run-llama/LlamaIndexTS
1
0
Fork 0
mirror of https://github.com/run-llama/LlamaIndexTS.git synced 2026-07-16 07:14:29 -04:00
Code Issues 138 Packages Projects Releases 10 Wiki Activity

feat: Voyage embeddings (#1574)

Browse Source 
Co-authored-by: Marcus Schiesser <mail@marcusschiesser.de>
This commit is contained in:
Brian Lange
2025-02-26 21:22:33 -06:00
committed by GitHub
parent 1914b52708
commit 034639153b
10 changed files with 373 additions and 2 deletions
Download Patch File Download Diff File Expand all files Collapse all files
.changeset/tender-candles-shop.md
+5
View File
@@ -0,0 +1,5 @@
---
"@llamaindex/voyage-ai": major
---
Adding VoyageAI embedding package
CONTRIBUTING.md
+7
View File
@@ -41,8 +41,15 @@ pnpm install
### Build the packages
You'll need Turbo to build the packages. If you don't have it, you can run it with `pnpx`.
To build all packages, run:
```shell
# Build all packages
pnpx turbo build --filter "./packages/*"
# Or if you have turbo installed, you can run:
turbo build --filter "./packages/*"
```
apps/next/src/content/docs/llamaindex/modules/embeddings/available_embeddings/voyageai.mdx
+46
View File
@@ -0,0 +1,46 @@
---
title: VoyageAI
---
To use VoyageAI embeddings, you need to import `VoyageAIEmbedding` from `@llamaindex/voyage-ai`.
## Installation
import { Tab, Tabs } from "fumadocs-ui/components/tabs";
<Tabs groupId="install" items={["npm", "yarn", "pnpm"]} persist>
```shell tab="npm"
npm install llamaindex @llamaindex/voyage-ai
```
```shell tab="yarn"
yarn add llamaindex @llamaindex/voyage-ai
```
```shell tab="pnpm"
pnpm add llamaindex @llamaindex/voyage-ai
```
</Tabs>
```ts
import { VoyageAIEmbedding } from "@llamaindex/voyage-ai";
import { Document, Settings, VectorStoreIndex } from "llamaindex";
Settings.embedModel = new VoyageAIEmbedding();
const document = new Document({ text: essay, id_: "essay" });
const index = await VectorStoreIndex.fromDocuments([document]);
const queryEngine = index.asQueryEngine();
const query = "What is the meaning of life?";
const results = await queryEngine.query({
query,
});
```
## API Reference
- [VoyageAIEmbedding](/docs/api/classes/VoyageAIEmbedding)
examples/package.json
+1
View File
@@ -41,6 +41,7 @@
"@llamaindex/upstash": "^0.0.7",
"@llamaindex/vercel": "^0.0.13",
"@llamaindex/vllm": "^0.0.24",
"@llamaindex/voyage-ai": "^0.0.1",
"@llamaindex/weaviate": "^0.0.7",
"@llamaindex/workflow": "^0.0.11",
"@notionhq/client": "^2.2.15",
examples/voyage-ai/embedding.ts
+17
View File
@@ -0,0 +1,17 @@
import { VoyageAIEmbedding } from "@llamaindex/voyage-ai";
async function main() {
// API token can be provided as an environment variable too
// using VOYAGE_API_TOKEN variable
const apiKey = process.env.VOYAGE_API_TOKEN ?? "YOUR_API_TOKEN";
const model = "voyage-3-lite";
const embedModel = new VoyageAIEmbedding({
model,
apiKey,
});
const texts = ["hello", "world"];
const embeddings = await embedModel.getTextEmbeddingsBatch(texts);
console.log(`\nWe have ${embeddings.length} embeddings`);
}
main().catch(console.error);
packages/providers/voyage-ai/package.json
+40
View File
@@ -0,0 +1,40 @@
{
"name": "@llamaindex/voyage-ai",
"description": "VoyageAI Adapter for LlamaIndex",
"version": "0.0.1",
"type": "module",
"main": "./dist/index.cjs",
"module": "./dist/index.js",
"exports": {
".": {
"require": {
"types": "./dist/index.d.cts",
"default": "./dist/index.cjs"
},
"import": {
"types": "./dist/index.d.ts",
"default": "./dist/index.js"
}
}
},
"files": [
"dist"
],
"repository": {
"type": "git",
"url": "https://github.com/run-llama/LlamaIndexTS.git",
"directory": "packages/providers/voyage-ai"
},
"scripts": {
"build": "bunchee",
"dev": "bunchee --watch"
},
"devDependencies": {
"bunchee": "6.0.3"
},
"dependencies": {
"@llamaindex/core": "workspace:*",
"@llamaindex/env": "workspace:*",
"voyageai": "0.0.3-1"
}
}
packages/providers/voyage-ai/src/embedding.ts
+137
View File
@@ -0,0 +1,137 @@
import { BaseEmbedding } from "@llamaindex/core/embeddings";
import type { MessageContentDetail } from "@llamaindex/core/llms";
import { extractSingleText } from "@llamaindex/core/utils";
import { getEnv } from "@llamaindex/env";
import { VoyageAI, VoyageAIClient } from "voyageai";
const DEFAULT_MODEL = "voyage-3";
const API_TOKEN_ENV_VARIABLE_NAME = "VOYAGE_API_TOKEN";
// const API_ROOT = "https://api.voyageai.com/v1/embeddings";
const DEFAULT_TIMEOUT = 60 * 1000;
const DEFAULT_MAX_RETRIES = 5;
/**
* VoyageAIEmbedding is an alias for VoyageAI that implements the BaseEmbedding interface.
*/
export class VoyageAIEmbedding extends BaseEmbedding {
/**
* VoyageAI model to use
* @default "voyage-3"
* @see https://docs.voyageai.com/docs/embeddings
*/
model: string;
/**
* VoyageAI API token
* @see https://docs.voyageai.com/docs/api-key-and-installation
* If not provided, it will try to get the token from the environment variable `VOYAGE_API_KEY`
*
*/
apiKey: string;
/**
* Maximum number of retries
* @default 5
*/
maxRetries: number;
/**
* Timeout in seconds
* @default 60
*/
timeout: number;
/**
* Whether to truncate the input texts to fit within the context length. Defaults to `true`.
* If `true`, over-length input texts will be truncated to fit within the context length, before vectorized by the embedding model.
* If `false`, an error will be raised if any given text exceeds the context length.
*/
truncation: boolean;
/**
* VoyageAI supports `document` and `query` as input types, or it can be left undefined. Using an input type prepends the input with a prompt before embedding.
* Example from their docs: using "query" adds "Represent the query for retrieving supporting documents:"
* VoyageAI says these types improve performance, but it will add to token usage. Embeddings with input types are compatible with those that don't use them.
* Setting this to `query` will use the `query` input type for getQueryEmbedding(s).
* Setting this to `document` will use the `document` input type for getTextEmbedding(s).
* Setting this to `both` will do both of the above.
* By default, this is undefined, which means no input types are used.
* @see https://docs.voyageai.com/docs/embeddings
* @default undefined
*/
useInputTypes: "query" | "document" | "both" | undefined;
/**
* VoyageAI client
*/
client: VoyageAIClient;
constructor(init?: Partial<VoyageAIEmbedding>) {
super();
this.model = init?.model ?? DEFAULT_MODEL;
this.apiKey = init?.apiKey ?? getEnv(API_TOKEN_ENV_VARIABLE_NAME) ?? "";
this.maxRetries = init?.maxRetries ?? DEFAULT_MAX_RETRIES;
this.timeout = init?.timeout ?? DEFAULT_TIMEOUT;
this.truncation = init?.truncation ?? true;
this.useInputTypes = init?.useInputTypes;
this.client = new VoyageAIClient({
apiKey: this.apiKey,
});
}
async getTextEmbedding(text: string): Promise<number[]> {
const embeddings = await this.getVoyageAIEmbedding([text], "document");
return embeddings[0]!;
}
async getQueryEmbedding(
query: MessageContentDetail,
): Promise<number[] | null> {
const text = extractSingleText(query);
if (text) {
const embeddings = await this.getVoyageAIEmbedding([text], "query");
return embeddings[0]!;
} else {
return null;
}
}
getTextEmbeddings = async (texts: string[]): Promise<number[][]> => {
return this.getVoyageAIEmbedding(texts, "document");
};
async getQueryEmbeddings(queries: string[]): Promise<number[][]> {
return this.getVoyageAIEmbedding(queries, "query");
}
private getInputType(requestType: "query" | "document") {
if (this.useInputTypes === "both") {
return requestType;
} else if (this.useInputTypes === requestType) {
return requestType;
} else {
return undefined;
}
}
private async getVoyageAIEmbedding(
inputs: VoyageAI.EmbedRequestInput,
inputType: VoyageAI.EmbedRequestInputType,
): Promise<number[][]> {
const request: VoyageAI.EmbedRequest = {
model: this.model,
input: inputs,
truncation: this.truncation,
};
const preferredInputType = this.getInputType(inputType);
if (preferredInputType) {
request.inputType = preferredInputType;
}
const response = await this.client.embed(request);
if (response.data) {
return response.data.map((item) => item.embedding ?? []);
} else {
throw new Error("Failed to get embeddings from VoyageAI");
}
}
}
packages/providers/voyage-ai/src/index.ts
+1
View File
@@ -0,0 +1 @@
export { VoyageAIEmbedding } from "./embedding";
packages/providers/voyage-ai/tsconfig.json
+19
View File
@@ -0,0 +1,19 @@
{
"extends": "../../../tsconfig.json",
"compilerOptions": {
"target": "ESNext",
"module": "ESNext",
"moduleResolution": "bundler",
"outDir": "./lib",
"tsBuildInfoFile": "./lib/.tsbuildinfo"
},
"include": ["./src"],
"references": [
{
"path": "../openai/tsconfig.json"
},
{
"path": "../../env/tsconfig.json"
}
]
}
pnpm-lock.yaml
Generated
+100 -2
View File
@@ -685,6 +685,9 @@ importers:
'@llamaindex/vllm':
specifier: ^0.0.24
version: link:../packages/providers/vllm
'@llamaindex/voyage-ai':
specifier: ^0.0.1
version: link:../packages/providers/voyage-ai
'@llamaindex/weaviate':
specifier: ^0.0.7
version: link:../packages/providers/storage/weaviate
@@ -1443,7 +1446,7 @@ importers:
version: link:../../../env
chromadb:
specifier: 1.10.3
version: 1.10.3(cohere-ai@7.14.0)(openai@4.83.0(ws@8.18.0(bufferutil@4.0.9))(zod@3.24.2))
version: 1.10.3(cohere-ai@7.14.0)(openai@4.83.0(ws@8.18.0(bufferutil@4.0.9))(zod@3.24.2))(voyageai@0.0.3-1)
chromadb-default-embed:
specifier: ^2.13.2
version: 2.13.2
@@ -1636,6 +1639,22 @@ importers:
specifier: 6.3.4
version: 6.3.4(patch_hash=pavboztthlgni7m5gzw7643oru)(typescript@5.7.3)
packages/providers/voyage-ai:
dependencies:
'@llamaindex/core':
specifier: workspace:*
version: link:../../core
'@llamaindex/env':
specifier: workspace:*
version: link:../../env
voyageai:
specifier: 0.0.3-1
version: 0.0.3-1
devDependencies:
bunchee:
specifier: 6.0.3
version: 6.0.3(typescript@5.7.3)
packages/readers:
dependencies:
'@azure/cosmos':
@@ -4263,6 +4282,15 @@ packages:
rollup:
optional: true
'@rollup/plugin-node-resolve@15.3.1':
resolution: {integrity: sha512-tgg6b91pAybXHJQMAAwW9VuWBO6Thi+q7BCNARLwSqlmsHz0XYURtGvh/AuwSADXSI4h/2uHbs7s4FzlZDGSGA==}
engines: {node: '>=14.0.0'}
peerDependencies:
rollup: ^2.78.0||^3.0.0||^4.0.0
peerDependenciesMeta:
rollup:
optional: true
'@rollup/plugin-node-resolve@16.0.0':
resolution: {integrity: sha512-0FPvAeVUT/zdWoO0jnb/V5BlBsUSNfkIOtFHzMO4H9MOklrmQFY6FduVHKucNb/aTFxvnGhj4MNj/T1oNdDfNg==}
engines: {node: '>=14.0.0'}
@@ -5926,6 +5954,16 @@ packages:
resolution: {integrity: sha512-WDtdLmJvAuNNPzByAYpRo2rF1Mmradw6gvWsQKf63476DDXmomT9zUiGypLcG4ibIM67vhAj8jJRdbmEws2Aqw==}
engines: {node: '>=6.14.2'}
bunchee@6.0.3:
resolution: {integrity: sha512-Yq/srd3ocXPAHv0KEdJvhFMNUOOVVqy0kNzaGVCirk/+MfnLdvZO5uf5BHugIHe/qSvWUQTJZ3SAfB/VABONeQ==}
engines: {node: '>= 18.0.0'}
hasBin: true
peerDependencies:
typescript: ^4.1 || ^5.0
peerDependenciesMeta:
typescript:
optional: true
bunchee@6.3.4:
resolution: {integrity: sha512-bMy2/+tdMPXOqBAX+9BI0HTNjOXOZ2TXjgFpp5Prt0ztP15xQQUcsECnU7wuBPpLH+4id3rXakH9icdbBRZHZQ==}
engines: {node: '>= 18.0.0'}
@@ -11534,6 +11572,9 @@ packages:
jsdom:
optional: true
voyageai@0.0.3-1:
resolution: {integrity: sha512-R3jN/xnILWoMBL3jPY61Ydm1JbpK3J+VmXBoHvlNg1Xz8h0xdX7sEffXeSu+sAEKQaPyWXVQDmM/jpBhPXw58g==}
vue-demi@0.14.10:
resolution: {integrity: sha512-nMZBOwuzabUO0nLgIcc6rycZEebF6eeUfaiQx9+WSk8e29IbLvPU9feI6tqW4kTo3hvoYAJkMh8n8D0fuISphg==}
engines: {node: '>=12'}
@@ -14803,6 +14844,16 @@ snapshots:
optionalDependencies:
rollup: 4.34.6
'@rollup/plugin-node-resolve@15.3.1(rollup@4.34.6)':
dependencies:
'@rollup/pluginutils': 5.1.4(rollup@4.34.6)
'@types/resolve': 1.20.2
deepmerge: 4.3.1
is-module: 1.0.0
resolve: 1.22.10
optionalDependencies:
rollup: 4.34.6
'@rollup/plugin-node-resolve@16.0.0(rollup@4.34.6)':
dependencies:
'@rollup/pluginutils': 5.1.4(rollup@4.34.6)
@@ -16932,6 +16983,31 @@ snapshots:
dependencies:
node-gyp-build: 4.8.4
bunchee@6.0.3(typescript@5.7.3):
dependencies:
'@rollup/plugin-commonjs': 28.0.2(rollup@4.34.6)
'@rollup/plugin-json': 6.1.0(rollup@4.34.6)
'@rollup/plugin-node-resolve': 15.3.1(rollup@4.34.6)
'@rollup/plugin-replace': 6.0.2(rollup@4.34.6)
'@rollup/plugin-wasm': 6.2.2(rollup@4.34.6)
'@rollup/pluginutils': 5.1.4(rollup@4.34.6)
'@swc/core': 1.10.16(@swc/helpers@0.5.15)
'@swc/helpers': 0.5.15
clean-css: 5.3.3
glob: 11.0.1
magic-string: 0.30.17
ora: 8.2.0
picomatch: 4.0.2
pretty-bytes: 5.6.0
rollup: 4.34.6
rollup-plugin-dts: 6.1.1(rollup@4.34.6)(typescript@5.7.3)
rollup-plugin-swc3: 0.11.2(@swc/core@1.10.16(@swc/helpers@0.5.15))(rollup@4.34.6)
rollup-preserve-directives: 1.1.3(rollup@4.34.6)
tslib: 2.8.1
yargs: 17.7.2
optionalDependencies:
typescript: 5.7.3
bunchee@6.3.4(patch_hash=pavboztthlgni7m5gzw7643oru)(typescript@5.7.2):
dependencies:
'@rollup/plugin-commonjs': 28.0.2(rollup@4.34.6)
@@ -17134,13 +17210,14 @@ snapshots:
transitivePeerDependencies:
- bare-buffer
chromadb@1.10.3(cohere-ai@7.14.0)(openai@4.83.0(ws@8.18.0(bufferutil@4.0.9))(zod@3.24.2)):
chromadb@1.10.3(cohere-ai@7.14.0)(openai@4.83.0(ws@8.18.0(bufferutil@4.0.9))(zod@3.24.2))(voyageai@0.0.3-1):
dependencies:
cliui: 8.0.1
isomorphic-fetch: 3.0.0
optionalDependencies:
cohere-ai: 7.14.0
openai: 4.83.0(ws@8.18.0(bufferutil@4.0.9))(zod@3.24.2)
voyageai: 0.0.3-1
transitivePeerDependencies:
- encoding
@@ -22738,6 +22815,15 @@ snapshots:
rollup: 4.34.6
rollup-preserve-directives: 1.1.3(rollup@4.34.6)
rollup-plugin-swc3@0.11.2(@swc/core@1.10.16(@swc/helpers@0.5.15))(rollup@4.34.6):
dependencies:
'@fastify/deepmerge': 1.3.0
'@rollup/pluginutils': 5.1.4(rollup@4.34.6)
'@swc/core': 1.10.16(@swc/helpers@0.5.15)
get-tsconfig: 4.10.0
rollup: 4.34.6
rollup-preserve-directives: 1.1.3(rollup@4.34.6)
rollup-pluginutils@2.8.2:
dependencies:
estree-walker: 0.6.1
@@ -24326,6 +24412,18 @@ snapshots:
- supports-color
- terser
voyageai@0.0.3-1:
dependencies:
form-data: 4.0.0
formdata-node: 6.0.3
js-base64: 3.7.2
node-fetch: 2.7.0
qs: 6.11.2
readable-stream: 4.7.0
url-join: 4.0.1
transitivePeerDependencies:
- encoding
vue-demi@0.14.10(vue@3.5.13(typescript@5.7.2)):
dependencies:
vue: 3.5.13(typescript@5.7.2)