Compare commits

...

22 Commits

Author SHA1 Message Date
yisding 2a27e21e00 changeset 2023-11-06 12:40:36 -08:00
yisding 3bc52a1f2c added 3.5 1106 2023-11-06 12:39:52 -08:00
yisding 9806b5a0a9 0.0.33 2023-11-06 10:59:52 -08:00
yisding 201cd0f5fc packages 2023-11-06 10:59:00 -08:00
yisding 5e2e92c11a changeset 2023-11-06 10:51:02 -08:00
yisding d57657599b new openai models from dev day 2023-11-06 10:50:22 -08:00
yisding 995db834b2 0.0.32 2023-11-02 18:05:56 -07:00
yisding dfd22aac46 changeset 2023-10-30 14:00:54 -07:00
yisding 72f62718f1 Merge pull request #160 from mtutty/add-observable-reader
Add observer/callback feature to SimpleDirectoryReader
2023-10-30 13:59:16 -07:00
yisding e938a4d154 minor changes 2023-10-30 13:52:15 -07:00
Michael Tutty 641019262e Add observer/callback feature to SimpleDirectoryReader 2023-10-30 13:52:15 -07:00
yisding fe9056f081 Merge pull request #164 from v4n/main
replace tiktoken with js-tiktoken
2023-10-30 10:56:34 -07:00
V4N fba49b8088 replace tiktoken with js-tiktoken 2023-10-30 10:00:02 -03:00
yisding 6e0ee9ec32 pinning babel/traverse for security 2023-10-26 15:50:55 -07:00
yisding a5e3e10e84 dynamic import of string-strip-html 2023-10-26 15:42:25 -07:00
yisding 99afbdd606 Merge pull request #154 from mtutty/add-html-reader
Add HTMLReader, sample app and HTML file
2023-10-26 15:06:51 -07:00
yisding 90c0b83c34 changeset 2023-10-26 15:04:51 -07:00
yisding 68f9dd1ce1 prettier 2023-10-26 15:04:08 -07:00
yisding 51e4b1de99 add HTMLReader to SimpleDirectoryReader 2023-10-26 15:02:04 -07:00
Michael Tutty 08f091a889 Revert .vscode/settings.json changes 2023-10-26 21:04:55 +00:00
Michael Tutty 692e3cc56e Add HTMLReader to core/src/readers, apps/simple example, and apps/simple/data HTML file 2023-10-26 20:21:59 +00:00
yisding bcfbccc381 0.0.31 2023-10-25 16:52:00 -07:00
26 changed files with 1996 additions and 612 deletions
-5
View File
@@ -1,5 +0,0 @@
---
"llamaindex": patch
---
Give HistoryChatEngine pluggable options (thanks @marcusschiesser)
-5
View File
@@ -1,5 +0,0 @@
---
"llamaindex": patch
---
Add SimilarityPostProcessor (thanks @TomPenguin)
+5
View File
@@ -0,0 +1,5 @@
---
"llamaindex": patch
---
Add support for gpt-3.5-turbo-1106
-5
View File
@@ -1,5 +0,0 @@
---
"llamaindex": patch
---
Added LLMMetadata (thanks @marcusschiesser)
+1
View File
@@ -3,6 +3,7 @@
# dependencies
node_modules
.pnp
.pnpm-store
.pnp.js
# testing
+24
View File
@@ -1,5 +1,29 @@
# simple
## 0.0.31
### Patch Changes
- Updated dependencies [5e2e92c]
- llamaindex@0.0.33
## 0.0.30
### Patch Changes
- Updated dependencies [90c0b83]
- Updated dependencies [dfd22aa]
- llamaindex@0.0.32
## 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
+24
View File
@@ -0,0 +1,24 @@
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 -1
View File
@@ -6,7 +6,7 @@ import readline from "node:readline/promises";
import { ChatMessage, LlamaDeuce, OpenAI } from "llamaindex";
(async () => {
const gpt4 = new OpenAI({ model: "gpt-4", temperature: 0.9 });
const gpt4 = new OpenAI({ model: "gpt-4-vision-preview", temperature: 0.9 });
const l2 = new LlamaDeuce({
model: "Llama-2-70b-chat-4bit",
temperature: 0.9,
+21
View File
@@ -0,0 +1,21 @@
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 -1
View File
@@ -1,7 +1,7 @@
import { OpenAI } from "llamaindex";
(async () => {
const llm = new OpenAI({ model: "gpt-3.5-turbo", temperature: 0.0 });
const llm = new OpenAI({ model: "gpt-3.5-turbo-1106", temperature: 0.1 });
// complete api
const response1 = await llm.complete("How are you?");
+3 -2
View File
@@ -1,5 +1,5 @@
{
"version": "0.0.28",
"version": "0.0.31",
"private": true,
"name": "simple",
"dependencies": {
@@ -9,7 +9,8 @@
"llamaindex": "workspace:*"
},
"devDependencies": {
"@types/node": "^18.18.6"
"@types/node": "^18.18.6",
"ts-node": "^10.9.1"
},
"scripts": {
"lint": "eslint ."
+15
View File
@@ -0,0 +1,15 @@
import { OpenAI } from "llamaindex";
(async () => {
const llm = new OpenAI({ model: "gpt-4-vision-preview", temperature: 0.1 });
// complete api
const response1 = await llm.complete("How are you?");
console.log(response1.message.content);
// chat api
const response2 = await llm.chat([
{ content: "Tell me a joke!", role: "user" },
]);
console.log(response2.message.content);
})();
+1 -1
View File
@@ -1,7 +1,7 @@
import { OpenAI } from "llamaindex";
(async () => {
const llm = new OpenAI({ model: "gpt-3.5-turbo", temperature: 0.0 });
const llm = new OpenAI({ model: "gpt-3.5-turbo-1106", temperature: 0.1 });
// complete api
const response1 = await llm.complete("How are you?");
+2 -1
View File
@@ -28,7 +28,8 @@
},
"pnpm": {
"overrides": {
"trim": "1.0.1"
"trim": "1.0.1",
"@babel/traverse": "7.23.2"
}
}
}
+21
View File
@@ -1,5 +1,26 @@
# llamaindex
## 0.0.33
### Patch Changes
- 5e2e92c: gpt-4-1106-preview and gpt-4-vision-preview from OpenAI dev day
## 0.0.32
### Patch Changes
- 90c0b83: Add HTMLReader (thanks @mtutty)
- dfd22aa: Add observer/filter to the SimpleDirectoryReader (thanks @mtutty)
## 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
+8 -7
View File
@@ -1,34 +1,35 @@
{
"name": "llamaindex",
"version": "0.0.30",
"version": "0.0.33",
"license": "MIT",
"dependencies": {
"@anthropic-ai/sdk": "^0.8.0",
"@anthropic-ai/sdk": "^0.9.0",
"@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",
"notion-md-crawler": "^0.0.2",
"openai": "^4.13.0",
"openai": "^4.15.4",
"papaparse": "^5.4.1",
"pdf-parse": "^1.1.1",
"portkey-ai": "^0.1.13",
"portkey-ai": "^0.1.16",
"rake-modified": "^1.0.8",
"replicate": "^0.20.1",
"tiktoken": "^1.0.10",
"string-strip-html": "^13.4.3",
"uuid": "^9.0.1",
"wink-nlp": "^1.14.3"
},
"devDependencies": {
"@types/lodash": "^4.14.200",
"@types/node": "^18.18.6",
"@types/node": "^18.18.8",
"@types/papaparse": "^5.3.10",
"@types/pdf-parse": "^1.1.3",
"@types/uuid": "^9.0.6",
"node-stdlib-browser": "^1.2.0",
"tsup": "^7.2.0",
"typescript": "^4.9.5"
"typescript": "^5.2.2"
},
"engines": {
"node": ">=18.0.0"
+10 -12
View File
@@ -1,5 +1,4 @@
import cl100k_base from "tiktoken/encoders/cl100k_base.json";
import { Tiktoken } from "tiktoken/lite";
import { encodingForModel, TiktokenModel } from "js-tiktoken";
import { v4 as uuidv4 } from "uuid";
import { Event, EventTag, EventType } from "./callbacks/CallbackManager";
@@ -18,18 +17,17 @@ class GlobalsHelper {
} | null = null;
private initDefaultTokenizer() {
const encoding = new Tiktoken(
cl100k_base.bpe_ranks,
cl100k_base.special_tokens,
cl100k_base.pat_str,
);
const encoding = encodingForModel("text-embedding-ada-002"); // cl100k_base
this.defaultTokenizer = {
encode: (text: string) => {
return encoding.encode(text);
return new Uint32Array(encoding.encode(text));
},
decode: (tokens: Uint32Array) => {
return new TextDecoder().decode(encoding.decode(tokens));
const numberArray = Array.from(tokens);
const text = encoding.decode(numberArray);
const uint8Array = new TextEncoder().encode(text);
return new TextDecoder().decode(uint8Array);
},
};
}
@@ -41,10 +39,10 @@ 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`);
@@ -52,7 +50,7 @@ class GlobalsHelper {
if (!this.defaultTokenizer) {
this.initDefaultTokenizer();
}
return this.defaultTokenizer!.decode.bind(this.defaultTokenizer);
}
+1
View File
@@ -18,6 +18,7 @@ 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";
+6 -3
View File
@@ -102,11 +102,14 @@ export interface LLM {
export const GPT4_MODELS = {
"gpt-4": { contextWindow: 8192 },
"gpt-4-32k": { contextWindow: 32768 },
"gpt-4-1106-preview": { contextWindow: 128000 },
"gpt-4-vision-preview": { contextWindow: 8192 },
};
export const TURBO_MODELS = {
export const GPT35_MODELS = {
"gpt-3.5-turbo": { contextWindow: 4096 },
"gpt-3.5-turbo-16k": { contextWindow: 16384 },
"gpt-3.5-turbo-1106": { contextWindow: 16384 },
};
/**
@@ -114,7 +117,7 @@ export const TURBO_MODELS = {
*/
export const ALL_AVAILABLE_OPENAI_MODELS = {
...GPT4_MODELS,
...TURBO_MODELS,
...GPT35_MODELS,
};
/**
@@ -648,7 +651,7 @@ export class Anthropic implements LLM {
this.callbackManager = init?.callbackManager;
}
tokens(messages: ChatMessage[]): number {
throw new Error("Method not implemented.");
}
+4 -1
View File
@@ -24,7 +24,10 @@ export class OpenAISession {
if (options.azure) {
this.openai = new AzureOpenAI(options);
} else {
this.openai = new OpenAI(options);
this.openai = new OpenAI({
...options,
// defaultHeaders: { "OpenAI-Beta": "assistants=v1" },
});
}
}
}
+77
View File
@@ -0,0 +1,77 @@
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);
// }
// },
};
}
}
@@ -1,12 +1,25 @@
import _ from "lodash";
import { Document } from "../Node";
import { DEFAULT_FS } from "../storage/constants";
import { CompleteFileSystem, walk } from "../storage/FileSystem";
import { BaseReader } from "./base";
import { DEFAULT_FS } from "../storage/constants";
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
@@ -21,12 +34,14 @@ export class TextFileReader implements BaseReader {
}
}
const FILE_EXT_TO_READER: Record<string, BaseReader> = {
export 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 = {
@@ -37,20 +52,37 @@ export type SimpleDirectoryReaderLoadDataProps = {
};
/**
* Read all of the documents in a directory. Currently supports PDF and TXT files.
* Read all of the documents in a directory.
* By default, supports the list of file types
* in the FILE_EXIT_TO_READER map.
*/
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) {
@@ -58,16 +90,52 @@ export class SimpleDirectoryReader implements BaseReader {
} else if (!_.isNil(defaultReader)) {
reader = defaultReader;
} else {
console.warn(`No reader for file extension of ${filePath}`);
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 [];
}
continue;
}
const fileDocs = await reader.loadData(filePath, fs);
docs.push(...fileDocs);
// Observer can still cancel addition of the resulting docs from this file
if (this.doObserverCheck("file", filePath, ReaderStatus.COMPLETE)) {
docs.push(...fileDocs);
}
} catch (e) {
console.error(`Error reading file ${filePath}: ${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 [];
}
}
}
// 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;
}
}
@@ -73,6 +73,7 @@ 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.",
+1
View File
@@ -3,6 +3,7 @@
"esModuleInterop": true,
"forceConsistentCasingInFileNames": true,
"isolatedModules": true,
"module": "esnext",
"moduleResolution": "node",
"preserveWatchOutput": true,
"skipLibCheck": true,
+608 -561
View File
File diff suppressed because it is too large Load Diff