./dist/TextSplitter.js"
-    },
-    "./tools": {
-      "types": "./dist/tools.d.mts",
-      "import": "./dist/tools.mjs",
-      "require": "./dist/tools.js"
-    },
-    "./objects": {
-      "types": "./dist/objects.d.mts",
-      "import": "./dist/objects.mjs",
-      "require": "./dist/objects.js"
-    },
-    "./readers": {
-      "types": "./dist/readers.d.mts",
-      "import": "./dist/readers.mjs",
-      "require": "./dist/readers.js"
-    },
-    "./readers/AssemblyAIReader": {
-      "types": "./dist/readers/AssemblyAIReader.d.mts",
-      "import": "./dist/readers/AssemblyAIReader.mjs",
-      "require": "./dist/readers/AssemblyAIReader.js"
-    },
-    "./readers/CSVReader": {
-      "types": "./dist/readers/CSVReader.d.mts",
-      "import": "./dist/readers/CSVReader.mjs",
-      "require": "./dist/readers/CSVReader.js"
-    },
-    "./readers/DocxReader": {
-      "types": "./dist/readers/DocxReader.d.mts",
-      "import": "./dist/readers/DocxReader.mjs",
-      "require": "./dist/readers/DocxReader.js"
-    },
-    "./readers/HTMLReader": {
-      "types": "./dist/readers/HTMLReader.d.mts",
-      "import": "./dist/readers/HTMLReader.mjs",
-      "require": "./dist/readers/HTMLReader.js"
-    },
-    "./readers/ImageReader": {
-      "types": "./dist/readers/ImageReader.d.mts",
-      "import": "./dist/readers/ImageReader.mjs",
-      "require": "./dist/readers/ImageReader.js"
-    },
-    "./readers/MarkdownReader": {
-      "types": "./dist/readers/MarkdownReader.d.mts",
-      "import": "./dist/readers/MarkdownReader.mjs",
-      "require": "./dist/readers/MarkdownReader.js"
-    },
-    "./readers/NotionReader": {
-      "types": "./dist/readers/NotionReader.d.mts",
-      "import": "./dist/readers/NotionReader.mjs",
-      "require": "./dist/readers/NotionReader.js"
-    },
-    "./readers/PDFReader": {
-      "types": "./dist/readers/PDFReader.d.mts",
-      "import": "./dist/readers/PDFReader.mjs",
-      "require": "./dist/readers/PDFReader.js"
-    },
-    "./readers/SimpleDirectoryReader": {
-      "types": "./dist/readers/SimpleDirectoryReader.d.mts",
-      "import": "./dist/readers/SimpleDirectoryReader.mjs",
-      "require": "./dist/readers/SimpleDirectoryReader.js"
-    },
-    "./readers/SimpleMongoReader": {
-      "types": "./dist/readers/SimpleMongoReader.d.mts",
-      "import": "./dist/readers/SimpleMongoReader.mjs",
-      "require": "./dist/readers/SimpleMongoReader.js"
-    },
-    "./cloud": {
-      "types": "./dist/cloud.d.mts",
-      "import": "./dist/cloud.mjs",
-      "require": "./dist/cloud.js"
+    "./*": {
+      "import": {
+        "types": "./dist/type/*.d.ts",
+        "default": "./dist/*.js"
+      },
+      "require": {
+        "types": "./dist/type/*.d.ts",
+        "default": "./dist/cjs/*.js"
+      }
     }
   },
   "files": [
-    "**"
+    "dist",
+    "CHANGELOG.md",
+    "examples"
   ],
   "repository": {
     "type": "git",
@@ -211,13 +91,12 @@
   },
   "scripts": {
     "lint": "eslint .",
-    "test": "jest",
-    "build": "rm -rf ./dist && NODE_OPTIONS=\"--max-old-space-size=12288\" bunchee",
-    "postbuild": "pnpm run copy && pnpm run modify-package-json",
-    "copy": "cp -r package.json CHANGELOG.md ../../README.md ../../LICENSE examples src dist/",
-    "modify-package-json": "node ./scripts/modify-package-json.mjs",
-    "prepublish": "pnpm run modify-package-json && echo \"please cd ./dist and run pnpm publish\" && exit 1",
-    "dev": "NODE_OPTIONS=\"--max-old-space-size=16384\" bunchee -w",
-    "circular-check": "madge -c ./src/index.ts"
+    "build": "rm -rf ./dist && pnpm run build:esm && pnpm run build:cjs && pnpm run build:type",
+    "build:esm": "swc src -d dist --strip-leading-paths --config-file .swcrc",
+    "build:cjs": "swc src -d dist/cjs --strip-leading-paths --config-file .cjs.swcrc",
+    "build:type": "tsc -p tsconfig.json",
+    "postbuild": "node -e \"require('fs').writeFileSync('./dist/cjs/package.json', JSON.stringify({ type: 'commonjs' }))\"",
+    "circular-check": "madge -c ./src/index.ts",
+    "dev": "concurrently \"pnpm run build:esm --watch\" \"pnpm run build:cjs --watch\" \"pnpm run build:type --watch\""
   }
 }

packages/core/scripts/modify-package-json.mjs

@@ -1,25 +0,0 @@
-#!/usr/bin/env node
-/**
- * This script is used to modify the package.json file in the dist folder
- * so that it can be published to npm.
- */
-import editJsonFile from "edit-json-file";
-import fs from "node:fs/promises";
-
-{
-  await fs.copyFile("./package.json", "./dist/package.json");
-  const file = editJsonFile("./dist/package.json");
-
-  file.unset("scripts");
-  file.unset("private");
-  await new Promise((resolve) => file.save(resolve));
-}
-{
-  const packageJson = await fs.readFile("./dist/package.json", "utf8");
-  const modifiedPackageJson = packageJson.replaceAll("./dist/", "./");
-  await fs.writeFile(
-    "./dist/package.json",
-    JSON.stringify(JSON.parse(modifiedPackageJson), null, 2),
-    "utf8",
-  );
-}

packages/core/src/ChatHistory.ts

@@ -1,10 +1,7 @@
-import { OpenAI } from "./llm/LLM";
-import { ChatMessage, LLM, MessageType } from "./llm/types";
-import {
-  defaultSummaryPrompt,
-  messagesToHistoryStr,
-  SummaryPrompt,
-} from "./Prompt";
+import { OpenAI } from "./llm/LLM.js";
+import type { ChatMessage, LLM, MessageType } from "./llm/types.js";
+import type { SummaryPrompt } from "./Prompt.js";
+import { defaultSummaryPrompt, messagesToHistoryStr } from "./Prompt.js";
 
 /**
  * A ChatHistory is used to keep the state of back and forth chat messages

packages/core/src/GlobalsHelper.ts

@@ -1,7 +1,11 @@
 import { encodingForModel } from "js-tiktoken";
 
-import { Event, EventTag, EventType } from "./callbacks/CallbackManager";
-import { randomUUID } from "./env";
+import { randomUUID } from "@llamaindex/env";
+import type {
+  Event,
+  EventTag,
+  EventType,
+} from "./callbacks/CallbackManager.js";
 
 export enum Tokenizers {
   CL100K_BASE = "cl100k_base",
@@ -32,7 +36,7 @@ class GlobalsHelper {
     };
   }
 
-  tokenizer(encoding?: string) {
+  tokenizer(encoding?: Tokenizers) {
     if (encoding && encoding !== Tokenizers.CL100K_BASE) {
       throw new Error(`Tokenizer encoding ${encoding} not yet supported`);
     }
@@ -43,7 +47,7 @@ class GlobalsHelper {
     return this.defaultTokenizer!.encode.bind(this.defaultTokenizer);
   }
 
-  tokenizerDecoder(encoding?: string) {
+  tokenizerDecoder(encoding?: Tokenizers) {
     if (encoding && encoding !== Tokenizers.CL100K_BASE) {
       throw new Error(`Tokenizer encoding ${encoding} not yet supported`);
     }

packages/core/src/Node.ts

@@ -1,5 +1,5 @@
+import { createSHA256, path, randomUUID } from "@llamaindex/env";
 import _ from "lodash";
-import { createSHA256, path, randomUUID } from "./env";
 
 export enum NodeRelationship {
   SOURCE = "SOURCE",
@@ -65,7 +65,8 @@ export abstract class BaseNode<T extends Metadata = Metadata> {
 
   abstract getContent(metadataMode: MetadataMode): string;
   abstract getMetadataStr(metadataMode: MetadataMode): string;
-  abstract setContent(value: any): void;
+  // todo: set value as a generic type
+  abstract setContent(value: unknown): void;
 
   get sourceNode(): RelatedNodeInfo<T> | undefined {
     const relationship = this.relationships[NodeRelationship.SOURCE];
@@ -353,10 +354,10 @@ export function splitNodesByType(nodes: BaseNode[]): {
   imageNodes: ImageNode[];
   textNodes: TextNode[];
 } {
-  let imageNodes: ImageNode[] = [];
-  let textNodes: TextNode[] = [];
+  const imageNodes: ImageNode[] = [];
+  const textNodes: TextNode[] = [];
 
-  for (let node of nodes) {
+  for (const node of nodes) {
     if (node instanceof ImageNode) {
       imageNodes.push(node);
     } else if (node instanceof TextNode) {

packages/core/src/OutputParser.ts

@@ -1,5 +1,5 @@
-import { SubQuestion } from "./engines/query/types";
-import { BaseOutputParser, StructuredOutput } from "./types";
+import type { SubQuestion } from "./engines/query/types.js";
+import type { BaseOutputParser, StructuredOutput } from "./types.js";
 
 /**
  * Error class for output parsing. Due to the nature of LLMs, anytime we use LLM
@@ -44,8 +44,8 @@ export function parseJsonMarkdown(text: string) {
   const left_square = text.indexOf("[");
   const left_brace = text.indexOf("{");
 
-  var left: number;
-  var right: number;
+  let left: number;
+  let right: number;
   if (left_square < left_brace && left_square != -1) {
     left = left_square;
     right = text.lastIndexOf("]");

packages/core/src/Prompt.ts

@@ -1,6 +1,6 @@
-import { SubQuestion } from "./engines/query/types";
-import { ChatMessage } from "./llm/types";
-import { ToolMetadata } from "./types";
+import type { SubQuestion } from "./engines/query/types.js";
+import type { ChatMessage } from "./llm/types.js";
+import type { ToolMetadata } from "./types.js";
 
 /**
  * A SimplePrompt is a function that takes a dictionary of inputs and returns a string.

packages/core/src/PromptHelper.ts

@@ -1,12 +1,12 @@
-import { globalsHelper } from "./GlobalsHelper";
-import { SimplePrompt } from "./Prompt";
-import { SentenceSplitter } from "./TextSplitter";
+import { globalsHelper } from "./GlobalsHelper.js";
+import type { SimplePrompt } from "./Prompt.js";
+import { SentenceSplitter } from "./TextSplitter.js";
 import {
   DEFAULT_CHUNK_OVERLAP_RATIO,
   DEFAULT_CONTEXT_WINDOW,
   DEFAULT_NUM_OUTPUTS,
   DEFAULT_PADDING,
-} from "./constants";
+} from "./constants.js";
 
 export function getEmptyPromptTxt(prompt: SimplePrompt) {
   return prompt({});

packages/core/src/QuestionGenerator.ts

@@ -1,14 +1,18 @@
-import { SubQuestionOutputParser } from "./OutputParser";
-import {
-  SubQuestionPrompt,
-  buildToolsText,
-  defaultSubQuestionPrompt,
-} from "./Prompt";
-import { BaseQuestionGenerator, SubQuestion } from "./engines/query/types";
-import { OpenAI } from "./llm/LLM";
-import { LLM } from "./llm/types";
-import { PromptMixin } from "./prompts";
-import { BaseOutputParser, StructuredOutput, ToolMetadata } from "./types";
+import { SubQuestionOutputParser } from "./OutputParser.js";
+import type { SubQuestionPrompt } from "./Prompt.js";
+import { buildToolsText, defaultSubQuestionPrompt } from "./Prompt.js";
+import type {
+  BaseQuestionGenerator,
+  SubQuestion,
+} from "./engines/query/types.js";
+import { OpenAI } from "./llm/LLM.js";
+import type { LLM } from "./llm/types.js";
+import { PromptMixin } from "./prompts/index.js";
+import type {
+  BaseOutputParser,
+  StructuredOutput,
+  ToolMetadata,
+} from "./types.js";
 
 /**
  * LLMQuestionGenerator uses the LLM to generate new questions for the LLM using tools and a user query.

packages/core/src/Response.ts

@@ -1,4 +1,4 @@
-import { BaseNode } from "./Node";
+import type { BaseNode } from "./Node.js";
 
 /**
  * Response is the output of a LLM

packages/core/src/Retriever.ts

@@ -1,6 +1,6 @@
-import { Event } from "./callbacks/CallbackManager";
-import { NodeWithScore } from "./Node";
-import { ServiceContext } from "./ServiceContext";
+import type { Event } from "./callbacks/CallbackManager.js";
+import type { NodeWithScore } from "./Node.js";
+import type { ServiceContext } from "./ServiceContext.js";
 
 /**
  * Retrievers retrieve the nodes that most closely match our query in similarity.

packages/core/src/ServiceContext.ts

@@ -1,8 +1,11 @@
-import { CallbackManager } from "./callbacks/CallbackManager";
-import { BaseEmbedding, OpenAIEmbedding } from "./embeddings";
-import { LLM, OpenAI } from "./llm";
-import { NodeParser, SimpleNodeParser } from "./nodeParsers";
-import { PromptHelper } from "./PromptHelper";
+import { PromptHelper } from "./PromptHelper.js";
+import { CallbackManager } from "./callbacks/CallbackManager.js";
+import { OpenAIEmbedding } from "./embeddings/OpenAIEmbedding.js";
+import type { BaseEmbedding } from "./embeddings/types.js";
+import type { LLM } from "./llm/index.js";
+import { OpenAI } from "./llm/index.js";
+import { SimpleNodeParser } from "./nodeParsers/SimpleNodeParser.js";
+import type { NodeParser } from "./nodeParsers/types.js";
 
 /**
  * The ServiceContext is a collection of components that are used in different parts of the application.

packages/core/src/TextSplitter.ts

@@ -1,7 +1,7 @@
-import { EOL } from "./env";
+import { EOL } from "@llamaindex/env";
 // GitHub translated
-import { globalsHelper } from "./GlobalsHelper";
-import { DEFAULT_CHUNK_OVERLAP, DEFAULT_CHUNK_SIZE } from "./constants";
+import { globalsHelper } from "./GlobalsHelper.js";
+import { DEFAULT_CHUNK_OVERLAP, DEFAULT_CHUNK_SIZE } from "./constants.js";
 
 class TextSplit {
   textChunk: string;
@@ -130,7 +130,7 @@ export class SentenceSplitter {
 
   getParagraphSplits(text: string, effectiveChunkSize?: number): string[] {
     // get paragraph splits
-    let paragraphSplits: string[] = text.split(this.paragraphSeparator);
+    const paragraphSplits: string[] = text.split(this.paragraphSeparator);
     let idx = 0;
     if (effectiveChunkSize == undefined) {
       return paragraphSplits;
@@ -155,9 +155,9 @@ export class SentenceSplitter {
   }
 
   getSentenceSplits(text: string, effectiveChunkSize?: number): string[] {
-    let paragraphSplits = this.getParagraphSplits(text, effectiveChunkSize);
+    const paragraphSplits = this.getParagraphSplits(text, effectiveChunkSize);
     // Next we split the text using the chunk tokenizer fn/
-    let splits = [];
+    const splits = [];
     for (const parText of paragraphSplits) {
       const sentenceSplits = this.chunkingTokenizerFn(parText);
 
@@ -194,9 +194,9 @@ export class SentenceSplitter {
       }));
     }
 
-    let newSplits: SplitRep[] = [];
+    const newSplits: SplitRep[] = [];
     for (const split of sentenceSplits) {
-      let splitTokens = this.tokenizer(split);
+      const splitTokens = this.tokenizer(split);
       const splitLen = splitTokens.length;
       if (splitLen <= effectiveChunkSize) {
         newSplits.push({ text: split, numTokens: splitLen });
@@ -219,7 +219,7 @@ export class SentenceSplitter {
     // go through sentence splits, combine to chunks that are within the chunk size
 
     // docs represents final list of text chunks
-    let docs: TextSplit[] = [];
+    const docs: TextSplit[] = [];
     // curChunkSentences represents the current list of sentence splits (that)
     // will be merged into a chunk
     let curChunkSentences: SplitRep[] = [];
@@ -287,18 +287,18 @@ export class SentenceSplitter {
       return [];
     }
 
-    let effectiveChunkSize = this.getEffectiveChunkSize(extraInfoStr);
-    let sentenceSplits = this.getSentenceSplits(text, effectiveChunkSize);
+    const effectiveChunkSize = this.getEffectiveChunkSize(extraInfoStr);
+    const sentenceSplits = this.getSentenceSplits(text, effectiveChunkSize);
 
     // Check if any sentences exceed the chunk size. If they don't,
     // force split by tokenizer
-    let newSentenceSplits = this.processSentenceSplits(
+    const newSentenceSplits = this.processSentenceSplits(
       sentenceSplits,
       effectiveChunkSize,
     );
 
     // combine sentence splits into chunks of text that can then be returned
-    let combinedTextSplits = this.combineTextSplits(
+    const combinedTextSplits = this.combineTextSplits(
       newSentenceSplits,
       effectiveChunkSize,
     );

packages/core/src/agent/index.ts

@@ -1,5 +1,5 @@
-export * from "./openai/base";
-export * from "./openai/worker";
-export * from "./react/base";
-export * from "./react/worker";
-export * from "./types";
+export * from "./openai/base.js";
+export * from "./openai/worker.js";
+export * from "./react/base.js";
+export * from "./react/worker.js";
+export * from "./types.js";

packages/core/src/agent/openai/base.ts

@@ -1,9 +1,10 @@
-import { CallbackManager } from "../../callbacks/CallbackManager";
-import { ChatMessage, OpenAI } from "../../llm";
-import { ObjectRetriever } from "../../objects/base";
-import { BaseTool } from "../../types";
-import { AgentRunner } from "../runner/base";
-import { OpenAIAgentWorker } from "./worker";
+import type { CallbackManager } from "../../callbacks/CallbackManager.js";
+import type { ChatMessage } from "../../llm/index.js";
+import { OpenAI } from "../../llm/index.js";
+import type { ObjectRetriever } from "../../objects/base.js";
+import type { BaseTool } from "../../types.js";
+import { AgentRunner } from "../runner/base.js";
+import { OpenAIAgentWorker } from "./worker.js";
 
 type OpenAIAgentParams = {
   tools?: BaseTool[];

packages/core/src/agent/openai/utils.ts

@@ -1,4 +1,4 @@
-import { ToolMetadata } from "../../types";
+import type { ToolMetadata } from "../../types.js";
 
 export type OpenAIFunction = {
   type: "function";

packages/core/src/agent/openai/worker.ts

@@ -1,23 +1,27 @@
 // Assuming that the necessary interfaces and classes (like BaseTool, OpenAI, ChatMessage, CallbackManager, etc.) are defined elsewhere
 
-import { CallbackManager } from "../../callbacks/CallbackManager";
-import { AgentChatResponse, ChatResponseMode } from "../../engines/chat";
-import { randomUUID } from "../../env";
+import { randomUUID } from "@llamaindex/env";
+import type { CallbackManager } from "../../callbacks/CallbackManager.js";
 import {
+  AgentChatResponse,
+  ChatResponseMode,
+} from "../../engines/chat/types.js";
+import type {
   ChatMessage,
   ChatResponse,
   ChatResponseChunk,
-  OpenAI,
-} from "../../llm";
-import { ChatMemoryBuffer } from "../../memory/ChatMemoryBuffer";
-import { ObjectRetriever } from "../../objects/base";
-import { ToolOutput } from "../../tools/types";
-import { callToolWithErrorHandling } from "../../tools/utils";
-import { BaseTool } from "../../types";
-import { AgentWorker, Task, TaskStep, TaskStepOutput } from "../types";
-import { addUserStepToMemory, getFunctionByName } from "../utils";
-import { OpenAIToolCall } from "./types/chat";
-import { toOpenAiTool } from "./utils";
+} from "../../llm/index.js";
+import { OpenAI } from "../../llm/index.js";
+import { ChatMemoryBuffer } from "../../memory/ChatMemoryBuffer.js";
+import type { ObjectRetriever } from "../../objects/base.js";
+import type { ToolOutput } from "../../tools/types.js";
+import { callToolWithErrorHandling } from "../../tools/utils.js";
+import type { BaseTool } from "../../types.js";
+import type { AgentWorker, Task } from "../types.js";
+import { TaskStep, TaskStepOutput } from "../types.js";
+import { addUserStepToMemory, getFunctionByName } from "../utils.js";
+import type { OpenAIToolCall } from "./types/chat.js";
+import { toOpenAiTool } from "./utils.js";
 
 const DEFAULT_MAX_FUNCTION_CALLS = 5;
 

packages/core/src/agent/react/base.ts

@@ -1,9 +1,9 @@
-import { CallbackManager } from "../../callbacks/CallbackManager";
-import { ChatMessage, LLM } from "../../llm";
-import { ObjectRetriever } from "../../objects/base";
-import { BaseTool } from "../../types";
-import { AgentRunner } from "../runner/base";
-import { ReActAgentWorker } from "./worker";
+import type { CallbackManager } from "../../callbacks/CallbackManager.js";
+import type { ChatMessage, LLM } from "../../llm/index.js";
+import type { ObjectRetriever } from "../../objects/base.js";
+import type { BaseTool } from "../../types.js";
+import { AgentRunner } from "../runner/base.js";
+import { ReActAgentWorker } from "./worker.js";
 
 type ReActAgentParams = {
   tools: BaseTool[];

packages/core/src/agent/react/formatter.ts

@@ -1,7 +1,8 @@
-import { ChatMessage } from "../../llm";
-import { BaseTool } from "../../types";
-import { getReactChatSystemHeader } from "./prompts";
-import { BaseReasoningStep, ObservationReasoningStep } from "./types";
+import type { ChatMessage } from "../../llm/index.js";
+import type { BaseTool } from "../../types.js";
+import { getReactChatSystemHeader } from "./prompts.js";
+import type { BaseReasoningStep } from "./types.js";
+import { ObservationReasoningStep } from "./types.js";
 
 function getReactToolDescriptions(tools: BaseTool[]): string[] {
   const toolDescs: string[] = [];

packages/core/src/agent/react/outputParser.ts

@@ -1,9 +1,9 @@
+import type { BaseReasoningStep } from "./types.js";
 import {
   ActionReasoningStep,
   BaseOutputParser,
-  BaseReasoningStep,
   ResponseReasoningStep,
-} from "./types";
+} from "./types.js";
 
 function extractJsonStr(text: string): string {
   const pattern = /\{.*\}/s;

packages/core/src/agent/react/types.ts

@@ -1,4 +1,4 @@
-import { ChatMessage } from "../../llm";
+import type { ChatMessage } from "../../llm/index.js";
 
 export interface BaseReasoningStep {
   getContent(): string;

packages/core/src/agent/react/worker.ts

@@ -1,20 +1,22 @@
 import { randomUUID } from "crypto";
-import { CallbackManager } from "../../callbacks/CallbackManager";
-import { AgentChatResponse } from "../../engines/chat";
-import { ChatResponse, LLM, OpenAI } from "../../llm";
-import { ChatMemoryBuffer } from "../../memory/ChatMemoryBuffer";
-import { ObjectRetriever } from "../../objects/base";
-import { ToolOutput } from "../../tools";
-import { BaseTool } from "../../types";
-import { AgentWorker, Task, TaskStep, TaskStepOutput } from "../types";
-import { ReActChatFormatter } from "./formatter";
-import { ReActOutputParser } from "./outputParser";
+import { CallbackManager } from "../../callbacks/CallbackManager.js";
+import { AgentChatResponse } from "../../engines/chat/index.js";
+import type { ChatResponse, LLM } from "../../llm/index.js";
+import { OpenAI } from "../../llm/index.js";
+import { ChatMemoryBuffer } from "../../memory/ChatMemoryBuffer.js";
+import type { ObjectRetriever } from "../../objects/base.js";
+import { ToolOutput } from "../../tools/index.js";
+import type { BaseTool } from "../../types.js";
+import type { AgentWorker, Task } from "../types.js";
+import { TaskStep, TaskStepOutput } from "../types.js";
+import { ReActChatFormatter } from "./formatter.js";
+import { ReActOutputParser } from "./outputParser.js";
+import type { BaseReasoningStep } from "./types.js";
 import {
   ActionReasoningStep,
-  BaseReasoningStep,
   ObservationReasoningStep,
   ResponseReasoningStep,
-} from "./types";
+} from "./types.js";
 
 type ReActAgentWorkerParams = {
   tools: BaseTool[];

packages/core/src/agent/runner/base.ts

@@ -1,15 +1,16 @@
 import { randomUUID } from "crypto";
-import { CallbackManager } from "../../callbacks/CallbackManager";
+import { CallbackManager } from "../../callbacks/CallbackManager.js";
+import type { ChatEngineAgentParams } from "../../engines/chat/index.js";
 import {
   AgentChatResponse,
-  ChatEngineAgentParams,
   ChatResponseMode,
-} from "../../engines/chat";
-import { ChatMessage, LLM } from "../../llm";
-import { ChatMemoryBuffer } from "../../memory/ChatMemoryBuffer";
-import { BaseMemory } from "../../memory/types";
-import { AgentWorker, Task, TaskStep, TaskStepOutput } from "../types";
-import { AgentState, BaseAgentRunner, TaskState } from "./types";
+} from "../../engines/chat/index.js";
+import type { ChatMessage, LLM } from "../../llm/index.js";
+import { ChatMemoryBuffer } from "../../memory/ChatMemoryBuffer.js";
+import type { BaseMemory } from "../../memory/types.js";
+import type { AgentWorker, TaskStepOutput } from "../types.js";
+import { Task, TaskStep } from "../types.js";
+import { AgentState, BaseAgentRunner, TaskState } from "./types.js";
 
 const validateStepFromArgs = (
   taskId: string,

packages/core/src/agent/runner/types.ts

@@ -1,5 +1,6 @@
-import { AgentChatResponse } from "../../engines/chat";
-import { BaseAgent, Task, TaskStep, TaskStepOutput } from "../types";
+import type { AgentChatResponse } from "../../engines/chat/index.js";
+import type { Task, TaskStep, TaskStepOutput } from "../types.js";
+import { BaseAgent } from "../types.js";
 
 export class TaskState {
   task!: Task;

packages/core/src/agent/types.ts

@@ -1,5 +1,8 @@
-import { AgentChatResponse, ChatEngineAgentParams } from "../engines/chat";
-import { QueryEngineParamsNonStreaming } from "../types";
+import type {
+  AgentChatResponse,
+  ChatEngineAgentParams,
+} from "../engines/chat/index.js";
+import type { QueryEngineParamsNonStreaming } from "../types.js";
 
 export interface AgentWorker {
   initializeStep(task: Task, kwargs?: any): TaskStep;

packages/core/src/agent/utils.ts

@@ -1,7 +1,7 @@
-import { ChatMessage } from "../llm";
-import { ChatMemoryBuffer } from "../memory/ChatMemoryBuffer";
-import { BaseTool } from "../types";
-import { TaskStep } from "./types";
+import type { ChatMessage } from "../llm/index.js";
+import type { ChatMemoryBuffer } from "../memory/ChatMemoryBuffer.js";
+import type { BaseTool } from "../types.js";
+import type { TaskStep } from "./types.js";
 
 /**
  * Adds the user's input to the memory.

packages/core/src/callbacks/CallbackManager.ts

@@ -1,5 +1,5 @@
 import type { Anthropic } from "@anthropic-ai/sdk";
-import { NodeWithScore } from "../Node";
+import type { NodeWithScore } from "../Node.js";
 
 /*
   An event is a wrapper that groups related operations.

packages/core/src/cloud/LlamaCloudIndex.ts

@@ -1,10 +1,11 @@
-import { BaseRetriever } from "../Retriever";
-import { RetrieverQueryEngine } from "../engines/query/RetrieverQueryEngine";
-import { BaseNodePostprocessor } from "../postprocessors";
-import { BaseSynthesizer } from "../synthesizers";
-import { BaseQueryEngine } from "../types";
-import { LlamaCloudRetriever, RetrieveParams } from "./LlamaCloudRetriever";
-import { CloudConstructorParams } from "./types";
+import type { BaseRetriever } from "../Retriever.js";
+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 { LlamaCloudRetriever } from "./LlamaCloudRetriever.js";
+import type { CloudConstructorParams } from "./types.js";
 
 export class LlamaCloudIndex {
   params: CloudConstructorParams;
@@ -13,7 +14,7 @@ export class LlamaCloudIndex {
     this.params = params;
   }
 
-  asRetriever(params: RetrieveParams): BaseRetriever {
+  asRetriever(params: RetrieveParams = {}): BaseRetriever {
     return new LlamaCloudRetriever({ ...this.params, ...params });
   }
 

packages/core/src/cloud/LlamaCloudRetriever.ts

@@ -1,15 +1,14 @@
-import { PlatformApi, PlatformApiClient } from "@llamaindex/cloud";
-import { globalsHelper } from "../GlobalsHelper";
-import { NodeWithScore, ObjectType, jsonToNode } from "../Node";
-import { BaseRetriever } from "../Retriever";
-import { ServiceContext, serviceContextFromDefaults } from "../ServiceContext";
-import { Event } from "../callbacks/CallbackManager";
-import {
-  ClientParams,
-  CloudConstructorParams,
-  DEFAULT_PROJECT_NAME,
-} from "./types";
-import { getClient } from "./utils";
+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 { 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<
   PlatformApi.RetrievalParams,
@@ -38,8 +37,9 @@ export class LlamaCloudRetriever implements BaseRetriever {
 
   constructor(params: CloudConstructorParams & RetrieveParams) {
     this.clientParams = { apiKey: params.apiKey, baseUrl: params.baseUrl };
-    params.denseSimilarityTopK =
-      params.similarityTopK ?? params.denseSimilarityTopK;
+    if (params.similarityTopK) {
+      params.denseSimilarityTopK = params.similarityTopK;
+    }
     this.retrieveParams = params;
     this.pipelineName = params.name;
     if (params.projectName) {

packages/core/src/cloud/index.ts

@@ -1,2 +1,2 @@
-export * from "./LlamaCloudIndex";
-export * from "./LlamaCloudRetriever";
+export * from "./LlamaCloudIndex.js";
+export * from "./LlamaCloudRetriever.js";

packages/core/src/cloud/types.ts

@@ -1,4 +1,4 @@
-import { ServiceContext } from "../ServiceContext";
+import type { ServiceContext } from "../ServiceContext.js";
 
 export const DEFAULT_PROJECT_NAME = "default";
 export const DEFAULT_BASE_URL = "https://api.cloud.llamaindex.ai";

packages/core/src/cloud/utils.ts

@@ -1,5 +1,6 @@
-import { PlatformApiClient } from "@llamaindex/cloud";
-import { ClientParams, DEFAULT_BASE_URL } from "./types";
+import type { PlatformApiClient } from "@llamaindex/cloud";
+import type { ClientParams } from "./types.js";
+import { DEFAULT_BASE_URL } from "./types.js";
 
 export async function getClient({
   apiKey,

packages/core/src/embeddings/ClipEmbedding.ts

@@ -1,6 +1,6 @@
-import { ImageType } from "../Node";
-import { MultiModalEmbedding } from "./MultiModalEmbedding";
-import { readImage } from "./utils";
+import type { ImageType } from "../Node.js";
+import { MultiModalEmbedding } from "./MultiModalEmbedding.js";
+import { readImage } from "./utils.js";
 
 export enum ClipEmbeddingModelType {
   XENOVA_CLIP_VIT_BASE_PATCH32 = "Xenova/clip-vit-base-patch32",

packages/core/src/embeddings/HuggingFaceEmbedding.ts

@@ -1,4 +1,4 @@
-import { BaseEmbedding } from "./types";
+import { BaseEmbedding } from "./types.js";
 
 export enum HuggingFaceEmbeddingModelType {
   XENOVA_ALL_MINILM_L6_V2 = "Xenova/all-MiniLM-L6-v2",

packages/core/src/embeddings/MistralAIEmbedding.ts

@@ -1,5 +1,5 @@
-import { MistralAISession } from "../llm/mistral";
-import { BaseEmbedding } from "./types";
+import { MistralAISession } from "../llm/mistral.js";
+import { BaseEmbedding } from "./types.js";
 
 export enum MistralAIEmbeddingModelType {
   MISTRAL_EMBED = "mistral-embed",

packages/core/src/embeddings/MultiModalEmbedding.ts

@@ -1,5 +1,5 @@
-import { ImageType } from "../Node";
-import { BaseEmbedding } from "./types";
+import type { ImageType } from "../Node.js";
+import { BaseEmbedding } from "./types.js";
 
 /*
  * Base class for Multi Modal embeddings.

packages/core/src/embeddings/OllamaEmbedding.ts

@@ -1,5 +1,5 @@
-import { Ollama } from "../llm/ollama";
-import { BaseEmbedding } from "./types";
+import { Ollama } from "../llm/ollama.js";
+import type { BaseEmbedding } from "./types.js";
 
 /**
  * OllamaEmbedding is an alias for Ollama that implements the BaseEmbedding interface.

packages/core/src/embeddings/OpenAIEmbedding.ts

@@ -1,13 +1,14 @@
-import { ClientOptions as OpenAIClientOptions } from "openai";
+import type { ClientOptions as OpenAIClientOptions } from "openai";
+import type { AzureOpenAIConfig } from "../llm/azure.js";
 import {
-  AzureOpenAIConfig,
   getAzureBaseUrl,
   getAzureConfigFromEnv,
   getAzureModel,
   shouldUseAzure,
-} from "../llm/azure";
-import { OpenAISession, getOpenAISession } from "../llm/open_ai";
-import { BaseEmbedding } from "./types";
+} from "../llm/azure.js";
+import type { OpenAISession } from "../llm/open_ai.js";
+import { getOpenAISession } from "../llm/open_ai.js";
+import { BaseEmbedding } from "./types.js";
 
 export const ALL_OPENAI_EMBEDDING_MODELS = {
   "text-embedding-ada-002": {

packages/core/src/embeddings/fireworks.ts

@@ -1,4 +1,4 @@
-import { OpenAIEmbedding } from "./OpenAIEmbedding";
+import { OpenAIEmbedding } from "./OpenAIEmbedding.js";
 
 export class FireworksEmbedding extends OpenAIEmbedding {
   constructor(init?: Partial<OpenAIEmbedding>) {

packages/core/src/embeddings/index.ts

@@ -1,10 +1,10 @@
-export * from "./ClipEmbedding";
-export * from "./HuggingFaceEmbedding";
-export * from "./MistralAIEmbedding";
-export * from "./MultiModalEmbedding";
-export { OllamaEmbedding } from "./OllamaEmbedding";
-export * from "./OpenAIEmbedding";
-export { FireworksEmbedding } from "./fireworks";
-export { TogetherEmbedding } from "./together";
-export * from "./types";
-export * from "./utils";
+export * from "./ClipEmbedding.js";
+export * from "./HuggingFaceEmbedding.js";
+export * from "./MistralAIEmbedding.js";
+export * from "./MultiModalEmbedding.js";
+export { OllamaEmbedding } from "./OllamaEmbedding.js";
+export * from "./OpenAIEmbedding.js";
+export { FireworksEmbedding } from "./fireworks.js";
+export { TogetherEmbedding } from "./together.js";
+export * from "./types.js";
+export * from "./utils.js";

packages/core/src/embeddings/together.ts

@@ -1,4 +1,4 @@
-import { OpenAIEmbedding } from "./OpenAIEmbedding";
+import { OpenAIEmbedding } from "./OpenAIEmbedding.js";
 
 export class TogetherEmbedding extends OpenAIEmbedding {
   constructor(init?: Partial<OpenAIEmbedding>) {

packages/core/src/embeddings/types.ts

@@ -1,6 +1,7 @@
-import { BaseNode, MetadataMode } from "../Node";
-import { TransformComponent } from "../ingestion";
-import { SimilarityType, similarity } from "./utils";
+import type { BaseNode } from "../Node.js";
+import { MetadataMode } from "../Node.js";
+import type { TransformComponent } from "../ingestion/types.js";
+import { SimilarityType, similarity } from "./utils.js";
 
 const DEFAULT_EMBED_BATCH_SIZE = 10;
 

packages/core/src/embeddings/utils.ts

@@ -1,8 +1,8 @@
+import { defaultFS } from "@llamaindex/env";
 import _ from "lodash";
-import { ImageType } from "../Node";
-import { DEFAULT_SIMILARITY_TOP_K } from "../constants";
-import { defaultFS } from "../env";
-import { VectorStoreQueryMode } from "../storage/vectorStore/types";
+import type { ImageType } from "../Node.js";
+import { DEFAULT_SIMILARITY_TOP_K } from "../constants.js";
+import { VectorStoreQueryMode } from "../storage/vectorStore/types.js";
 
 /**
  * Similarity type
@@ -46,7 +46,7 @@ export function similarity(
 
   switch (mode) {
     case SimilarityType.EUCLIDEAN: {
-      let difference = embedding1.map((x, i) => x - embedding2[i]);
+      const difference = embedding1.map((x, i) => x - embedding2[i]);
       return -norm(difference);
     }
     case SimilarityType.DOT_PRODUCT: {
@@ -94,7 +94,7 @@ export function getTopKEmbeddings(
     );
   }
 
-  let similarities: { similarity: number; id: number }[] = [];
+  const similarities: { similarity: number; id: number }[] = [];
 
   for (let i = 0; i < embeddings.length; i++) {
     const sim = similarity(queryEmbedding, embeddings[i]);
@@ -105,8 +105,8 @@ export function getTopKEmbeddings(
 
   similarities.sort((a, b) => b.similarity - a.similarity); // Reverse sort
 
-  let resultSimilarities: number[] = [];
-  let resultIds: any[] = [];
+  const resultSimilarities: number[] = [];
+  const resultIds: any[] = [];
 
   for (let i = 0; i < similarityTopK; i++) {
     if (i >= similarities.length) {
@@ -142,21 +142,21 @@ export function getTopKMMREmbeddings(
   _similarityCutoff: number | null = null,
   mmrThreshold: number | null = null,
 ): [number[], any[]] {
-  let threshold = mmrThreshold || 0.5;
+  const threshold = mmrThreshold || 0.5;
   similarityFn = similarityFn || similarity;
 
   if (embeddingIds === null || embeddingIds.length === 0) {
     embeddingIds = Array.from({ length: embeddings.length }, (_, i) => i);
   }
-  let fullEmbedMap = new Map(embeddingIds.map((value, i) => [value, i]));
-  let embedMap = new Map(fullEmbedMap);
-  let embedSimilarity: Map<any, number> = new Map();
+  const fullEmbedMap = new Map(embeddingIds.map((value, i) => [value, i]));
+  const embedMap = new Map(fullEmbedMap);
+  const embedSimilarity: Map<any, number> = new Map();
   let score: number = Number.NEGATIVE_INFINITY;
   let highScoreId: any | null = null;
 
   for (let i = 0; i < embeddings.length; i++) {
-    let emb = embeddings[i];
-    let similarity = similarityFn(queryEmbedding, emb);
+    const emb = embeddings[i];
+    const similarity = similarityFn(queryEmbedding, emb);
     embedSimilarity.set(embeddingIds[i], similarity);
     if (similarity * threshold > score) {
       highScoreId = embeddingIds[i];
@@ -164,20 +164,20 @@ export function getTopKMMREmbeddings(
     }
   }
 
-  let results: [number, any][] = [];
+  const results: [number, any][] = [];
 
-  let embeddingLength = embeddings.length;
-  let similarityTopKCount = similarityTopK || embeddingLength;
+  const embeddingLength = embeddings.length;
+  const similarityTopKCount = similarityTopK || embeddingLength;
 
   while (results.length < Math.min(similarityTopKCount, embeddingLength)) {
     results.push([score, highScoreId]);
-    embedMap.delete(highScoreId!);
-    let recentEmbeddingId = highScoreId;
+    embedMap.delete(highScoreId);
+    const recentEmbeddingId = highScoreId;
     score = Number.NEGATIVE_INFINITY;
-    for (let embedId of Array.from(embedMap.keys())) {
-      let overlapWithRecent = similarityFn(
+    for (const embedId of Array.from(embedMap.keys())) {
+      const overlapWithRecent = similarityFn(
         embeddings[embedMap.get(embedId)!],
-        embeddings[fullEmbedMap.get(recentEmbeddingId!)!],
+        embeddings[fullEmbedMap.get(recentEmbeddingId)!],
       );
       if (
         threshold * embedSimilarity.get(embedId)! -
@@ -192,8 +192,8 @@ export function getTopKMMREmbeddings(
     }
   }
 
-  let resultSimilarities = results.map(([s, _]) => s);
-  let resultIds = results.map(([_, n]) => n);
+  const resultSimilarities = results.map(([s, _]) => s);
+  const resultIds = results.map(([_, n]) => n);
 
   return [resultSimilarities, resultIds];
 }

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

@@ -1,23 +1,22 @@
-import { ChatHistory, getHistory } from "../../ChatHistory";
+import type { ChatHistory } from "../../ChatHistory.js";
+import { getHistory } from "../../ChatHistory.js";
+import type { CondenseQuestionPrompt } from "../../Prompt.js";
 import {
-  CondenseQuestionPrompt,
   defaultCondenseQuestionPrompt,
   messagesToHistoryStr,
-} from "../../Prompt";
-import { Response } from "../../Response";
-import {
-  ServiceContext,
-  serviceContextFromDefaults,
-} from "../../ServiceContext";
-import { ChatMessage, LLM } from "../../llm";
-import { extractText, streamReducer } from "../../llm/utils";
-import { PromptMixin } from "../../prompts";
-import { BaseQueryEngine } from "../../types";
-import {
+} from "../../Prompt.js";
+import type { Response } from "../../Response.js";
+import type { ServiceContext } from "../../ServiceContext.js";
+import { serviceContextFromDefaults } from "../../ServiceContext.js";
+import type { ChatMessage, LLM } from "../../llm/index.js";
+import { extractText, streamReducer } from "../../llm/utils.js";
+import { PromptMixin } from "../../prompts/index.js";
+import type { BaseQueryEngine } from "../../types.js";
+import type {
   ChatEngine,
   ChatEngineParamsNonStreaming,
   ChatEngineParamsStreaming,
-} from "./types";
+} from "./types.js";
 
 /**
  * CondenseQuestionChatEngine is used in conjunction with a Index (for example VectorStoreIndex).

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

@@ -1,21 +1,27 @@
-import { ChatHistory, getHistory } from "../../ChatHistory";
-import { ContextSystemPrompt } from "../../Prompt";
-import { Response } from "../../Response";
-import { BaseRetriever } from "../../Retriever";
-import { Event } from "../../callbacks/CallbackManager";
-import { randomUUID } from "../../env";
-import { ChatMessage, ChatResponseChunk, LLM, OpenAI } from "../../llm";
-import { MessageContent } from "../../llm/types";
-import { extractText, streamConverter, streamReducer } from "../../llm/utils";
-import { BaseNodePostprocessor } from "../../postprocessors";
-import { PromptMixin } from "../../prompts";
-import { DefaultContextGenerator } from "./DefaultContextGenerator";
+import { randomUUID } from "@llamaindex/env";
+import type { ChatHistory } from "../../ChatHistory.js";
+import { getHistory } from "../../ChatHistory.js";
+import type { ContextSystemPrompt } from "../../Prompt.js";
+import { Response } from "../../Response.js";
+import type { BaseRetriever } from "../../Retriever.js";
+import type { Event } from "../../callbacks/CallbackManager.js";
+import type { ChatMessage, ChatResponseChunk, LLM } from "../../llm/index.js";
+import { OpenAI } from "../../llm/index.js";
+import type { MessageContent } from "../../llm/types.js";
 import {
+  extractText,
+  streamConverter,
+  streamReducer,
+} from "../../llm/utils.js";
+import type { BaseNodePostprocessor } from "../../postprocessors/index.js";
+import { PromptMixin } from "../../prompts/Mixin.js";
+import { DefaultContextGenerator } from "./DefaultContextGenerator.js";
+import type {
   ChatEngine,
   ChatEngineParamsNonStreaming,
   ChatEngineParamsStreaming,
   ContextGenerator,
-} from "./types";
+} from "./types.js";
 
 /**
  * ContextChatEngine uses the Index to get the appropriate context for each query.

packages/core/src/engines/chat/DefaultContextGenerator.ts

@@ -1,11 +1,12 @@
-import { NodeWithScore, TextNode } from "../../Node";
-import { ContextSystemPrompt, defaultContextSystemPrompt } from "../../Prompt";
-import { BaseRetriever } from "../../Retriever";
-import { Event } from "../../callbacks/CallbackManager";
-import { randomUUID } from "../../env";
-import { BaseNodePostprocessor } from "../../postprocessors";
-import { PromptMixin } from "../../prompts";
-import { Context, ContextGenerator } from "./types";
+import { randomUUID } from "@llamaindex/env";
+import type { NodeWithScore, TextNode } from "../../Node.js";
+import type { ContextSystemPrompt } from "../../Prompt.js";
+import { defaultContextSystemPrompt } from "../../Prompt.js";
+import type { BaseRetriever } from "../../Retriever.js";
+import type { Event } from "../../callbacks/CallbackManager.js";
+import type { BaseNodePostprocessor } from "../../postprocessors/index.js";
+import { PromptMixin } from "../../prompts/index.js";
+import type { Context, ContextGenerator } from "./types.js";
 
 export class DefaultContextGenerator
   extends PromptMixin

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

@@ -1,12 +1,14 @@
-import { ChatHistory, getHistory } from "../../ChatHistory";
-import { Response } from "../../Response";
-import { ChatResponseChunk, LLM, OpenAI } from "../../llm";
-import { streamConverter, streamReducer } from "../../llm/utils";
-import {
+import type { ChatHistory } from "../../ChatHistory.js";
+import { getHistory } from "../../ChatHistory.js";
+import { Response } from "../../Response.js";
+import type { ChatResponseChunk, LLM } from "../../llm/index.js";
+import { OpenAI } from "../../llm/index.js";
+import { streamConverter, streamReducer } from "../../llm/utils.js";
+import type {
   ChatEngine,
   ChatEngineParamsNonStreaming,
   ChatEngineParamsStreaming,
-} from "./types";
+} from "./types.js";
 
 /**
  * SimpleChatEngine is the simplest possible chat engine. Useful for using your own custom prompts.

packages/core/src/engines/chat/index.ts

@@ -1,4 +1,4 @@
-export { CondenseQuestionChatEngine } from "./CondenseQuestionChatEngine";
-export { ContextChatEngine } from "./ContextChatEngine";
-export { SimpleChatEngine } from "./SimpleChatEngine";
-export * from "./types";
+export { CondenseQuestionChatEngine } from "./CondenseQuestionChatEngine.js";
+export { ContextChatEngine } from "./ContextChatEngine.js";
+export { SimpleChatEngine } from "./SimpleChatEngine.js";
+export * from "./types.js";

packages/core/src/engines/chat/types.ts

@@ -1,10 +1,10 @@
-import { ChatHistory } from "../../ChatHistory";
-import { BaseNode, NodeWithScore } from "../../Node";
-import { Response } from "../../Response";
-import { Event } from "../../callbacks/CallbackManager";
-import { ChatMessage } from "../../llm";
-import { MessageContent } from "../../llm/types";
-import { ToolOutput } from "../../tools/types";
+import type { ChatHistory } from "../../ChatHistory.js";
+import type { BaseNode, NodeWithScore } from "../../Node.js";
+import type { Response } from "../../Response.js";
+import type { Event } from "../../callbacks/CallbackManager.js";
+import type { ChatMessage } from "../../llm/index.js";
+import type { MessageContent } from "../../llm/types.js";
+import type { ToolOutput } from "../../tools/types.js";
 
 /**
  * Represents the base parameters for ChatEngine.

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

@@ -1,17 +1,18 @@
-import { NodeWithScore } from "../../Node";
-import { Response } from "../../Response";
-import { BaseRetriever } from "../../Retriever";
-import { ServiceContext } from "../../ServiceContext";
-import { Event } from "../../callbacks/CallbackManager";
-import { randomUUID } from "../../env";
-import { BaseNodePostprocessor } from "../../postprocessors";
-import { PromptMixin } from "../../prompts";
-import { BaseSynthesizer, ResponseSynthesizer } from "../../synthesizers";
-import {
+import { randomUUID } from "@llamaindex/env";
+import type { NodeWithScore } from "../../Node.js";
+import type { Response } from "../../Response.js";
+import type { BaseRetriever } from "../../Retriever.js";
+import type { ServiceContext } from "../../ServiceContext.js";
+import type { Event } from "../../callbacks/CallbackManager.js";
+import type { BaseNodePostprocessor } from "../../postprocessors/index.js";
+import { PromptMixin } from "../../prompts/Mixin.js";
+import type { BaseSynthesizer } from "../../synthesizers/index.js";
+import { ResponseSynthesizer } from "../../synthesizers/index.js";
+import type {
   BaseQueryEngine,
   QueryEngineParamsNonStreaming,
   QueryEngineParamsStreaming,
-} from "../../types";
+} from "../../types.js";
 
 /**
  * A query engine that uses a retriever to query an index and then synthesizes the response.

packages/core/src/engines/query/RouterQueryEngine.ts

@@ -1,18 +1,17 @@
-import { BaseNode } from "../../Node";
-import { Response } from "../../Response";
-import {
-  ServiceContext,
-  serviceContextFromDefaults,
-} from "../../ServiceContext";
-import { PromptMixin } from "../../prompts";
-import { BaseSelector, LLMSingleSelector } from "../../selectors";
-import { TreeSummarize } from "../../synthesizers";
-import {
+import type { BaseNode } from "../../Node.js";
+import { Response } from "../../Response.js";
+import type { ServiceContext } from "../../ServiceContext.js";
+import { serviceContextFromDefaults } from "../../ServiceContext.js";
+import { PromptMixin } from "../../prompts/index.js";
+import type { BaseSelector } from "../../selectors/index.js";
+import { LLMSingleSelector } from "../../selectors/index.js";
+import { TreeSummarize } from "../../synthesizers/index.js";
+import type {
   BaseQueryEngine,
   QueryBundle,
   QueryEngineParamsNonStreaming,
   QueryEngineParamsStreaming,
-} from "../../types";
+} from "../../types.js";
 
 type RouterQueryEngineTool = {
   queryEngine: BaseQueryEngine;

packages/core/src/engines/query/SubQuestionQueryEngine.ts

@@ -1,28 +1,27 @@
-import { NodeWithScore, TextNode } from "../../Node";
-import { LLMQuestionGenerator } from "../../QuestionGenerator";
-import { Response } from "../../Response";
+import { randomUUID } from "@llamaindex/env";
+import type { NodeWithScore } from "../../Node.js";
+import { TextNode } from "../../Node.js";
+import { LLMQuestionGenerator } from "../../QuestionGenerator.js";
+import type { Response } from "../../Response.js";
+import type { ServiceContext } from "../../ServiceContext.js";
+import { serviceContextFromDefaults } from "../../ServiceContext.js";
+import type { Event } from "../../callbacks/CallbackManager.js";
+import { PromptMixin } from "../../prompts/Mixin.js";
+import type { BaseSynthesizer } from "../../synthesizers/index.js";
 import {
-  ServiceContext,
-  serviceContextFromDefaults,
-} from "../../ServiceContext";
-import { Event } from "../../callbacks/CallbackManager";
-import { randomUUID } from "../../env";
-import { PromptMixin } from "../../prompts";
-import {
-  BaseSynthesizer,
   CompactAndRefine,
   ResponseSynthesizer,
-} from "../../synthesizers";
+} from "../../synthesizers/index.js";
 
-import {
+import type {
   BaseQueryEngine,
   BaseTool,
   QueryEngineParamsNonStreaming,
   QueryEngineParamsStreaming,
   ToolMetadata,
-} from "../../types";
+} from "../../types.js";
 
-import { BaseQuestionGenerator, SubQuestion } from "./types";
+import type { BaseQuestionGenerator, SubQuestion } from "./types.js";
 
 /**
  * SubQuestionQueryEngine decomposes a question into subquestions and then

packages/core/src/engines/query/index.ts

@@ -1,3 +1,3 @@
-export * from "./RetrieverQueryEngine";
-export * from "./RouterQueryEngine";
-export * from "./SubQuestionQueryEngine";
+export * from "./RetrieverQueryEngine.js";
+export * from "./RouterQueryEngine.js";
+export * from "./SubQuestionQueryEngine.js";

packages/core/src/engines/query/types.ts

@@ -1,4 +1,4 @@
-import { ToolMetadata } from "../../types";
+import type { ToolMetadata } from "../../types.js";
 
 /**
  * QuestionGenerators generate new questions for the LLM using tools and a user query.

packages/core/src/evaluation/Correctness.ts

@@ -0,0 +1,123 @@
+import { MetadataMode } from "../Node.js";
+import type { ServiceContext } from "../ServiceContext.js";
+import { serviceContextFromDefaults } from "../ServiceContext.js";
+import type { ChatMessage } from "../llm/types.js";
+import { PromptMixin } from "../prompts/Mixin.js";
+import type { CorrectnessSystemPrompt } from "./prompts.js";
+import {
+  defaultCorrectnessSystemPrompt,
+  defaultUserPrompt,
+} from "./prompts.js";
+import type {
+  BaseEvaluator,
+  EvaluationResult,
+  EvaluatorParams,
+  EvaluatorResponseParams,
+} from "./types.js";
+import { defaultEvaluationParser } from "./utils.js";
+
+type CorrectnessParams = {
+  serviceContext?: ServiceContext;
+  scoreThreshold?: number;
+  parserFunction?: (str: string) => [number, string];
+};
+
+/** Correctness Evaluator */
+export class CorrectnessEvaluator extends PromptMixin implements BaseEvaluator {
+  private serviceContext: ServiceContext;
+  private scoreThreshold: number;
+  private parserFunction: (str: string) => [number, string];
+
+  private correctnessPrompt: CorrectnessSystemPrompt =
+    defaultCorrectnessSystemPrompt;
+
+  constructor(params: CorrectnessParams) {
+    super();
+
+    this.serviceContext = params.serviceContext || serviceContextFromDefaults();
+    this.correctnessPrompt = defaultCorrectnessSystemPrompt;
+    this.scoreThreshold = params.scoreThreshold || 4.0;
+    this.parserFunction = params.parserFunction || defaultEvaluationParser;
+  }
+
+  _updatePrompts(prompts: {
+    correctnessPrompt: CorrectnessSystemPrompt;
+  }): void {
+    if ("correctnessPrompt" in prompts) {
+      this.correctnessPrompt = prompts["correctnessPrompt"];
+    }
+  }
+
+  /**
+   *
+   * @param query Query to evaluate
+   * @param response  Response to evaluate
+   * @param contexts Array of contexts
+   * @param reference  Reference response
+   */
+  async evaluate({
+    query,
+    response,
+    contexts,
+    reference,
+  }: EvaluatorParams): Promise<EvaluationResult> {
+    if (query === null || response === null) {
+      throw new Error("query, and response must be provided");
+    }
+
+    const messages: ChatMessage[] = [
+      {
+        role: "system",
+        content: this.correctnessPrompt(),
+      },
+      {
+        role: "user",
+        content: defaultUserPrompt({
+          query,
+          generatedAnswer: response,
+          referenceAnswer: reference || "(NO REFERENCE ANSWER SUPPLIED)",
+        }),
+      },
+    ];
+
+    const evalResponse = await this.serviceContext.llm.chat({
+      messages,
+    });
+
+    const [score, reasoning] = this.parserFunction(
+      evalResponse.message.content,
+    );
+
+    return {
+      query: query,
+      response: response,
+      passing: score >= this.scoreThreshold || score === null,
+      score: score,
+      feedback: reasoning,
+    };
+  }
+
+  /**
+   * @param query Query to evaluate
+   * @param response  Response to evaluate
+   */
+  async evaluateResponse({
+    query,
+    response,
+  }: EvaluatorResponseParams): Promise<EvaluationResult> {
+    const responseStr = response?.response;
+    const contexts = [];
+
+    if (response) {
+      for (const node of response.sourceNodes || []) {
+        contexts.push(node.getContent(MetadataMode.ALL));
+      }
+    }
+
+    return this.evaluate({
+      query,
+      response: responseStr,
+      contexts,
+    });
+  }
+}

packages/core/src/evaluation/Faithfulness.ts

@@ -0,0 +1,151 @@
+import { Document, MetadataMode } from "../Node.js";
+import type { ServiceContext } from "../ServiceContext.js";
+import { serviceContextFromDefaults } from "../ServiceContext.js";
+import { SummaryIndex } from "../indices/summary/index.js";
+import { PromptMixin } from "../prompts/Mixin.js";
+import type {
+  FaithfulnessRefinePrompt,
+  FaithfulnessTextQAPrompt,
+} from "./prompts.js";
+import {
+  defaultFaithfulnessRefinePrompt,
+  defaultFaithfulnessTextQaPrompt,
+} from "./prompts.js";
+import type {
+  BaseEvaluator,
+  EvaluationResult,
+  EvaluatorParams,
+  EvaluatorResponseParams,
+} from "./types.js";
+
+export class FaithfulnessEvaluator
+  extends PromptMixin
+  implements BaseEvaluator
+{
+  private serviceContext: ServiceContext;
+  private raiseError: boolean;
+  private evalTemplate: FaithfulnessTextQAPrompt;
+  private refineTemplate: FaithfulnessRefinePrompt;
+
+  constructor(params: {
+    serviceContext?: ServiceContext;
+    raiseError?: boolean;
+    faithfulnessSystemPrompt?: FaithfulnessTextQAPrompt;
+    faithFulnessRefinePrompt?: FaithfulnessRefinePrompt;
+  }) {
+    super();
+    this.serviceContext = params.serviceContext || serviceContextFromDefaults();
+    this.raiseError = params.raiseError || false;
+
+    this.evalTemplate =
+      params.faithfulnessSystemPrompt || defaultFaithfulnessTextQaPrompt;
+    this.refineTemplate =
+      params.faithFulnessRefinePrompt || defaultFaithfulnessRefinePrompt;
+  }
+
+  protected _getPrompts(): { [x: string]: any } {
+    return {
+      faithfulnessSystemPrompt: this.evalTemplate,
+      faithFulnessRefinePrompt: this.refineTemplate,
+    };
+  }
+
+  protected _updatePrompts(promptsDict: {
+    faithfulnessSystemPrompt: FaithfulnessTextQAPrompt;
+    faithFulnessRefinePrompt: FaithfulnessRefinePrompt;
+  }): void {
+    if (promptsDict.faithfulnessSystemPrompt) {
+      this.evalTemplate = promptsDict.faithfulnessSystemPrompt;
+    }
+
+    if (promptsDict.faithFulnessRefinePrompt) {
+      this.refineTemplate = promptsDict.faithFulnessRefinePrompt;
+    }
+  }
+
+  /**
+   * @param query Query to evaluate
+   * @param response  Response to evaluate
+   * @param contexts Array of contexts
+   * @param reference  Reference response
+   * @param sleepTimeInSeconds  Sleep time in seconds
+   */
+  async evaluate({
+    query,
+    response,
+    contexts = [],
+    reference,
+    sleepTimeInSeconds = 0,
+  }: EvaluatorParams): Promise<EvaluationResult> {
+    if (query === null || response === null) {
+      throw new Error("query, and response must be provided");
+    }
+
+    await new Promise((resolve) =>
+      setTimeout(resolve, sleepTimeInSeconds * 1000),
+    );
+
+    const docs = contexts?.map((context) => new Document({ text: context }));
+
+    const index = await SummaryIndex.fromDocuments(docs, {
+      serviceContext: this.serviceContext,
+    });
+
+    const queryEngine = index.asQueryEngine();
+
+    queryEngine.updatePrompts({
+      "responseSynthesizer:textQATemplate": this.evalTemplate,
+      "responseSynthesizer:refineTemplate": this.refineTemplate,
+    });
+
+    const responseObj = await queryEngine.query({
+      query: response,
+    });
+
+    const rawResponseTxt = responseObj.toString();
+
+    let passing: boolean;
+
+    if (rawResponseTxt.toLowerCase().includes("yes")) {
+      passing = true;
+    } else {
+      passing = false;
+      if (this.raiseError) {
+        throw new Error("The response is invalid");
+      }
+    }
+
+    return {
+      query,
+      contexts,
+      response,
+      passing,
+      score: passing ? 1.0 : 0.0,
+      feedback: rawResponseTxt,
+    };
+  }
+
+  /**
+   * @param query Query to evaluate
+   * @param response  Response to evaluate
+   */
+  async evaluateResponse({
+    query,
+    response,
+  }: EvaluatorResponseParams): Promise<EvaluationResult> {
+    const responseStr = response?.response;
+    const contexts = [];
+
+    if (response) {
+      for (const node of response.sourceNodes || []) {
+        contexts.push(node.getContent(MetadataMode.ALL));
+      }
+    }
+
+    return this.evaluate({
+      query,
+      response: responseStr,
+      contexts,
+    });
+  }
+}

packages/core/src/evaluation/Relevancy.ts

@@ -0,0 +1,139 @@
+import { Document, MetadataMode } from "../Node.js";
+import type { ServiceContext } from "../ServiceContext.js";
+import { serviceContextFromDefaults } from "../ServiceContext.js";
+import { SummaryIndex } from "../indices/summary/index.js";
+import { PromptMixin } from "../prompts/Mixin.js";
+import type { RelevancyEvalPrompt, RelevancyRefinePrompt } from "./prompts.js";
+import {
+  defaultRelevancyEvalPrompt,
+  defaultRelevancyRefinePrompt,
+} from "./prompts.js";
+import type {
+  BaseEvaluator,
+  EvaluationResult,
+  EvaluatorParams,
+  EvaluatorResponseParams,
+} from "./types.js";
+
+type RelevancyParams = {
+  serviceContext?: ServiceContext;
+  raiseError?: boolean;
+  evalTemplate?: RelevancyEvalPrompt;
+  refineTemplate?: RelevancyRefinePrompt;
+};
+
+export class RelevancyEvaluator extends PromptMixin implements BaseEvaluator {
+  private serviceContext: ServiceContext;
+  private raiseError: boolean;
+
+  private evalTemplate: RelevancyEvalPrompt;
+  private refineTemplate: RelevancyRefinePrompt;
+
+  constructor(params: RelevancyParams) {
+    super();
+
+    this.serviceContext = params.serviceContext ?? serviceContextFromDefaults();
+    this.raiseError = params.raiseError ?? false;
+    this.evalTemplate = params.evalTemplate ?? defaultRelevancyEvalPrompt;
+    this.refineTemplate = params.refineTemplate ?? defaultRelevancyRefinePrompt;
+  }
+
+  _getPrompts() {
+    return {
+      evalTemplate: this.evalTemplate,
+      refineTemplate: this.refineTemplate,
+    };
+  }
+
+  _updatePrompts(prompts: {
+    evalTemplate: RelevancyEvalPrompt;
+    refineTemplate: RelevancyRefinePrompt;
+  }): void {
+    if ("evalTemplate" in prompts) {
+      this.evalTemplate = prompts["evalTemplate"];
+    }
+    if ("refineTemplate" in prompts) {
+      this.refineTemplate = prompts["refineTemplate"];
+    }
+  }
+
+  async evaluate({
+    query,
+    response,
+    contexts = [],
+    sleepTimeInSeconds = 0,
+  }: EvaluatorParams): Promise<EvaluationResult> {
+    if (query === null || response === null) {
+      throw new Error("query, contexts, and response must be provided");
+    }
+
+    await new Promise((resolve) =>
+      setTimeout(resolve, sleepTimeInSeconds * 1000),
+    );
+
+    const docs = contexts?.map((context) => new Document({ text: context }));
+
+    const index = await SummaryIndex.fromDocuments(docs, {
+      serviceContext: this.serviceContext,
+    });
+
+    const queryResponse = `Question: ${query}\nResponse: ${response}`;
+
+    const queryEngine = index.asQueryEngine();
+
+    queryEngine.updatePrompts({
+      "responseSynthesizer:textQATemplate": this.evalTemplate,
+      "responseSynthesizer:refineTemplate": this.refineTemplate,
+    });
+
+    const responseObj = await queryEngine.query({
+      query: queryResponse,
+    });
+
+    const rawResponseTxt = responseObj.toString();
+
+    let passing: boolean;
+
+    if (rawResponseTxt.toLowerCase().includes("yes")) {
+      passing = true;
+    } else {
+      passing = false;
+      if (this.raiseError) {
+        throw new Error("The response is invalid");
+      }
+    }
+
+    return {
+      query,
+      contexts,
+      response,
+      passing,
+      score: passing ? 1.0 : 0.0,
+      feedback: rawResponseTxt,
+    };
+  }
+
+  /**
+   * @param query Query to evaluate
+   * @param response  Response to evaluate
+   */
+  async evaluateResponse({
+    query,
+    response,
+  }: EvaluatorResponseParams): Promise<EvaluationResult> {
+    const responseStr = response?.response;
+    const contexts = [];
+
+    if (response) {
+      for (const node of response.sourceNodes || []) {
+        contexts.push(node.getContent(MetadataMode.ALL));
+      }
+    }
+
+    return this.evaluate({
+      query,
+      response: responseStr,
+      contexts,
+    });
+  }
+}

packages/core/src/evaluation/index.ts

@@ -0,0 +1,5 @@
+export * from "./Correctness.js";
+export * from "./Faithfulness.js";
+export * from "./Relevancy.js";
+export * from "./prompts.js";
+export * from "./utils.js";

packages/core/src/evaluation/prompts.ts

@@ -0,0 +1,155 @@
+export const defaultUserPrompt = ({
+  query,
+  referenceAnswer,
+  generatedAnswer,
+}: {
+  query: string;
+  referenceAnswer: string;
+  generatedAnswer: string;
+}) => `
+## User Query
+${query}
+
+## Reference Answer
+${referenceAnswer}
+
+## Generated Answer
+${generatedAnswer}
+`;
+
+export type UserPrompt = typeof defaultUserPrompt;
+
+export const defaultCorrectnessSystemPrompt =
+  () => `You are an expert evaluation system for a question answering chatbot.
+
+You are given the following information:
+- a user query, and
+- a generated answer
+
+You may also be given a reference answer to use for reference in your evaluation.
+
+Your job is to judge the relevance and correctness of the generated answer.
+Output a single score that represents a holistic evaluation.
+You must return your response in a line with only the score.
+Do not return answers in any other format.
+On a separate line provide your reasoning for the score as well.
+
+Follow these guidelines for scoring:
+- Your score has to be between 1 and 5, where 1 is the worst and 5 is the best.
+- If the generated answer is not relevant to the user query,
+you should give a score of 1.
+- If the generated answer is relevant but contains mistakes,
+you should give a score between 2 and 3.
+- If the generated answer is relevant and fully correct,
+you should give a score between 4 and 5.
+
+Example Response:
+4.0
+The generated answer has the exact same metrics as the reference answer
+but it is not as concise.
+`;
+
+export type CorrectnessSystemPrompt = typeof defaultCorrectnessSystemPrompt;
+
+export const defaultFaithfulnessRefinePrompt = ({
+  query,
+  context,
+  existingAnswer,
+}: {
+  query: string;
+  context: string;
+  existingAnswer: string;
+}) => `
+We want to understand if the following information is present
+in the context information: ${query}
+We have provided an existing YES/NO answer: ${existingAnswer}
+We have the opportunity to refine the existing answer
+(only if needed) with some more context below.
+------------
+${context}
+------------
+If the existing answer was already YES, still answer YES.
+If the information is present in the new context, answer YES.
+Otherwise answer NO.
+`;
+
+export type FaithfulnessRefinePrompt = typeof defaultFaithfulnessRefinePrompt;
+
+export const defaultFaithfulnessTextQaPrompt = ({
+  query,
+  context,
+}: {
+  query: string;
+  context: string;
+}) => `
+Please tell if a given piece of information
+is supported by the context.
+You need to answer with either YES or NO.
+Answer YES if any of the context supports the information, even
+if most of the context is unrelated.
+Some examples are provided below.
+
+Information: Apple pie is generally double-crusted.
+Context: An apple pie is a fruit pie in which the principal filling
+ingredient is apples.
+Apple pie is often served with whipped cream, ice cream
+('apple pie à la mode'), custard or cheddar cheese.
+It is generally double-crusted, with pastry both above
+and below the filling; the upper crust may be solid or
+latticed (woven of crosswise strips).
+Answer: YES
+Information: Apple pies tastes bad.
+Context: An apple pie is a fruit pie in which the principal filling
+ingredient is apples.
+Apple pie is often served with whipped cream, ice cream
+('apple pie à la mode'), custard or cheddar cheese.
+It is generally double-crusted, with pastry both above
+and below the filling; the upper crust may be solid or
+latticed (woven of crosswise strips).
+Answer: NO
+Information: ${query}
+Context: ${context}
+Answer:
+`;
+
+export type FaithfulnessTextQAPrompt = typeof defaultFaithfulnessTextQaPrompt;
+
+export const defaultRelevancyEvalPrompt = ({
+  query,
+  context,
+}: {
+  query: string;
+  context: string;
+}) => `Your task is to evaluate if the response for the query is in line with the context information provided.
+You have two options to answer. Either YES/ NO.
+Answer - YES, if the response for the query is in line with context information otherwise NO.
+Query and Response: ${query}
+Context: ${context}
+Answer: `;
+
+export type RelevancyEvalPrompt = typeof defaultRelevancyEvalPrompt;
+
+export const defaultRelevancyRefinePrompt = ({
+  query,
+  existingAnswer,
+  contextMsg,
+}: {
+  query: string;
+  existingAnswer: string;
+  contextMsg: string;
+}) => `We want to understand if the following query and response is
+in line with the context information: 
+${query}
+We have provided an existing YES/NO answer: 
+${existingAnswer}
+We have the opportunity to refine the existing answer
+(only if needed) with some more context below.
+------------
+${contextMsg}
+------------
+If the existing answer was already YES, still answer YES.
+If the information is present in the new context, answer YES.
+Otherwise answer NO.
+`;
+
+export type RelevancyRefinePrompt = typeof defaultRelevancyRefinePrompt;

packages/core/src/evaluation/types.ts

@@ -0,0 +1,30 @@
+import { Response } from "../Response.js";
+
+export type EvaluationResult = {
+  query?: string;
+  contexts?: string[];
+  response: string | null;
+  score: number;
+  scoreSecondary?: number;
+  scoreSecondaryType?: string;
+  meta?: any;
+  passing: boolean;
+  feedback: string;
+};
+
+export type EvaluatorParams = {
+  query: string | null;
+  response: string;
+  contexts?: string[];
+  reference?: string;
+  sleepTimeInSeconds?: number;
+};
+
+export type EvaluatorResponseParams = {
+  query: string | null;
+  response: Response;
+};
+export interface BaseEvaluator {
+  evaluate(params: EvaluatorParams): Promise<EvaluationResult>;
+  evaluateResponse?(params: EvaluatorResponseParams): Promise<EvaluationResult>;
+}

packages/core/src/evaluation/utils.ts

@@ -0,0 +1,8 @@
+export const defaultEvaluationParser = (
+  evalResponse: string,
+): [number, string] => {
+  const [scoreStr, reasoningStr] = evalResponse.split("\n");
+  const score = parseFloat(scoreStr);
+  const reasoning = reasoningStr.trim();
+  return [score, reasoning];
+};

packages/core/src/extractors/MetadataExtractors.ts

@@ -1,13 +1,15 @@
-import { BaseNode, MetadataMode, TextNode } from "../Node";
-import { LLM, OpenAI } from "../llm";
+import type { BaseNode } from "../Node.js";
+import { MetadataMode, TextNode } from "../Node.js";
+import type { LLM } from "../llm/index.js";
+import { OpenAI } from "../llm/index.js";
 import {
   defaultKeywordExtractorPromptTemplate,
   defaultQuestionAnswerPromptTemplate,
   defaultSummaryExtractorPromptTemplate,
   defaultTitleCombinePromptTemplate,
   defaultTitleExtractorPromptTemplate,
-} from "./prompts";
-import { BaseExtractor } from "./types";
+} from "./prompts.js";
+import { BaseExtractor } from "./types.js";
 
 const STRIP_REGEX = /(\r\n|\n|\r)/gm;
 
@@ -172,7 +174,7 @@ export class TitleExtractor extends BaseExtractor {
 
     if (nodesToExtractTitle.length === 0) return [];
 
-    let titlesCandidates: string[] = [];
+    const titlesCandidates: string[] = [];
     let title: string = "";
 
     for (let i = 0; i < nodesToExtractTitle.length; i++) {
@@ -411,7 +413,7 @@ export class SummaryExtractor extends BaseExtractor {
       nodes.map((node) => this.generateNodeSummary(node)),
     );
 
-    let metadataList: any[] = nodes.map(() => ({}));
+    const metadataList: any[] = nodes.map(() => ({}));
 
     for (let i = 0; i < nodes.length; i++) {
       if (i > 0 && this._prevSummary && nodeSummaries[i - 1]) {

packages/core/src/extractors/index.ts

@@ -3,5 +3,5 @@ export {
   QuestionsAnsweredExtractor,
   SummaryExtractor,
   TitleExtractor,
-} from "./MetadataExtractors";
-export { BaseExtractor } from "./types";
+} from "./MetadataExtractors.js";
+export { BaseExtractor } from "./types.js";

packages/core/src/extractors/types.ts

@@ -1,6 +1,7 @@
-import { BaseNode, MetadataMode, TextNode } from "../Node";
-import { TransformComponent } from "../ingestion";
-import { defaultNodeTextTemplate } from "./prompts";
+import type { BaseNode } from "../Node.js";
+import { MetadataMode, TextNode } from "../Node.js";
+import type { TransformComponent } from "../ingestion/types.js";
+import { defaultNodeTextTemplate } from "./prompts.js";
 
 /*
  * Abstract class for all extractors.
@@ -43,16 +44,16 @@ export abstract class BaseExtractor implements TransformComponent {
       newNodes = nodes.slice();
     }
 
-    let curMetadataList = await this.extract(newNodes);
+    const curMetadataList = await this.extract(newNodes);
 
-    for (let idx in newNodes) {
+    for (const idx in newNodes) {
       newNodes[idx].metadata = {
         ...newNodes[idx].metadata,
         ...curMetadataList[idx],
       };
     }
 
-    for (let idx in newNodes) {
+    for (const idx in newNodes) {
       if (excludedEmbedMetadataKeys) {
         newNodes[idx].excludedEmbedMetadataKeys.concat(
           excludedEmbedMetadataKeys,

packages/core/src/index.ts

@@ -1,31 +1,32 @@
-export * from "./ChatHistory";
-export * from "./GlobalsHelper";
-export * from "./Node";
-export * from "./OutputParser";
-export * from "./Prompt";
-export * from "./PromptHelper";
-export * from "./QuestionGenerator";
-export * from "./Response";
-export * from "./Retriever";
-export * from "./ServiceContext";
-export * from "./TextSplitter";
-export * from "./agent";
-export * from "./callbacks/CallbackManager";
-export * from "./cloud";
-export * from "./constants";
-export * from "./embeddings";
-export * from "./engines/chat";
-export * from "./engines/query";
-export * from "./extractors";
-export * from "./indices";
-export * from "./ingestion";
-export * from "./llm";
-export * from "./nodeParsers";
-export * from "./objects";
-export * from "./postprocessors";
-export * from "./prompts";
-export * from "./readers";
-export * from "./selectors";
-export * from "./storage";
-export * from "./synthesizers";
-export * from "./tools";
+export * from "./ChatHistory.js";
+export * from "./GlobalsHelper.js";
+export * from "./Node.js";
+export * from "./OutputParser.js";
+export * from "./Prompt.js";
+export * from "./PromptHelper.js";
+export * from "./QuestionGenerator.js";
+export * from "./Response.js";
+export * from "./Retriever.js";
+export * from "./ServiceContext.js";
+export * from "./TextSplitter.js";
+export * from "./agent/index.js";
+export * from "./callbacks/CallbackManager.js";
+export * from "./cloud/index.js";
+export * from "./constants.js";
+export * from "./embeddings/index.js";
+export * from "./engines/chat/index.js";
+export * from "./engines/query/index.js";
+export * from "./evaluation/index.js";
+export * from "./extractors/index.js";
+export * from "./indices/index.js";
+export * from "./ingestion/index.js";
+export * from "./llm/index.js";
+export * from "./nodeParsers/index.js";
+export * from "./objects/index.js";
+export * from "./postprocessors/index.js";
+export * from "./prompts/index.js";
+export * from "./readers/index.js";
+export * from "./selectors/index.js";
+export * from "./storage/index.js";
+export * from "./synthesizers/index.js";
+export * from "./tools/index.js";

packages/core/src/indices/BaseIndex.ts

@@ -1,112 +1,15 @@
-import { BaseNode, Document, jsonToNode } from "../Node";
-import { BaseRetriever } from "../Retriever";
-import { ServiceContext } from "../ServiceContext";
-import { randomUUID } from "../env";
-import { runTransformations } from "../ingestion";
-import { StorageContext } from "../storage/StorageContext";
-import { BaseDocumentStore } from "../storage/docStore/types";
-import { BaseIndexStore } from "../storage/indexStore/types";
-import { VectorStore } from "../storage/vectorStore/types";
-import { BaseSynthesizer } from "../synthesizers";
-import { BaseQueryEngine } from "../types";
-
-/**
- * The underlying structure of each index.
- */
-export abstract class IndexStruct {
-  indexId: string;
-  summary?: string;
-
-  constructor(indexId = randomUUID(), summary = undefined) {
-    this.indexId = indexId;
-    this.summary = summary;
-  }
-
-  toJson(): Record<string, unknown> {
-    return {
-      indexId: this.indexId,
-      summary: this.summary,
-    };
-  }
-
-  getSummary(): string {
-    if (this.summary === undefined) {
-      throw new Error("summary field of the index dict is not set");
-    }
-    return this.summary;
-  }
-}
-
-export enum IndexStructType {
-  SIMPLE_DICT = "simple_dict",
-  LIST = "list",
-  KEYWORD_TABLE = "keyword_table",
-}
-
-export class IndexDict extends IndexStruct {
-  nodesDict: Record<string, BaseNode> = {};
-  type: IndexStructType = IndexStructType.SIMPLE_DICT;
-
-  getSummary(): string {
-    if (this.summary === undefined) {
-      throw new Error("summary field of the index dict is not set");
-    }
-    return this.summary;
-  }
-
-  addNode(node: BaseNode, textId?: string) {
-    const vectorId = textId ?? node.id_;
-    this.nodesDict[vectorId] = node;
-  }
-
-  toJson(): Record<string, unknown> {
-    return {
-      ...super.toJson(),
-      nodesDict: this.nodesDict,
-      type: this.type,
-    };
-  }
-
-  delete(nodeId: string) {
-    delete this.nodesDict[nodeId];
-  }
-}
-
-export function jsonToIndexStruct(json: any): IndexStruct {
-  if (json.type === IndexStructType.LIST) {
-    const indexList = new IndexList(json.indexId, json.summary);
-    indexList.nodes = json.nodes;
-    return indexList;
-  } else if (json.type === IndexStructType.SIMPLE_DICT) {
-    const indexDict = new IndexDict(json.indexId, json.summary);
-    indexDict.nodesDict = Object.entries(json.nodesDict).reduce<
-      Record<string, BaseNode>
-    >((acc, [key, value]) => {
-      acc[key] = jsonToNode(value);
-      return acc;
-    }, {});
-    return indexDict;
-  } else {
-    throw new Error(`Unknown index struct type: ${json.type}`);
-  }
-}
-
-export class IndexList extends IndexStruct {
-  nodes: string[] = [];
-  type: IndexStructType = IndexStructType.LIST;
-
-  addNode(node: BaseNode) {
-    this.nodes.push(node.id_);
-  }
-
-  toJson(): Record<string, unknown> {
-    return {
-      ...super.toJson(),
-      nodes: this.nodes,
-      type: this.type,
-    };
-  }
-}
+import type { BaseNode, Document } from "../Node.js";
+import type { BaseRetriever } from "../Retriever.js";
+import type { ServiceContext } from "../ServiceContext.js";
+import { runTransformations } from "../ingestion/IngestionPipeline.js";
+import type { StorageContext } from "../storage/StorageContext.js";
+import type { BaseDocumentStore } from "../storage/docStore/types.js";
+import type { BaseIndexStore } from "../storage/indexStore/types.js";
+import type { VectorStore } from "../storage/vectorStore/types.js";
+import type { BaseSynthesizer } from "../synthesizers/types.js";
+import type { BaseQueryEngine } from "../types.js";
+import { IndexStruct } from "./IndexStruct.js";
+import { IndexStructType } from "./json-to-index-struct.js";
 
 // A table of keywords mapping keywords to text chunks.
 export class KeywordTable extends IndexStruct {

packages/core/src/indices/IndexStruct.ts

@@ -0,0 +1,28 @@
+import { randomUUID } from "@llamaindex/env";
+
+/**
+ * The underlying structure of each index.
+ */
+export abstract class IndexStruct {
+  indexId: string;
+  summary?: string;
+
+  constructor(indexId = randomUUID(), summary = undefined) {
+    this.indexId = indexId;
+    this.summary = summary;
+  }
+
+  toJson(): Record<string, unknown> {
+    return {
+      indexId: this.indexId,
+      summary: this.summary,
+    };
+  }
+
+  getSummary(): string {
+    if (this.summary === undefined) {
+      throw new Error("summary field of the index dict is not set");
+    }
+    return this.summary;
+  }
+}

packages/core/src/indices/index.ts

@@ -1,4 +1,4 @@
-export * from "./BaseIndex";
-export * from "./keyword";
-export * from "./summary";
-export * from "./vectorStore";
+export * from "./BaseIndex.js";
+export * from "./keyword/index.js";
+export * from "./summary/index.js";
+export * from "./vectorStore/index.js";

packages/core/src/indices/json-to-index-struct.ts

@@ -0,0 +1,72 @@
+import type { BaseNode } from "../Node.js";
+import { jsonToNode } from "../Node.js";
+import { IndexStruct } from "./IndexStruct.js";
+
+export enum IndexStructType {
+  SIMPLE_DICT = "simple_dict",
+  LIST = "list",
+  KEYWORD_TABLE = "keyword_table",
+}
+export class IndexDict extends IndexStruct {
+  nodesDict: Record<string, BaseNode> = {};
+  type: IndexStructType = IndexStructType.SIMPLE_DICT;
+
+  getSummary(): string {
+    if (this.summary === undefined) {
+      throw new Error("summary field of the index dict is not set");
+    }
+    return this.summary;
+  }
+
+  addNode(node: BaseNode, textId?: string) {
+    const vectorId = textId ?? node.id_;
+    this.nodesDict[vectorId] = node;
+  }
+
+  toJson(): Record<string, unknown> {
+    return {
+      ...super.toJson(),
+      nodesDict: this.nodesDict,
+      type: this.type,
+    };
+  }
+
+  delete(nodeId: string) {
+    delete this.nodesDict[nodeId];
+  }
+}
+export function jsonToIndexStruct(json: any): IndexStruct {
+  if (json.type === IndexStructType.LIST) {
+    const indexList = new IndexList(json.indexId, json.summary);
+    indexList.nodes = json.nodes;
+    return indexList;
+  } else if (json.type === IndexStructType.SIMPLE_DICT) {
+    const indexDict = new IndexDict(json.indexId, json.summary);
+    indexDict.nodesDict = Object.entries(json.nodesDict).reduce<
+      Record<string, BaseNode>
+    >((acc, [key, value]) => {
+      acc[key] = jsonToNode(value);
+      return acc;
+    }, {});
+    return indexDict;
+  } else {
+    throw new Error(`Unknown index struct type: ${json.type}`);
+  }
+}
+
+export class IndexList extends IndexStruct {
+  nodes: string[] = [];
+  type: IndexStructType = IndexStructType.LIST;
+
+  addNode(node: BaseNode) {
+    this.nodes.push(node.id_);
+  }
+
+  toJson(): Record<string, unknown> {
+    return {
+      ...super.toJson(),
+      nodes: this.nodes,
+      type: this.type,
+    };
+  }
+}

packages/core/src/indices/keyword/KeywordTableIndex.ts

@@ -1,278 +0,0 @@
-import { BaseNode, Document, MetadataMode } from "../../Node";
-import { defaultKeywordExtractPrompt } from "../../Prompt";
-import { BaseRetriever } from "../../Retriever";
-import {
-  ServiceContext,
-  serviceContextFromDefaults,
-} from "../../ServiceContext";
-import { RetrieverQueryEngine } from "../../engines/query";
-import { BaseNodePostprocessor } from "../../postprocessors";
-import {
-  BaseDocumentStore,
-  StorageContext,
-  storageContextFromDefaults,
-} from "../../storage";
-import { BaseSynthesizer } from "../../synthesizers";
-import { BaseQueryEngine } from "../../types";
-import {
-  BaseIndex,
-  BaseIndexInit,
-  IndexStructType,
-  KeywordTable,
-} from "../BaseIndex";
-import {
-  KeywordTableLLMRetriever,
-  KeywordTableRAKERetriever,
-  KeywordTableSimpleRetriever,
-} from "./KeywordTableIndexRetriever";
-import { extractKeywordsGivenResponse } from "./utils";
-
-export interface KeywordIndexOptions {
-  nodes?: BaseNode[];
-  indexStruct?: KeywordTable;
-  indexId?: string;
-  serviceContext?: ServiceContext;
-  storageContext?: StorageContext;
-}
-export enum KeywordTableRetrieverMode {
-  DEFAULT = "DEFAULT",
-  SIMPLE = "SIMPLE",
-  RAKE = "RAKE",
-}
-
-const KeywordTableRetrieverMap = {
-  [KeywordTableRetrieverMode.DEFAULT]: KeywordTableLLMRetriever,
-  [KeywordTableRetrieverMode.SIMPLE]: KeywordTableSimpleRetriever,
-  [KeywordTableRetrieverMode.RAKE]: KeywordTableRAKERetriever,
-};
-
-/**
- * The KeywordTableIndex, an index that extracts keywords from each Node and builds a mapping from each keyword to the corresponding Nodes of that keyword.
- */
-export class KeywordTableIndex extends BaseIndex<KeywordTable> {
-  constructor(init: BaseIndexInit<KeywordTable>) {
-    super(init);
-  }
-
-  static async init(options: KeywordIndexOptions): Promise<KeywordTableIndex> {
-    const storageContext =
-      options.storageContext ?? (await storageContextFromDefaults({}));
-    const serviceContext =
-      options.serviceContext ?? serviceContextFromDefaults({});
-    const { docStore, indexStore } = storageContext;
-
-    // Setup IndexStruct from storage
-    let indexStructs = (await indexStore.getIndexStructs()) as KeywordTable[];
-    let indexStruct: KeywordTable | null;
-
-    if (options.indexStruct && indexStructs.length > 0) {
-      throw new Error(
-        "Cannot initialize index with both indexStruct and indexStore",
-      );
-    }
-
-    if (options.indexStruct) {
-      indexStruct = options.indexStruct;
-    } else if (indexStructs.length == 1) {
-      indexStruct = indexStructs[0];
-    } else if (indexStructs.length > 1 && options.indexId) {
-      indexStruct = (await indexStore.getIndexStruct(
-        options.indexId,
-      )) as KeywordTable;
-    } else {
-      indexStruct = null;
-    }
-
-    // check indexStruct type
-    if (indexStruct && indexStruct.type !== IndexStructType.KEYWORD_TABLE) {
-      throw new Error(
-        "Attempting to initialize KeywordTableIndex with non-keyword table indexStruct",
-      );
-    }
-
-    if (indexStruct) {
-      if (options.nodes) {
-        throw new Error(
-          "Cannot initialize KeywordTableIndex with both nodes and indexStruct",
-        );
-      }
-    } else {
-      if (!options.nodes) {
-        throw new Error(
-          "Cannot initialize KeywordTableIndex without nodes or indexStruct",
-        );
-      }
-      indexStruct = await KeywordTableIndex.buildIndexFromNodes(
-        options.nodes,
-        storageContext.docStore,
-        serviceContext,
-      );
-
-      await indexStore.addIndexStruct(indexStruct);
-    }
-
-    return new KeywordTableIndex({
-      storageContext,
-      serviceContext,
-      docStore,
-      indexStore,
-      indexStruct,
-    });
-  }
-
-  asRetriever(options?: any): BaseRetriever {
-    const { mode = KeywordTableRetrieverMode.DEFAULT, ...otherOptions } =
-      options ?? {};
-    const KeywordTableRetriever =
-      KeywordTableRetrieverMap[mode as KeywordTableRetrieverMode];
-    if (KeywordTableRetriever) {
-      return new KeywordTableRetriever({ index: this, ...otherOptions });
-    }
-    throw new Error(`Unknown retriever mode: ${mode}`);
-  }
-
-  asQueryEngine(options?: {
-    retriever?: BaseRetriever;
-    responseSynthesizer?: BaseSynthesizer;
-    preFilters?: unknown;
-    nodePostprocessors?: BaseNodePostprocessor[];
-  }): BaseQueryEngine {
-    const { retriever, responseSynthesizer } = options ?? {};
-    return new RetrieverQueryEngine(
-      retriever ?? this.asRetriever(),
-      responseSynthesizer,
-      options?.preFilters,
-      options?.nodePostprocessors,
-    );
-  }
-
-  static async extractKeywords(
-    text: string,
-    serviceContext: ServiceContext,
-  ): Promise<Set<string>> {
-    const response = await serviceContext.llm.complete({
-      prompt: defaultKeywordExtractPrompt({
-        context: text,
-      }),
-    });
-    return extractKeywordsGivenResponse(response.text, "KEYWORDS:");
-  }
-
-  /**
-   * High level API: split documents, get keywords, and build index.
-   * @param documents
-   * @param storageContext
-   * @param serviceContext
-   * @returns
-   */
-  static async fromDocuments(
-    documents: Document[],
-    args: {
-      storageContext?: StorageContext;
-      serviceContext?: ServiceContext;
-    } = {},
-  ): Promise<KeywordTableIndex> {
-    let { storageContext, serviceContext } = args;
-    storageContext = storageContext ?? (await storageContextFromDefaults({}));
-    serviceContext = serviceContext ?? serviceContextFromDefaults({});
-    const docStore = storageContext.docStore;
-
-    docStore.addDocuments(documents, true);
-    for (const doc of documents) {
-      docStore.setDocumentHash(doc.id_, doc.hash);
-    }
-
-    const nodes = serviceContext.nodeParser.getNodesFromDocuments(documents);
-    const index = await KeywordTableIndex.init({
-      nodes,
-      storageContext,
-      serviceContext,
-    });
-    return index;
-  }
-
-  /**
-   * Get keywords for nodes and place them into the index.
-   * @param nodes
-   * @param serviceContext
-   * @param vectorStore
-   * @returns
-   */
-  static async buildIndexFromNodes(
-    nodes: BaseNode[],
-    docStore: BaseDocumentStore,
-    serviceContext: ServiceContext,
-  ): Promise<KeywordTable> {
-    const indexStruct = new KeywordTable();
-    await docStore.addDocuments(nodes, true);
-    for (const node of nodes) {
-      const keywords = await KeywordTableIndex.extractKeywords(
-        node.getContent(MetadataMode.LLM),
-        serviceContext,
-      );
-      indexStruct.addNode([...keywords], node.id_);
-    }
-    return indexStruct;
-  }
-
-  async insertNodes(nodes: BaseNode[]) {
-    for (let node of nodes) {
-      const keywords = await KeywordTableIndex.extractKeywords(
-        node.getContent(MetadataMode.LLM),
-        this.serviceContext,
-      );
-      this.indexStruct.addNode([...keywords], node.id_);
-    }
-  }
-
-  deleteNode(nodeId: string): void {
-    const keywordsToDelete: Set<string> = new Set();
-    for (const [keyword, existingNodeIds] of Object.entries(
-      this.indexStruct.table,
-    )) {
-      const index = existingNodeIds.indexOf(nodeId);
-      if (index !== -1) {
-        existingNodeIds.splice(index, 1);
-
-        // Delete keywords that have zero nodes
-        if (existingNodeIds.length === 0) {
-          keywordsToDelete.add(keyword);
-        }
-      }
-    }
-    this.indexStruct.deleteNode([...keywordsToDelete], nodeId);
-  }
-
-  async deleteNodes(nodeIds: string[], deleteFromDocStore: boolean) {
-    nodeIds.forEach((nodeId) => {
-      this.deleteNode(nodeId);
-    });
-
-    if (deleteFromDocStore) {
-      for (const nodeId of nodeIds) {
-        await this.docStore.deleteDocument(nodeId, false);
-      }
-    }
-
-    await this.storageContext.indexStore.addIndexStruct(this.indexStruct);
-  }
-
-  async deleteRefDoc(
-    refDocId: string,
-    deleteFromDocStore?: boolean,
-  ): Promise<void> {
-    const refDocInfo = await this.docStore.getRefDocInfo(refDocId);
-
-    if (!refDocInfo) {
-      return;
-    }
-
-    await this.deleteNodes(refDocInfo.nodeIds, false);
-
-    if (deleteFromDocStore) {
-      await this.docStore.deleteRefDoc(refDocId, false);
-    }
-
-    return;
-  }
-}

packages/core/src/indices/keyword/KeywordTableIndexRetriever.ts

@@ -1,116 +0,0 @@
-import { NodeWithScore } from "../../Node";
-import {
-  defaultKeywordExtractPrompt,
-  defaultQueryKeywordExtractPrompt,
-  KeywordExtractPrompt,
-  QueryKeywordExtractPrompt,
-} from "../../Prompt";
-import { BaseRetriever } from "../../Retriever";
-import { ServiceContext } from "../../ServiceContext";
-import { BaseDocumentStore } from "../../storage/docStore/types";
-import { KeywordTable } from "../BaseIndex";
-import { KeywordTableIndex } from "./KeywordTableIndex";
-import {
-  extractKeywordsGivenResponse,
-  rakeExtractKeywords,
-  simpleExtractKeywords,
-} from "./utils";
-
-// Base Keyword Table Retriever
-abstract class BaseKeywordTableRetriever implements BaseRetriever {
-  protected index: KeywordTableIndex;
-  protected indexStruct: KeywordTable;
-  protected docstore: BaseDocumentStore;
-  protected serviceContext: ServiceContext;
-
-  protected maxKeywordsPerQuery: number; // Maximum number of keywords to extract from query.
-  protected numChunksPerQuery: number; // Maximum number of text chunks to query.
-  protected keywordExtractTemplate: KeywordExtractPrompt; // A Keyword Extraction Prompt
-  protected queryKeywordExtractTemplate: QueryKeywordExtractPrompt; // A Query Keyword Extraction Prompt
-
-  constructor({
-    index,
-    keywordExtractTemplate,
-    queryKeywordExtractTemplate,
-    maxKeywordsPerQuery = 10,
-    numChunksPerQuery = 10,
-  }: {
-    index: KeywordTableIndex;
-    keywordExtractTemplate?: KeywordExtractPrompt;
-    queryKeywordExtractTemplate?: QueryKeywordExtractPrompt;
-    maxKeywordsPerQuery: number;
-    numChunksPerQuery: number;
-  }) {
-    this.index = index;
-    this.indexStruct = index.indexStruct;
-    this.docstore = index.docStore;
-    this.serviceContext = index.serviceContext;
-
-    this.maxKeywordsPerQuery = maxKeywordsPerQuery;
-    this.numChunksPerQuery = numChunksPerQuery;
-    this.keywordExtractTemplate =
-      keywordExtractTemplate || defaultKeywordExtractPrompt;
-    this.queryKeywordExtractTemplate =
-      queryKeywordExtractTemplate || defaultQueryKeywordExtractPrompt;
-  }
-
-  abstract getKeywords(query: string): Promise<string[]>;
-
-  async retrieve(query: string): Promise<NodeWithScore[]> {
-    const keywords = await this.getKeywords(query);
-    const chunkIndicesCount: { [key: string]: number } = {};
-    const filteredKeywords = keywords.filter((keyword) =>
-      this.indexStruct.table.has(keyword),
-    );
-
-    for (let keyword of filteredKeywords) {
-      for (let nodeId of this.indexStruct.table.get(keyword) || []) {
-        chunkIndicesCount[nodeId] = (chunkIndicesCount[nodeId] ?? 0) + 1;
-      }
-    }
-
-    const sortedChunkIndices = Object.keys(chunkIndicesCount)
-      .sort((a, b) => chunkIndicesCount[b] - chunkIndicesCount[a])
-      .slice(0, this.numChunksPerQuery);
-
-    const sortedNodes = await this.docstore.getNodes(sortedChunkIndices);
-
-    return sortedNodes.map((node) => ({ node }));
-  }
-
-  getServiceContext(): ServiceContext {
-    return this.index.serviceContext;
-  }
-}
-
-// Extracts keywords using LLMs.
-export class KeywordTableLLMRetriever extends BaseKeywordTableRetriever {
-  async getKeywords(query: string): Promise<string[]> {
-    const response = await this.serviceContext.llm.complete({
-      prompt: this.queryKeywordExtractTemplate({
-        question: query,
-        maxKeywords: this.maxKeywordsPerQuery,
-      }),
-    });
-    const keywords = extractKeywordsGivenResponse(response.text, "KEYWORDS:");
-    return [...keywords];
-  }
-}
-
-// Extracts keywords using simple regex-based keyword extractor.
-export class KeywordTableSimpleRetriever extends BaseKeywordTableRetriever {
-  getKeywords(query: string): Promise<string[]> {
-    return Promise.resolve([
-      ...simpleExtractKeywords(query, this.maxKeywordsPerQuery),
-    ]);
-  }
-}
-
-// Extracts keywords using RAKE keyword extractor
-export class KeywordTableRAKERetriever extends BaseKeywordTableRetriever {
-  getKeywords(query: string): Promise<string[]> {
-    return Promise.resolve([
-      ...rakeExtractKeywords(query, this.maxKeywordsPerQuery),
-    ]);
-  }
-}

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

@@ -1,9 +1,376 @@
-export {
-  KeywordTableIndex,
-  KeywordTableRetrieverMode,
-} from "./KeywordTableIndex";
-export {
-  KeywordTableLLMRetriever,
-  KeywordTableRAKERetriever,
-  KeywordTableSimpleRetriever,
-} from "./KeywordTableIndexRetriever";
+import type { BaseNode, Document, NodeWithScore } from "../../Node.js";
+import { MetadataMode } from "../../Node.js";
+import type {
+  KeywordExtractPrompt,
+  QueryKeywordExtractPrompt,
+} from "../../Prompt.js";
+import {
+  defaultKeywordExtractPrompt,
+  defaultQueryKeywordExtractPrompt,
+} from "../../Prompt.js";
+import type { BaseRetriever } from "../../Retriever.js";
+import type { ServiceContext } from "../../ServiceContext.js";
+import { serviceContextFromDefaults } from "../../ServiceContext.js";
+import { RetrieverQueryEngine } from "../../engines/query/index.js";
+import type { BaseNodePostprocessor } from "../../postprocessors/index.js";
+import type { BaseDocumentStore, StorageContext } from "../../storage/index.js";
+import { storageContextFromDefaults } from "../../storage/index.js";
+import type { BaseSynthesizer } from "../../synthesizers/index.js";
+import type { BaseQueryEngine } from "../../types.js";
+import type { BaseIndexInit } from "../BaseIndex.js";
+import { BaseIndex, KeywordTable } from "../BaseIndex.js";
+import { IndexStructType } from "../json-to-index-struct.js";
+import {
+  extractKeywordsGivenResponse,
+  rakeExtractKeywords,
+  simpleExtractKeywords,
+} from "./utils.js";
+
+export interface KeywordIndexOptions {
+  nodes?: BaseNode[];
+  indexStruct?: KeywordTable;
+  indexId?: string;
+  serviceContext?: ServiceContext;
+  storageContext?: StorageContext;
+}
+export enum KeywordTableRetrieverMode {
+  DEFAULT = "DEFAULT",
+  SIMPLE = "SIMPLE",
+  RAKE = "RAKE",
+}
+
+// Base Keyword Table Retriever
+abstract class BaseKeywordTableRetriever implements BaseRetriever {
+  protected index: KeywordTableIndex;
+  protected indexStruct: KeywordTable;
+  protected docstore: BaseDocumentStore;
+  protected serviceContext: ServiceContext;
+
+  protected maxKeywordsPerQuery: number; // Maximum number of keywords to extract from query.
+  protected numChunksPerQuery: number; // Maximum number of text chunks to query.
+  protected keywordExtractTemplate: KeywordExtractPrompt; // A Keyword Extraction Prompt
+  protected queryKeywordExtractTemplate: QueryKeywordExtractPrompt; // A Query Keyword Extraction Prompt
+
+  constructor({
+    index,
+    keywordExtractTemplate,
+    queryKeywordExtractTemplate,
+    maxKeywordsPerQuery = 10,
+    numChunksPerQuery = 10,
+  }: {
+    index: KeywordTableIndex;
+    keywordExtractTemplate?: KeywordExtractPrompt;
+    queryKeywordExtractTemplate?: QueryKeywordExtractPrompt;
+    maxKeywordsPerQuery: number;
+    numChunksPerQuery: number;
+  }) {
+    this.index = index;
+    this.indexStruct = index.indexStruct;
+    this.docstore = index.docStore;
+    this.serviceContext = index.serviceContext;
+
+    this.maxKeywordsPerQuery = maxKeywordsPerQuery;
+    this.numChunksPerQuery = numChunksPerQuery;
+    this.keywordExtractTemplate =
+      keywordExtractTemplate || defaultKeywordExtractPrompt;
+    this.queryKeywordExtractTemplate =
+      queryKeywordExtractTemplate || defaultQueryKeywordExtractPrompt;
+  }
+
+  abstract getKeywords(query: string): Promise<string[]>;
+
+  async retrieve(query: string): Promise<NodeWithScore[]> {
+    const keywords = await this.getKeywords(query);
+    const chunkIndicesCount: { [key: string]: number } = {};
+    const filteredKeywords = keywords.filter((keyword) =>
+      this.indexStruct.table.has(keyword),
+    );
+
+    for (const keyword of filteredKeywords) {
+      for (const nodeId of this.indexStruct.table.get(keyword) || []) {
+        chunkIndicesCount[nodeId] = (chunkIndicesCount[nodeId] ?? 0) + 1;
+      }
+    }
+
+    const sortedChunkIndices = Object.keys(chunkIndicesCount)
+      .sort((a, b) => chunkIndicesCount[b] - chunkIndicesCount[a])
+      .slice(0, this.numChunksPerQuery);
+
+    const sortedNodes = await this.docstore.getNodes(sortedChunkIndices);
+
+    return sortedNodes.map((node) => ({ node }));
+  }
+
+  getServiceContext(): ServiceContext {
+    return this.index.serviceContext;
+  }
+}
+
+// Extracts keywords using LLMs.
+export class KeywordTableLLMRetriever extends BaseKeywordTableRetriever {
+  async getKeywords(query: string): Promise<string[]> {
+    const response = await this.serviceContext.llm.complete({
+      prompt: this.queryKeywordExtractTemplate({
+        question: query,
+        maxKeywords: this.maxKeywordsPerQuery,
+      }),
+    });
+    const keywords = extractKeywordsGivenResponse(response.text, "KEYWORDS:");
+    return [...keywords];
+  }
+}
+
+// Extracts keywords using simple regex-based keyword extractor.
+export class KeywordTableSimpleRetriever extends BaseKeywordTableRetriever {
+  getKeywords(query: string): Promise<string[]> {
+    return Promise.resolve([
+      ...simpleExtractKeywords(query, this.maxKeywordsPerQuery),
+    ]);
+  }
+}
+
+// Extracts keywords using RAKE keyword extractor
+export class KeywordTableRAKERetriever extends BaseKeywordTableRetriever {
+  getKeywords(query: string): Promise<string[]> {
+    return Promise.resolve([
+      ...rakeExtractKeywords(query, this.maxKeywordsPerQuery),
+    ]);
+  }
+}
+
+const KeywordTableRetrieverMap = {
+  [KeywordTableRetrieverMode.DEFAULT]: KeywordTableLLMRetriever,
+  [KeywordTableRetrieverMode.SIMPLE]: KeywordTableSimpleRetriever,
+  [KeywordTableRetrieverMode.RAKE]: KeywordTableRAKERetriever,
+};
+
+/**
+ * The KeywordTableIndex, an index that extracts keywords from each Node and builds a mapping from each keyword to the corresponding Nodes of that keyword.
+ */
+export class KeywordTableIndex extends BaseIndex<KeywordTable> {
+  constructor(init: BaseIndexInit<KeywordTable>) {
+    super(init);
+  }
+
+  static async init(options: KeywordIndexOptions): Promise<KeywordTableIndex> {
+    const storageContext =
+      options.storageContext ?? (await storageContextFromDefaults({}));
+    const serviceContext =
+      options.serviceContext ?? serviceContextFromDefaults({});
+    const { docStore, indexStore } = storageContext;
+
+    // Setup IndexStruct from storage
+    const indexStructs = (await indexStore.getIndexStructs()) as KeywordTable[];
+    let indexStruct: KeywordTable | null;
+
+    if (options.indexStruct && indexStructs.length > 0) {
+      throw new Error(
+        "Cannot initialize index with both indexStruct and indexStore",
+      );
+    }
+
+    if (options.indexStruct) {
+      indexStruct = options.indexStruct;
+    } else if (indexStructs.length == 1) {
+      indexStruct = indexStructs[0];
+    } else if (indexStructs.length > 1 && options.indexId) {
+      indexStruct = (await indexStore.getIndexStruct(
+        options.indexId,
+      )) as KeywordTable;
+    } else {
+      indexStruct = null;
+    }
+
+    // check indexStruct type
+    if (indexStruct && indexStruct.type !== IndexStructType.KEYWORD_TABLE) {
+      throw new Error(
+        "Attempting to initialize KeywordTableIndex with non-keyword table indexStruct",
+      );
+    }
+
+    if (indexStruct) {
+      if (options.nodes) {
+        throw new Error(
+          "Cannot initialize KeywordTableIndex with both nodes and indexStruct",
+        );
+      }
+    } else {
+      if (!options.nodes) {
+        throw new Error(
+          "Cannot initialize KeywordTableIndex without nodes or indexStruct",
+        );
+      }
+      indexStruct = await KeywordTableIndex.buildIndexFromNodes(
+        options.nodes,
+        storageContext.docStore,
+        serviceContext,
+      );
+
+      await indexStore.addIndexStruct(indexStruct);
+    }
+
+    return new KeywordTableIndex({
+      storageContext,
+      serviceContext,
+      docStore,
+      indexStore,
+      indexStruct,
+    });
+  }
+
+  asRetriever(options?: any): BaseRetriever {
+    const { mode = KeywordTableRetrieverMode.DEFAULT, ...otherOptions } =
+      options ?? {};
+    const KeywordTableRetriever =
+      KeywordTableRetrieverMap[mode as KeywordTableRetrieverMode];
+    if (KeywordTableRetriever) {
+      return new KeywordTableRetriever({ index: this, ...otherOptions });
+    }
+    throw new Error(`Unknown retriever mode: ${mode}`);
+  }
+
+  asQueryEngine(options?: {
+    retriever?: BaseRetriever;
+    responseSynthesizer?: BaseSynthesizer;
+    preFilters?: unknown;
+    nodePostprocessors?: BaseNodePostprocessor[];
+  }): BaseQueryEngine {
+    const { retriever, responseSynthesizer } = options ?? {};
+    return new RetrieverQueryEngine(
+      retriever ?? this.asRetriever(),
+      responseSynthesizer,
+      options?.preFilters,
+      options?.nodePostprocessors,
+    );
+  }
+
+  static async extractKeywords(
+    text: string,
+    serviceContext: ServiceContext,
+  ): Promise<Set<string>> {
+    const response = await serviceContext.llm.complete({
+      prompt: defaultKeywordExtractPrompt({
+        context: text,
+      }),
+    });
+    return extractKeywordsGivenResponse(response.text, "KEYWORDS:");
+  }
+
+  /**
+   * High level API: split documents, get keywords, and build index.
+   * @param documents
+   * @param storageContext
+   * @param serviceContext
+   * @returns
+   */
+  static async fromDocuments(
+    documents: Document[],
+    args: {
+      storageContext?: StorageContext;
+      serviceContext?: ServiceContext;
+    } = {},
+  ): Promise<KeywordTableIndex> {
+    let { storageContext, serviceContext } = args;
+    storageContext = storageContext ?? (await storageContextFromDefaults({}));
+    serviceContext = serviceContext ?? serviceContextFromDefaults({});
+    const docStore = storageContext.docStore;
+
+    docStore.addDocuments(documents, true);
+    for (const doc of documents) {
+      docStore.setDocumentHash(doc.id_, doc.hash);
+    }
+
+    const nodes = serviceContext.nodeParser.getNodesFromDocuments(documents);
+    const index = await KeywordTableIndex.init({
+      nodes,
+      storageContext,
+      serviceContext,
+    });
+    return index;
+  }
+
+  /**
+   * Get keywords for nodes and place them into the index.
+   * @param nodes
+   * @param serviceContext
+   * @param vectorStore
+   * @returns
+   */
+  static async buildIndexFromNodes(
+    nodes: BaseNode[],
+    docStore: BaseDocumentStore,
+    serviceContext: ServiceContext,
+  ): Promise<KeywordTable> {
+    const indexStruct = new KeywordTable();
+    await docStore.addDocuments(nodes, true);
+    for (const node of nodes) {
+      const keywords = await KeywordTableIndex.extractKeywords(
+        node.getContent(MetadataMode.LLM),
+        serviceContext,
+      );
+      indexStruct.addNode([...keywords], node.id_);
+    }
+    return indexStruct;
+  }
+
+  async insertNodes(nodes: BaseNode[]) {
+    for (const node of nodes) {
+      const keywords = await KeywordTableIndex.extractKeywords(
+        node.getContent(MetadataMode.LLM),
+        this.serviceContext,
+      );
+      this.indexStruct.addNode([...keywords], node.id_);
+    }
+  }
+
+  deleteNode(nodeId: string): void {
+    const keywordsToDelete: Set<string> = new Set();
+    for (const [keyword, existingNodeIds] of Object.entries(
+      this.indexStruct.table,
+    )) {
+      const index = existingNodeIds.indexOf(nodeId);
+      if (index !== -1) {
+        existingNodeIds.splice(index, 1);
+
+        // Delete keywords that have zero nodes
+        if (existingNodeIds.length === 0) {
+          keywordsToDelete.add(keyword);
+        }
+      }
+    }
+    this.indexStruct.deleteNode([...keywordsToDelete], nodeId);
+  }
+
+  async deleteNodes(nodeIds: string[], deleteFromDocStore: boolean) {
+    nodeIds.forEach((nodeId) => {
+      this.deleteNode(nodeId);
+    });
+
+    if (deleteFromDocStore) {
+      for (const nodeId of nodeIds) {
+        await this.docStore.deleteDocument(nodeId, false);
+      }
+    }
+
+    await this.storageContext.indexStore.addIndexStruct(this.indexStruct);
+  }
+
+  async deleteRefDoc(
+    refDocId: string,
+    deleteFromDocStore?: boolean,
+  ): Promise<void> {
+    const refDocInfo = await this.docStore.getRefDocInfo(refDocId);
+
+    if (!refDocInfo) {
+      return;
+    }
+
+    await this.deleteNodes(refDocInfo.nodeIds, false);
+
+    if (deleteFromDocStore) {
+      await this.docStore.deleteRefDoc(refDocId, false);
+    }
+
+    return;
+  }
+}

packages/core/src/indices/keyword/utils.ts

@@ -6,11 +6,11 @@ export function expandTokensWithSubtokens(tokens: Set<string>): Set<string> {
   const results: Set<string> = new Set();
   const regex: RegExp = /\w+/g;
 
-  for (let token of tokens) {
+  for (const token of tokens) {
     results.add(token);
     const subTokens: RegExpMatchArray | null = token.match(regex);
     if (subTokens && subTokens.length > 1) {
-      for (let w of subTokens) {
+      for (const w of subTokens) {
         results.add(w);
       }
     }
@@ -31,7 +31,7 @@ export function extractKeywordsGivenResponse(
   }
 
   const keywords: string[] = response.split(",");
-  for (let k of keywords) {
+  for (const k of keywords) {
     let rk: string = k;
     if (lowercase) {
       rk = rk.toLowerCase();
@@ -47,13 +47,13 @@ export function simpleExtractKeywords(
   maxKeywords?: number,
 ): Set<string> {
   const regex: RegExp = /\w+/g;
-  let tokens: string[] = [...textChunk.matchAll(regex)].map((token) =>
+  const tokens: string[] = [...textChunk.matchAll(regex)].map((token) =>
     token[0].toLowerCase().trim(),
   );
 
   // Creating a frequency map
   const valueCounts: { [key: string]: number } = {};
-  for (let token of tokens) {
+  for (const token of tokens) {
     valueCounts[token] = (valueCounts[token] || 0) + 1;
   }