Skip to content

[openvino-langchain] Use TextEmbeddingPipeline #978

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Merged
Retribution98 merged 3 commits into from
Sep 4, 2025
Merged

[openvino-langchain] Use TextEmbeddingPipeline #978

Changes from all commits
Commits
Show all changes
3 commits
Select commit Hold shift + click to select a range
995d976
[openvino-langchain] Use TextEmbeddingPipeline
Retribution98 Jul 1, 2025
1a17df0
Merge branch 'master' into textEmbeddingPipeline
Retribution98 Jul 18, 2025
b014913
Merge branch 'master' into textEmbeddingPipeline
Retribution98 Sep 3, 2025
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
122 changes: 16 additions & 106 deletions modules/openvino-langchain/src/embeddings.ts
View file
Open in desktop
Original file line number Diff line number Diff line change
@@ -1,34 +1,8 @@
import { join } from 'node:path';
import { addon as ov } from 'openvino-node';
// CVS-146344
type CompiledModel = {
createInferRequest: () => InferRequest;
output: (index: number) => any;
outputs: any[];
inputs: any[];
};
type InferRequest = {
inferAsync: (inputs: any) => Promise<{ [outputName: string]: Tensor }>;
};
type Tensor = {
getShape: () => number[];
data: SupportedTypedArray;
};

type SupportedTypedArray =
| Int8Array
| Uint8Array
| Int16Array
| Uint16Array
| Int32Array
| Uint32Array
| Float32Array
| Float64Array;

// Could not find a declaration file for module.
// It will be fixed in the next release.
// @ts-ignore
import { path as tokenizerExtensionPath } from 'openvino-tokenizers-node';
import {
EmbeddingResult,
EmbeddingResults,
TextEmbeddingPipeline,
} from 'openvino-genai-node';
import { Embeddings, EmbeddingsParams } from '@langchain/core/embeddings';

export interface OvEmbeddingsParams extends EmbeddingsParams {
Expand All @@ -37,91 +11,27 @@ export interface OvEmbeddingsParams extends EmbeddingsParams {
}

export class OpenVINOEmbeddings extends Embeddings {

private tokenizerModelCompiled: Promise<CompiledModel>;

private modelCompiled: Promise<CompiledModel>;
private pipeline: Promise<any>;

constructor(fields: OvEmbeddingsParams) {
super(fields);
const modelPath = join(fields.modelPath, 'openvino_model.xml');
const tokenizerPath = join(fields.modelPath, 'openvino_tokenizer.xml');
const core = new ov.Core();
core.addExtension(tokenizerExtensionPath);
const device = fields?.device || 'CPU';
this.tokenizerModelCompiled = core.readModel(tokenizerPath)
.then(result => core.compileModel(result, device))
.catch((err) => {
console.error('Failed to compile tokenizer model:', err);
throw err;
});
this.modelCompiled = core.compileModel(modelPath, device)
.catch((err) => {
console.error('Failed to compile main model:', err);
throw err;
});
this.pipeline = TextEmbeddingPipeline(
fields.modelPath,
fields?.device || 'CPU',
);
}

async embedDocuments(texts: string[]): Promise<number[][]> {
const tokensArray = [];

for (const text of texts) {
const encodings = await this.ovCall(text);

tokensArray.push(encodings);
}

const embeddings = [];

for (const tokens of tokensArray) {
const embedArray = [];

for (const token of tokens) {
embedArray.push(+token);
}
const pipeline = await this.pipeline;
const result: EmbeddingResults = await pipeline.embedDocuments(texts);

embeddings.push(embedArray);
}

return embeddings;
return Array.from(result.map(x => Array.from(x)));
}

async embedQuery(text: string): Promise<number[]> {
const tokens = [];
const encodings = await this.ovCall(text);

for (const token of encodings) {
tokens.push(+token);
}

return tokens;
}

async ovCall(text: string) {
const tokenizerModelCompiled = await this.tokenizerModelCompiled;
const irTokenizer = tokenizerModelCompiled.createInferRequest();

const inputTensor = new ov.Tensor([text]);
const tokenizedInput = await irTokenizer.inferAsync([inputTensor]);

const modelCompiled = await this.modelCompiled;

const tokenizerOutputs = tokenizerModelCompiled.outputs
.map(o => o.getAnyName())
.sort();
const embeddingInputs = modelCompiled.inputs
.map(i => i.getAnyName())
.sort();
if (!tokenizerOutputs.every(
(value, index) => value === embeddingInputs[index],
)) {
throw Error('Embedding model is incorrect. Tokenizer outputs should be the same as embedding model inputs.');
}
const ir = modelCompiled.createInferRequest();
const embeddings = await ir.inferAsync(tokenizedInput);

const outputName = modelCompiled.output(0);
const pipeline = await this.pipeline;
const result: EmbeddingResult = await pipeline.embedQuery(text);

return embeddings[outputName].data;
return Array.from(result);
}
}
Loading