mirror of
https://github.com/run-llama/LlamaIndexTS.git
synced 2026-07-18 16:44:33 -04:00
Compare commits
8 Commits
| Author | SHA1 | Date | |
|---|---|---|---|
| 205d16699d | |||
| 41177c798d | |||
| be79d0fd6c | |||
| c3950e8fca | |||
| e7770142ab | |||
| 88c31b2c9b | |||
| ceda4652ea | |||
| 019fa024ca |
@@ -0,0 +1,5 @@
|
||||
---
|
||||
"llamaindex": patch
|
||||
---
|
||||
|
||||
Give HistoryChatEngine pluggable options (thanks @marcusschiesser)
|
||||
@@ -1,5 +0,0 @@
|
||||
---
|
||||
"llamaindex": patch
|
||||
---
|
||||
|
||||
Add HTMLReader (thanks @mtutty)
|
||||
@@ -0,0 +1,5 @@
|
||||
---
|
||||
"llamaindex": patch
|
||||
---
|
||||
|
||||
Added LLMMetadata (thanks @marcusschiesser)
|
||||
@@ -1,5 +0,0 @@
|
||||
---
|
||||
"llamaindex": patch
|
||||
---
|
||||
|
||||
Add observer/filter to the SimpleDirectoryReader (thanks @mtutty)
|
||||
@@ -3,7 +3,6 @@
|
||||
# dependencies
|
||||
node_modules
|
||||
.pnp
|
||||
.pnpm-store
|
||||
.pnp.js
|
||||
|
||||
# testing
|
||||
|
||||
@@ -25,14 +25,15 @@ pnpm install llamaindex
|
||||
pnpm install @types/node
|
||||
```
|
||||
|
||||
Create the file example.ts
|
||||
|
||||
```ts
|
||||
// example.ts
|
||||
import fs from "fs/promises";
|
||||
import { Document, VectorStoreIndex } from "llamaindex";
|
||||
import { Tiktoken } from "@dqbd/tiktoken";
|
||||
|
||||
async function main() {
|
||||
// Initialize Tiktoken with the WASM binary file
|
||||
const tiktoken = new Tiktoken({ wasmURL: '/path/to/tiktoken_bg.wasm' });
|
||||
|
||||
// Load essay from abramov.txt in Node
|
||||
const essay = await fs.readFile(
|
||||
"node_modules/llamaindex/examples/abramov.txt",
|
||||
@@ -56,7 +57,6 @@ async function main() {
|
||||
}
|
||||
|
||||
main();
|
||||
```
|
||||
|
||||
Then you can run it using
|
||||
|
||||
|
||||
@@ -20,8 +20,12 @@ Create the file `example.ts`. This code will load some example data, create a do
|
||||
// example.ts
|
||||
import fs from "fs/promises";
|
||||
import { Document, VectorStoreIndex } from "llamaindex";
|
||||
import { Tiktoken } from "@dqbd/tiktoken";
|
||||
|
||||
async function main() {
|
||||
// Initialize Tiktoken with the WASM binary file
|
||||
const tiktoken = new Tiktoken({ wasmURL: '/path/to/tiktoken_bg.wasm' });
|
||||
|
||||
// Load essay from abramov.txt in Node
|
||||
const essay = await fs.readFile(
|
||||
"node_modules/llamaindex/examples/abramov.txt",
|
||||
|
||||
@@ -1,14 +1,5 @@
|
||||
# simple
|
||||
|
||||
## 0.0.29
|
||||
|
||||
### Patch Changes
|
||||
|
||||
- Updated dependencies [6c55b2d]
|
||||
- Updated dependencies [8aa8c65]
|
||||
- Updated dependencies [6c55b2d]
|
||||
- llamaindex@0.0.31
|
||||
|
||||
## 0.0.28
|
||||
|
||||
### Patch Changes
|
||||
|
||||
File diff suppressed because it is too large
Load Diff
@@ -1,24 +0,0 @@
|
||||
import { SimpleDirectoryReader } from "llamaindex";
|
||||
|
||||
function callback(
|
||||
category: string,
|
||||
name: string,
|
||||
status: any,
|
||||
message?: string,
|
||||
): boolean {
|
||||
console.log(category, name, status, message);
|
||||
if (name.endsWith(".pdf")) {
|
||||
console.log("I DON'T WANT PDF FILES!");
|
||||
return false;
|
||||
}
|
||||
return true;
|
||||
}
|
||||
|
||||
async function main() {
|
||||
// Load page
|
||||
const reader = new SimpleDirectoryReader(callback);
|
||||
const params = { directoryPath: "./data" };
|
||||
await reader.loadData(params);
|
||||
}
|
||||
|
||||
main().catch(console.error);
|
||||
@@ -1,21 +0,0 @@
|
||||
import { HTMLReader, VectorStoreIndex } from "llamaindex";
|
||||
|
||||
async function main() {
|
||||
// Load page
|
||||
const reader = new HTMLReader();
|
||||
const documents = await reader.loadData("data/18-1_Changelog.html");
|
||||
|
||||
// Split text and create embeddings. Store them in a VectorStoreIndex
|
||||
const index = await VectorStoreIndex.fromDocuments(documents);
|
||||
|
||||
// Query the index
|
||||
const queryEngine = index.asQueryEngine();
|
||||
const response = await queryEngine.query(
|
||||
"What were the notable changes in 18.1?",
|
||||
);
|
||||
|
||||
// Output response
|
||||
console.log(response.toString());
|
||||
}
|
||||
|
||||
main().catch(console.error);
|
||||
@@ -1,16 +1,15 @@
|
||||
{
|
||||
"version": "0.0.29",
|
||||
"version": "0.0.28",
|
||||
"private": true,
|
||||
"name": "simple",
|
||||
"dependencies": {
|
||||
"@notionhq/client": "^2.2.13",
|
||||
"@pinecone-database/pinecone": "^1.1.2",
|
||||
"commander": "^11.1.0",
|
||||
"@notionhq/client": "^2.2.12",
|
||||
"@pinecone-database/pinecone": "^1.0.1",
|
||||
"commander": "^11.0.0",
|
||||
"llamaindex": "workspace:*"
|
||||
},
|
||||
"devDependencies": {
|
||||
"@types/node": "^18.18.6",
|
||||
"ts-node": "^10.9.1"
|
||||
"@types/node": "^18.17.12"
|
||||
},
|
||||
"scripts": {
|
||||
"lint": "eslint ."
|
||||
|
||||
@@ -3,7 +3,6 @@ import {
|
||||
OpenAI,
|
||||
RetrieverQueryEngine,
|
||||
serviceContextFromDefaults,
|
||||
SimilarityPostprocessor,
|
||||
VectorStoreIndex,
|
||||
} from "llamaindex";
|
||||
import essay from "./essay";
|
||||
@@ -22,16 +21,8 @@ async function main() {
|
||||
|
||||
const retriever = index.asRetriever();
|
||||
retriever.similarityTopK = 5;
|
||||
const nodePostprocessor = new SimilarityPostprocessor({
|
||||
similarityCutoff: 0.7,
|
||||
});
|
||||
// TODO: cannot pass responseSynthesizer into retriever query engine
|
||||
const queryEngine = new RetrieverQueryEngine(
|
||||
retriever,
|
||||
undefined,
|
||||
undefined,
|
||||
[nodePostprocessor],
|
||||
);
|
||||
const queryEngine = new RetrieverQueryEngine(retriever);
|
||||
|
||||
const response = await queryEngine.query(
|
||||
"What did the author do growing up?",
|
||||
|
||||
+5
-6
@@ -11,16 +11,16 @@
|
||||
"publish-snapshot": "turbo run build lint test && changeset version --snapshot && changeset publish"
|
||||
},
|
||||
"devDependencies": {
|
||||
"@turbo/gen": "^1.10.16",
|
||||
"@types/jest": "^29.5.6",
|
||||
"eslint": "^8.52.0",
|
||||
"@turbo/gen": "^1.10.15",
|
||||
"@types/jest": "^29.5.5",
|
||||
"eslint": "^7.32.0",
|
||||
"eslint-config-custom": "workspace:*",
|
||||
"husky": "^8.0.3",
|
||||
"jest": "^29.7.0",
|
||||
"prettier": "^3.0.3",
|
||||
"prettier-plugin-organize-imports": "^3.2.3",
|
||||
"ts-jest": "^29.1.1",
|
||||
"turbo": "^1.10.16"
|
||||
"turbo": "^1.10.15"
|
||||
},
|
||||
"packageManager": "pnpm@7.15.0",
|
||||
"dependencies": {
|
||||
@@ -28,8 +28,7 @@
|
||||
},
|
||||
"pnpm": {
|
||||
"overrides": {
|
||||
"trim": "1.0.1",
|
||||
"@babel/traverse": "7.23.2"
|
||||
"trim": "1.0.1"
|
||||
}
|
||||
}
|
||||
}
|
||||
|
||||
@@ -1,13 +1,5 @@
|
||||
# llamaindex
|
||||
|
||||
## 0.0.31
|
||||
|
||||
### Patch Changes
|
||||
|
||||
- 6c55b2d: Give HistoryChatEngine pluggable options (thanks @marcusschiesser)
|
||||
- 8aa8c65: Add SimilarityPostProcessor (thanks @TomPenguin)
|
||||
- 6c55b2d: Added LLMMetadata (thanks @marcusschiesser)
|
||||
|
||||
## 0.0.30
|
||||
|
||||
### Patch Changes
|
||||
|
||||
+11
-12
@@ -1,35 +1,34 @@
|
||||
{
|
||||
"name": "llamaindex",
|
||||
"version": "0.0.31",
|
||||
"version": "0.0.30",
|
||||
"license": "MIT",
|
||||
"dependencies": {
|
||||
"@anthropic-ai/sdk": "^0.8.1",
|
||||
"@anthropic-ai/sdk": "^0.6.2",
|
||||
"@notionhq/client": "^2.2.13",
|
||||
"js-tiktoken": "^1.0.7",
|
||||
"lodash": "^4.17.21",
|
||||
"mammoth": "^1.6.0",
|
||||
"md-utils-ts": "^2.0.0",
|
||||
"mongodb": "^6.2.0",
|
||||
"mongodb": "^6.1.0",
|
||||
"notion-md-crawler": "^0.0.2",
|
||||
"openai": "^4.14.0",
|
||||
"openai": "^4.11.1",
|
||||
"papaparse": "^5.4.1",
|
||||
"pdf-parse": "^1.1.1",
|
||||
"portkey-ai": "^0.1.13",
|
||||
"rake-modified": "^1.0.8",
|
||||
"replicate": "^0.20.1",
|
||||
"string-strip-html": "^13.4.3",
|
||||
"tiktoken": "^1.0.10",
|
||||
"uuid": "^9.0.1",
|
||||
"wink-nlp": "^1.14.3"
|
||||
},
|
||||
"devDependencies": {
|
||||
"@types/lodash": "^4.14.200",
|
||||
"@types/node": "^18.18.7",
|
||||
"@types/papaparse": "^5.3.10",
|
||||
"@types/pdf-parse": "^1.1.3",
|
||||
"@types/uuid": "^9.0.6",
|
||||
"@types/lodash": "^4.14.199",
|
||||
"@types/node": "^18.18.4",
|
||||
"@types/papaparse": "^5.3.9",
|
||||
"@types/pdf-parse": "^1.1.2",
|
||||
"@types/uuid": "^9.0.5",
|
||||
"node-stdlib-browser": "^1.2.0",
|
||||
"tsup": "^7.2.0",
|
||||
"typescript": "^5.2.2"
|
||||
"typescript": "^4.9.5"
|
||||
},
|
||||
"engines": {
|
||||
"node": ">=18.0.0"
|
||||
|
||||
@@ -1,7 +1,6 @@
|
||||
import { v4 as uuidv4 } from "uuid";
|
||||
import { Event } from "./callbacks/CallbackManager";
|
||||
import { ChatHistory } from "./ChatHistory";
|
||||
import { BaseNodePostprocessor } from "./indices/BaseNodePostprocessor";
|
||||
import { ChatMessage, LLM, OpenAI } from "./llm/LLM";
|
||||
import { NodeWithScore, TextNode } from "./Node";
|
||||
import {
|
||||
@@ -179,24 +178,14 @@ export interface ContextGenerator {
|
||||
export class DefaultContextGenerator implements ContextGenerator {
|
||||
retriever: BaseRetriever;
|
||||
contextSystemPrompt: ContextSystemPrompt;
|
||||
nodePostprocessors: BaseNodePostprocessor[];
|
||||
|
||||
constructor(init: {
|
||||
retriever: BaseRetriever;
|
||||
contextSystemPrompt?: ContextSystemPrompt;
|
||||
nodePostprocessors?: BaseNodePostprocessor[];
|
||||
}) {
|
||||
this.retriever = init.retriever;
|
||||
this.contextSystemPrompt =
|
||||
init?.contextSystemPrompt ?? defaultContextSystemPrompt;
|
||||
this.nodePostprocessors = init.nodePostprocessors || [];
|
||||
}
|
||||
|
||||
private applyNodePostprocessors(nodes: NodeWithScore[]) {
|
||||
return this.nodePostprocessors.reduce(
|
||||
(nodes, nodePostprocessor) => nodePostprocessor.postprocessNodes(nodes),
|
||||
nodes,
|
||||
);
|
||||
}
|
||||
|
||||
async generate(message: string, parentEvent?: Event): Promise<Context> {
|
||||
@@ -212,16 +201,16 @@ export class DefaultContextGenerator implements ContextGenerator {
|
||||
parentEvent,
|
||||
);
|
||||
|
||||
const nodes = this.applyNodePostprocessors(sourceNodesWithScore);
|
||||
|
||||
return {
|
||||
message: {
|
||||
content: this.contextSystemPrompt({
|
||||
context: nodes.map((r) => (r.node as TextNode).text).join("\n\n"),
|
||||
context: sourceNodesWithScore
|
||||
.map((r) => (r.node as TextNode).text)
|
||||
.join("\n\n"),
|
||||
}),
|
||||
role: "system",
|
||||
},
|
||||
nodes,
|
||||
nodes: sourceNodesWithScore,
|
||||
};
|
||||
}
|
||||
}
|
||||
@@ -241,7 +230,6 @@ export class ContextChatEngine implements ChatEngine {
|
||||
chatModel?: LLM;
|
||||
chatHistory?: ChatMessage[];
|
||||
contextSystemPrompt?: ContextSystemPrompt;
|
||||
nodePostprocessors?: BaseNodePostprocessor[];
|
||||
}) {
|
||||
this.chatModel =
|
||||
init.chatModel ?? new OpenAI({ model: "gpt-3.5-turbo-16k" });
|
||||
|
||||
@@ -1,4 +1,5 @@
|
||||
import { encodingForModel, TiktokenModel } from "js-tiktoken";
|
||||
import cl100k_base from "tiktoken/encoders/cl100k_base.json";
|
||||
import { Tiktoken } from "tiktoken/lite";
|
||||
|
||||
import { v4 as uuidv4 } from "uuid";
|
||||
import { Event, EventTag, EventType } from "./callbacks/CallbackManager";
|
||||
@@ -17,17 +18,18 @@ class GlobalsHelper {
|
||||
} | null = null;
|
||||
|
||||
private initDefaultTokenizer() {
|
||||
const encoding = encodingForModel("text-embedding-ada-002"); // cl100k_base
|
||||
const encoding = new Tiktoken(
|
||||
cl100k_base.bpe_ranks,
|
||||
cl100k_base.special_tokens,
|
||||
cl100k_base.pat_str,
|
||||
);
|
||||
|
||||
this.defaultTokenizer = {
|
||||
encode: (text: string) => {
|
||||
return new Uint32Array(encoding.encode(text));
|
||||
return encoding.encode(text);
|
||||
},
|
||||
decode: (tokens: Uint32Array) => {
|
||||
const numberArray = Array.from(tokens);
|
||||
const text = encoding.decode(numberArray);
|
||||
const uint8Array = new TextEncoder().encode(text);
|
||||
return new TextDecoder().decode(uint8Array);
|
||||
return new TextDecoder().decode(encoding.decode(tokens));
|
||||
},
|
||||
};
|
||||
}
|
||||
@@ -39,10 +41,10 @@ class GlobalsHelper {
|
||||
if (!this.defaultTokenizer) {
|
||||
this.initDefaultTokenizer();
|
||||
}
|
||||
|
||||
|
||||
return this.defaultTokenizer!.encode.bind(this.defaultTokenizer);
|
||||
}
|
||||
|
||||
|
||||
tokenizerDecoder(encoding?: string) {
|
||||
if (encoding && encoding !== Tokenizers.CL100K_BASE) {
|
||||
throw new Error(`Tokenizer encoding ${encoding} not yet supported`);
|
||||
@@ -50,7 +52,7 @@ class GlobalsHelper {
|
||||
if (!this.defaultTokenizer) {
|
||||
this.initDefaultTokenizer();
|
||||
}
|
||||
|
||||
|
||||
return this.defaultTokenizer!.decode.bind(this.defaultTokenizer);
|
||||
}
|
||||
|
||||
|
||||
@@ -1,6 +1,5 @@
|
||||
import { v4 as uuidv4 } from "uuid";
|
||||
import { Event } from "./callbacks/CallbackManager";
|
||||
import { BaseNodePostprocessor } from "./indices/BaseNodePostprocessor";
|
||||
import { NodeWithScore, TextNode } from "./Node";
|
||||
import {
|
||||
BaseQuestionGenerator,
|
||||
@@ -31,14 +30,12 @@ export interface BaseQueryEngine {
|
||||
export class RetrieverQueryEngine implements BaseQueryEngine {
|
||||
retriever: BaseRetriever;
|
||||
responseSynthesizer: ResponseSynthesizer;
|
||||
nodePostprocessors: BaseNodePostprocessor[];
|
||||
preFilters?: unknown;
|
||||
|
||||
constructor(
|
||||
retriever: BaseRetriever,
|
||||
responseSynthesizer?: ResponseSynthesizer,
|
||||
preFilters?: unknown,
|
||||
nodePostprocessors?: BaseNodePostprocessor[],
|
||||
) {
|
||||
this.retriever = retriever;
|
||||
const serviceContext: ServiceContext | undefined =
|
||||
@@ -46,24 +43,6 @@ export class RetrieverQueryEngine implements BaseQueryEngine {
|
||||
this.responseSynthesizer =
|
||||
responseSynthesizer || new ResponseSynthesizer({ serviceContext });
|
||||
this.preFilters = preFilters;
|
||||
this.nodePostprocessors = nodePostprocessors || [];
|
||||
}
|
||||
|
||||
private applyNodePostprocessors(nodes: NodeWithScore[]) {
|
||||
return this.nodePostprocessors.reduce(
|
||||
(nodes, nodePostprocessor) => nodePostprocessor.postprocessNodes(nodes),
|
||||
nodes,
|
||||
);
|
||||
}
|
||||
|
||||
private async retrieve(query: string, parentEvent: Event) {
|
||||
const nodes = await this.retriever.retrieve(
|
||||
query,
|
||||
parentEvent,
|
||||
this.preFilters,
|
||||
);
|
||||
|
||||
return this.applyNodePostprocessors(nodes);
|
||||
}
|
||||
|
||||
async query(query: string, parentEvent?: Event) {
|
||||
@@ -72,7 +51,11 @@ export class RetrieverQueryEngine implements BaseQueryEngine {
|
||||
type: "wrapper",
|
||||
tags: ["final"],
|
||||
};
|
||||
const nodes = await this.retrieve(query, _parentEvent);
|
||||
const nodes = await this.retriever.retrieve(
|
||||
query,
|
||||
_parentEvent,
|
||||
this.preFilters,
|
||||
);
|
||||
return this.responseSynthesizer.synthesize(query, nodes, _parentEvent);
|
||||
}
|
||||
}
|
||||
|
||||
@@ -18,7 +18,6 @@ export * from "./readers/CSVReader";
|
||||
export * from "./readers/MarkdownReader";
|
||||
export * from "./readers/NotionReader";
|
||||
export * from "./readers/PDFReader";
|
||||
export * from "./readers/HTMLReader";
|
||||
export * from "./readers/SimpleDirectoryReader";
|
||||
export * from "./Response";
|
||||
export * from "./ResponseSynthesizer";
|
||||
|
||||
@@ -1,20 +0,0 @@
|
||||
import { NodeWithScore } from "../Node";
|
||||
|
||||
export interface BaseNodePostprocessor {
|
||||
postprocessNodes: (nodes: NodeWithScore[]) => NodeWithScore[];
|
||||
}
|
||||
|
||||
export class SimilarityPostprocessor implements BaseNodePostprocessor {
|
||||
similarityCutoff?: number;
|
||||
|
||||
constructor(options?: { similarityCutoff?: number }) {
|
||||
this.similarityCutoff = options?.similarityCutoff;
|
||||
}
|
||||
|
||||
postprocessNodes(nodes: NodeWithScore[]) {
|
||||
if (this.similarityCutoff === undefined) return nodes;
|
||||
|
||||
const cutoff = this.similarityCutoff || 0;
|
||||
return nodes.filter((node) => node.score && node.score >= cutoff);
|
||||
}
|
||||
}
|
||||
@@ -1,5 +1,4 @@
|
||||
export * from "./BaseIndex";
|
||||
export * from "./BaseNodePostprocessor";
|
||||
export * from "./keyword";
|
||||
export * from "./summary";
|
||||
export * from "./vectorStore";
|
||||
|
||||
@@ -15,7 +15,6 @@ import {
|
||||
IndexStructType,
|
||||
KeywordTable,
|
||||
} from "../BaseIndex";
|
||||
import { BaseNodePostprocessor } from "../BaseNodePostprocessor";
|
||||
import {
|
||||
KeywordTableLLMRetriever,
|
||||
KeywordTableRAKERetriever,
|
||||
@@ -130,15 +129,11 @@ export class KeywordTableIndex extends BaseIndex<KeywordTable> {
|
||||
asQueryEngine(options?: {
|
||||
retriever?: BaseRetriever;
|
||||
responseSynthesizer?: ResponseSynthesizer;
|
||||
preFilters?: unknown;
|
||||
nodePostprocessors?: BaseNodePostprocessor[];
|
||||
}): BaseQueryEngine {
|
||||
const { retriever, responseSynthesizer } = options ?? {};
|
||||
return new RetrieverQueryEngine(
|
||||
retriever ?? this.asRetriever(),
|
||||
responseSynthesizer,
|
||||
options?.preFilters,
|
||||
options?.nodePostprocessors,
|
||||
);
|
||||
}
|
||||
|
||||
|
||||
@@ -10,18 +10,17 @@ import {
|
||||
ServiceContext,
|
||||
serviceContextFromDefaults,
|
||||
} from "../../ServiceContext";
|
||||
import { BaseDocumentStore, RefDocInfo } from "../../storage/docStore/types";
|
||||
import {
|
||||
StorageContext,
|
||||
storageContextFromDefaults,
|
||||
} from "../../storage/StorageContext";
|
||||
import { BaseDocumentStore, RefDocInfo } from "../../storage/docStore/types";
|
||||
import {
|
||||
BaseIndex,
|
||||
BaseIndexInit,
|
||||
IndexList,
|
||||
IndexStructType,
|
||||
} from "../BaseIndex";
|
||||
import { BaseNodePostprocessor } from "../BaseNodePostprocessor";
|
||||
import {
|
||||
SummaryIndexLLMRetriever,
|
||||
SummaryIndexRetriever,
|
||||
@@ -156,8 +155,6 @@ export class SummaryIndex extends BaseIndex<IndexList> {
|
||||
asQueryEngine(options?: {
|
||||
retriever?: BaseRetriever;
|
||||
responseSynthesizer?: ResponseSynthesizer;
|
||||
preFilters?: unknown;
|
||||
nodePostprocessors?: BaseNodePostprocessor[];
|
||||
}): BaseQueryEngine {
|
||||
let { retriever, responseSynthesizer } = options ?? {};
|
||||
|
||||
@@ -173,12 +170,7 @@ export class SummaryIndex extends BaseIndex<IndexList> {
|
||||
});
|
||||
}
|
||||
|
||||
return new RetrieverQueryEngine(
|
||||
retriever,
|
||||
responseSynthesizer,
|
||||
options?.preFilters,
|
||||
options?.nodePostprocessors,
|
||||
);
|
||||
return new RetrieverQueryEngine(retriever, responseSynthesizer);
|
||||
}
|
||||
|
||||
static async buildIndexFromNodes(
|
||||
|
||||
@@ -18,7 +18,6 @@ import {
|
||||
IndexDict,
|
||||
IndexStructType,
|
||||
} from "../BaseIndex";
|
||||
import { BaseNodePostprocessor } from "../BaseNodePostprocessor";
|
||||
import { VectorIndexRetriever } from "./VectorIndexRetriever";
|
||||
|
||||
export interface VectorIndexOptions {
|
||||
@@ -247,15 +246,11 @@ export class VectorStoreIndex extends BaseIndex<IndexDict> {
|
||||
asQueryEngine(options?: {
|
||||
retriever?: BaseRetriever;
|
||||
responseSynthesizer?: ResponseSynthesizer;
|
||||
preFilters?: unknown;
|
||||
nodePostprocessors?: BaseNodePostprocessor[];
|
||||
}): BaseQueryEngine {
|
||||
const { retriever, responseSynthesizer } = options ?? {};
|
||||
return new RetrieverQueryEngine(
|
||||
retriever ?? this.asRetriever(),
|
||||
responseSynthesizer,
|
||||
options?.preFilters,
|
||||
options?.nodePostprocessors,
|
||||
);
|
||||
}
|
||||
|
||||
|
||||
@@ -1,77 +0,0 @@
|
||||
import { Document } from "../Node";
|
||||
import { DEFAULT_FS } from "../storage/constants";
|
||||
import { GenericFileSystem } from "../storage/FileSystem";
|
||||
import { BaseReader } from "./base";
|
||||
|
||||
/**
|
||||
* Extract the significant text from an arbitrary HTML document.
|
||||
* The contents of any head, script, style, and xml tags are removed completely.
|
||||
* The URLs for a[href] tags are extracted, along with the inner text of the tag.
|
||||
* All other tags are removed, and the inner text is kept intact.
|
||||
* Html entities (e.g., &) are not decoded.
|
||||
*/
|
||||
export class HTMLReader implements BaseReader {
|
||||
/**
|
||||
* Public method for this reader.
|
||||
* Required by BaseReader interface.
|
||||
* @param file Path/name of the file to be loaded.
|
||||
* @param fs fs wrapper interface for getting the file content.
|
||||
* @returns Promise<Document[]> A Promise object, eventually yielding zero or one Document parsed from the HTML content of the specified file.
|
||||
*/
|
||||
async loadData(
|
||||
file: string,
|
||||
fs: GenericFileSystem = DEFAULT_FS,
|
||||
): Promise<Document[]> {
|
||||
const dataBuffer = await fs.readFile(file, "utf-8");
|
||||
const htmlOptions = this.getOptions();
|
||||
const content = await this.parseContent(dataBuffer, htmlOptions);
|
||||
return [new Document({ text: content, id_: file })];
|
||||
}
|
||||
|
||||
/**
|
||||
* Wrapper for string-strip-html usage.
|
||||
* @param html Raw HTML content to be parsed.
|
||||
* @param options An object of options for the underlying library
|
||||
* @see getOptions
|
||||
* @returns The HTML content, stripped of unwanted tags and attributes
|
||||
*/
|
||||
async parseContent(html: string, options: any = {}): Promise<string> {
|
||||
const { stripHtml } = await import("string-strip-html"); // ESM only
|
||||
return stripHtml(html).result;
|
||||
}
|
||||
|
||||
/**
|
||||
* Wrapper for our configuration options passed to string-strip-html library
|
||||
* @see https://codsen.com/os/string-strip-html/examples
|
||||
* @returns An object of options for the underlying library
|
||||
*/
|
||||
getOptions() {
|
||||
return {
|
||||
skipHtmlDecoding: true,
|
||||
stripTogetherWithTheirContents: [
|
||||
"script", // default
|
||||
"style", // default
|
||||
"xml", // default
|
||||
"head", // <-- custom-added
|
||||
],
|
||||
// Keep the URLs for embedded links
|
||||
// cb: (tag: any, deleteFrom: number, deleteTo: number, insert: string, rangesArr: any, proposedReturn: string) => {
|
||||
// let temp;
|
||||
// if (
|
||||
// tag.name === "a" &&
|
||||
// tag.attributes &&
|
||||
// tag.attributes.some((attr: any) => {
|
||||
// if (attr.name === "href") {
|
||||
// temp = attr.value;
|
||||
// return true;
|
||||
// }
|
||||
// })
|
||||
// ) {
|
||||
// rangesArr.push([deleteFrom, deleteTo, `${temp} ${insert || ""}`]);
|
||||
// } else {
|
||||
// rangesArr.push(proposedReturn);
|
||||
// }
|
||||
// },
|
||||
};
|
||||
}
|
||||
}
|
||||
@@ -1,25 +1,12 @@
|
||||
import _ from "lodash";
|
||||
import { Document } from "../Node";
|
||||
import { CompleteFileSystem, walk } from "../storage/FileSystem";
|
||||
import { DEFAULT_FS } from "../storage/constants";
|
||||
import { CompleteFileSystem, walk } from "../storage/FileSystem";
|
||||
import { BaseReader } from "./base";
|
||||
import { PapaCSVReader } from "./CSVReader";
|
||||
import { DocxReader } from "./DocxReader";
|
||||
import { HTMLReader } from "./HTMLReader";
|
||||
import { MarkdownReader } from "./MarkdownReader";
|
||||
import { PDFReader } from "./PDFReader";
|
||||
import { BaseReader } from "./base";
|
||||
|
||||
type ReaderCallback = (
|
||||
category: "file" | "directory",
|
||||
name: string,
|
||||
status: ReaderStatus,
|
||||
message?: string,
|
||||
) => boolean;
|
||||
enum ReaderStatus {
|
||||
STARTED = 0,
|
||||
COMPLETE,
|
||||
ERROR,
|
||||
}
|
||||
|
||||
/**
|
||||
* Read a .txt file
|
||||
@@ -34,14 +21,12 @@ export class TextFileReader implements BaseReader {
|
||||
}
|
||||
}
|
||||
|
||||
export const FILE_EXT_TO_READER: Record<string, BaseReader> = {
|
||||
const FILE_EXT_TO_READER: Record<string, BaseReader> = {
|
||||
txt: new TextFileReader(),
|
||||
pdf: new PDFReader(),
|
||||
csv: new PapaCSVReader(),
|
||||
md: new MarkdownReader(),
|
||||
docx: new DocxReader(),
|
||||
htm: new HTMLReader(),
|
||||
html: new HTMLReader(),
|
||||
};
|
||||
|
||||
export type SimpleDirectoryReaderLoadDataProps = {
|
||||
@@ -52,37 +37,20 @@ export type SimpleDirectoryReaderLoadDataProps = {
|
||||
};
|
||||
|
||||
/**
|
||||
* Read all of the documents in a directory.
|
||||
* By default, supports the list of file types
|
||||
* in the FILE_EXIT_TO_READER map.
|
||||
* Read all of the documents in a directory. Currently supports PDF and TXT files.
|
||||
*/
|
||||
export class SimpleDirectoryReader implements BaseReader {
|
||||
constructor(private observer?: ReaderCallback) {}
|
||||
|
||||
async loadData({
|
||||
directoryPath,
|
||||
fs = DEFAULT_FS as CompleteFileSystem,
|
||||
defaultReader = new TextFileReader(),
|
||||
fileExtToReader = FILE_EXT_TO_READER,
|
||||
}: SimpleDirectoryReaderLoadDataProps): Promise<Document[]> {
|
||||
// Observer can decide to skip the directory
|
||||
if (
|
||||
!this.doObserverCheck("directory", directoryPath, ReaderStatus.STARTED)
|
||||
) {
|
||||
return [];
|
||||
}
|
||||
|
||||
let docs: Document[] = [];
|
||||
for await (const filePath of walk(fs, directoryPath)) {
|
||||
try {
|
||||
const fileExt = _.last(filePath.split(".")) || "";
|
||||
|
||||
// Observer can decide to skip each file
|
||||
if (!this.doObserverCheck("file", filePath, ReaderStatus.STARTED)) {
|
||||
// Skip this file
|
||||
continue;
|
||||
}
|
||||
|
||||
let reader = null;
|
||||
|
||||
if (fileExt in fileExtToReader) {
|
||||
@@ -90,52 +58,16 @@ export class SimpleDirectoryReader implements BaseReader {
|
||||
} else if (!_.isNil(defaultReader)) {
|
||||
reader = defaultReader;
|
||||
} else {
|
||||
const msg = `No reader for file extension of ${filePath}`;
|
||||
console.warn(msg);
|
||||
|
||||
// In an error condition, observer's false cancels the whole process.
|
||||
if (
|
||||
!this.doObserverCheck("file", filePath, ReaderStatus.ERROR, msg)
|
||||
) {
|
||||
return [];
|
||||
}
|
||||
|
||||
console.warn(`No reader for file extension of ${filePath}`);
|
||||
continue;
|
||||
}
|
||||
|
||||
const fileDocs = await reader.loadData(filePath, fs);
|
||||
|
||||
// Observer can still cancel addition of the resulting docs from this file
|
||||
if (this.doObserverCheck("file", filePath, ReaderStatus.COMPLETE)) {
|
||||
docs.push(...fileDocs);
|
||||
}
|
||||
docs.push(...fileDocs);
|
||||
} catch (e) {
|
||||
const msg = `Error reading file ${filePath}: ${e}`;
|
||||
console.error(msg);
|
||||
|
||||
// In an error condition, observer's false cancels the whole process.
|
||||
if (!this.doObserverCheck("file", filePath, ReaderStatus.ERROR, msg)) {
|
||||
return [];
|
||||
}
|
||||
console.error(`Error reading file ${filePath}: ${e}`);
|
||||
}
|
||||
}
|
||||
|
||||
// After successful import of all files, directory completion
|
||||
// is only a notification for observer, cannot be cancelled.
|
||||
this.doObserverCheck("directory", directoryPath, ReaderStatus.COMPLETE);
|
||||
|
||||
return docs;
|
||||
}
|
||||
|
||||
private doObserverCheck(
|
||||
category: "file" | "directory",
|
||||
name: string,
|
||||
status: ReaderStatus,
|
||||
message?: string,
|
||||
): boolean {
|
||||
if (this.observer) {
|
||||
return this.observer(category, name, status, message);
|
||||
}
|
||||
return true;
|
||||
}
|
||||
}
|
||||
|
||||
@@ -73,7 +73,6 @@ describe("SentenceSplitter", () => {
|
||||
let splits = sentenceSplitter.splitText(
|
||||
"This is a sentence. This is another sentence. 1.0",
|
||||
);
|
||||
|
||||
expect(splits).toEqual([
|
||||
"This is a sentence.",
|
||||
"This is another sentence.",
|
||||
|
||||
@@ -3,7 +3,6 @@
|
||||
"esModuleInterop": true,
|
||||
"forceConsistentCasingInFileNames": true,
|
||||
"isolatedModules": true,
|
||||
"module": "esnext",
|
||||
"moduleResolution": "node",
|
||||
"preserveWatchOutput": true,
|
||||
"skipLibCheck": true,
|
||||
|
||||
@@ -8,7 +8,8 @@
|
||||
"eslint-config-next": "^13.4.1",
|
||||
"eslint-config-prettier": "^8.3.0",
|
||||
"eslint-config-turbo": "^1.9.3",
|
||||
"eslint-plugin-react": "7.28.0"
|
||||
"eslint-plugin-react": "7.28.0",
|
||||
"@dqbd/tiktoken": "^1.0.7"
|
||||
},
|
||||
"publishConfig": {
|
||||
"access": "public"
|
||||
|
||||
Generated
+538
-392
File diff suppressed because it is too large
Load Diff
Reference in New Issue
Block a user