feat: add feature parity check tests [RET-2382] [RET-2186] (#13)

* chore: add gripmock dockerfile

* feat: add one ranker stub

* feat: add stub generator

* feat: add batch inference for stub generation

* feat: add batch inference stub embedder

* feat: use monkey patch stub generator for chunker

* feat: add all chunker stubs

* feat: add ranker generator

* refactor: cleanup embedder generator

* refactor: cleanup ranker generator

* chore: remove duplicated stubs

* refactor: remove unused request

* feat: add test ranker with mock

* feat: add stub for go

* fix: add tests and fix id for javascript

* feat: add js embedder tests

* refactor: cleanup chunker go tests

* feat: add embedder stubs

* feat: add ci with mock server

* chore: remove wrong condition on go workflow

* chore: fix go workflow workdir

* chore: rename go workflow

* chore: run mock server on the same go job

* chore: run mock server on the same go job

* chore: run mock server on the same go job

* chore: test stubs ci

* chore: only push stubs on merge

* feat: add variable embedder generator

* feat: add variable embedder generator

* feat: add 368 embedding stubs

* feat: add 368 embedding test for js

* feat: remove unnecessary file

* feat: test ranker mock server python

* feat: stubs

* feat: move files for better structure

* feat(js): make id be present only on the request

* feat: rename test to tests

* feat: remove stubs from generator

* feat: move docker compose to grpc

* fix: rectify stubs path

* fix: rectify stubs path

* fix: rectify stubs and proto path

* feat: add beginning of test generator

* feat: add ranker test generation for all

* feat: iterate through all template of function

* feat: remove placeholder comments

* feat: generate embedder test for py

* fix: fixes embedder dimension change

* doc: add doctest to reshaper

* doc: add doctest to reshaper

* feat: add embedder template for js

* feat: make naming consistent

* feat: add go template for embedder

* feat: generate all embedder tests

* feat: add chunker tests python

* feat: add go chunker test generation

* feat: use generated tests

* feat: add make file to generate stubs and tests

* feat: update stubs and tests to generated ones

* feat: use formatters on tests generated

* feat: run pre-commit for new python tests

* feat: add chunker js [RET-2186] (#14)

* feat: add stub for go

* fix: add tests and fix id for javascript

* refactor: cleanup chunker go tests

* feat: add 368 embedding stubs

* feat: move files for better structure

* feat: remove stubs from generator

* feat(js): add chunker

* refactor: add build request in requester

* refactor(js): add process response in requester

* feat(js): unpack bytes correctly as strings

* feat: remove debug logging

* feat: add playground chunker

* feat(js): add initial chunker test file

* feat(js): add chunker test generator

* feat(js): add chunker jinja template

* feat(js): add complete test to js

* feat: copy chunker tests

* feat(js): add postprocess new tests

* feat: revert workflow changes

* feat: revert embedder

* feat: remove playground and add readme

* feat: update stubs to latest shape

* refactor: remove id from input

* feat: make functions private in requester

* refactor: replace if else by switch

* chore: move id to requester in python

---------

Co-authored-by: Daniel Buades Marcos <[email protected]>

Author: clinia-brando

README.md

@@ -36,17 +36,66 @@ You can now import the Clinia Models client in your project and play with it.
 
 ## Playground Examples
 
-### Embedder Model
+### Embedder
 
-```typescript
-// TODO
-```
+``` typescript
+import { embedder } from '@clinia/models-client-embedder';
+
+async function runEmbedderExample() {
+  const myEmbedder = embedder({
+    host: {
+      Url: '127.0.0.1',
+      Scheme: 'http',
+      Port: 8001,
+    },
+  });
 
-### Ranker Model
+  const result = await myEmbedder.embed(
+    'embedder_medical_journals_qa',
+    '120240905185426',
+    {
+      texts: ['Clinia is based in Montreal'],
+      id: 'request-123',
+    },
+  );
+
+  console.log(JSON.stringify(result, null, 2));
+}
+
+runEmbedderExample().catch(console.error);
+```
 
+### Chunker
 ```typescript
+import { chunker } from '@clinia/models-client-chunker';
+
+async function runChunkerExample() {
+  const myChunker = chunker({
+    host: {
+      Url: '127.0.0.1',
+      Scheme: 'http',
+      Port: 8001,
+    },
+  });
+
+  const result = await myChunker.chunk(
+    'chunker',
+    '120252801110000',
+    {
+      texts: ['Clinia is based in Montreal'],
+      id: 'request-123',
+    },
+  );
+
+  console.log(JSON.stringify(result, null, 2));
+}
+
+runChunkerExample().catch(console.error);
+```
+
+### Ranker
+``` typescript
 import { ranker } from '@clinia/models-client-ranker';
-import { v4 as uuidv4 } from 'uuid';
 
 async function runRankerExample() {
   const myRanker = ranker({
@@ -57,34 +106,22 @@ async function runRankerExample() {
     },
   });
 
-  // Get model name and version from environment variables.
-  const modelName = process.env.CLINIA_MODEL_NAME;
-  const modelVersion = process.env.CLINIA_MODEL_VERSION;
-  if (!modelName || !modelVersion) {
-    throw new Error('Missing required environment variables: CLINIA_MODEL_NAME or CLINIA_MODEL_VERSION');
-  }
-
-  const rankRequest = {
-    id: uuidv4(),
-    query: 'Where is Clinia based?',
-    texts: ['Clinia is based in Montreal'],
-  };
-
-  const result = await myRanker.rank(modelName, modelVersion, rankRequest);
-  console.log('Rank result:', result);
+  const result = await myRanker.rank(
+    'ranker_medical_journals_qa',
+    '120240905185925',
+    {
+      query: 'hello, how are you?',
+      texts: ['Clinia is based in Montreal'],
+      id: 'request-123',
+    },
+  );
 
   console.log('Rank result:', result);
 }
 
 runRankerExample().catch(console.error);
 ```
 
-### Chunker Model
-
-```typescript
-// TODO
-```
-
 ## Note
 
 This repository is automatically generated from a private repository within Clinia that contains additional resources including tests, mock servers, and development tools.

packages/models-client-chunker/LICENSE

@@ -0,0 +1 @@
+../../LICENSE

packages/models-client-chunker/builds/browser.ts

@@ -0,0 +1,9 @@
+import type { RequesterConfig } from '@clinia/models-client-common';
+import { Chunker } from '../src/chunker';
+import { createGrpcRequester } from '@clinia/models-requester-grpc';
+
+export const chunker = (options: RequesterConfig): Chunker => {
+  return new Chunker({
+    requester: createGrpcRequester(options),
+  });
+};

packages/models-client-chunker/builds/node.ts

@@ -0,0 +1,9 @@
+import type { RequesterConfig } from '@clinia/models-client-common';
+import { Chunker } from '../src/chunker';
+import { createGrpcRequester } from '@clinia/models-requester-grpc';
+
+export const chunker = (options: RequesterConfig): Chunker => {
+  return new Chunker({
+    requester: createGrpcRequester(options),
+  });
+};

packages/models-client-chunker/index.d.ts

@@ -0,0 +1,2 @@
+// eslint-disable-next-line import/no-unresolved
+export * from './dist/builds/node';

packages/models-client-chunker/index.js

@@ -0,0 +1,2 @@
+// eslint-disable-next-line import/no-commonjs,import/extensions
+module.exports = require('./dist/models-client-chunker.cjs');

packages/models-client-chunker/package.json

@@ -0,0 +1,63 @@
+{
+  "name": "@clinia/models-client-chunker",
+  "version": "1.0.0",
+  "description": "Javascript client for Clinia embed model",
+  "repository": {
+    "type": "git",
+    "url": "git+https://github.com/clinia/models-client-javascript.git"
+  },
+  "license": "MIT",
+  "author": "Clinia",
+  "type": "module",
+  "exports": {
+    ".": {
+      "node": {
+        "types": {
+          "import": "./dist/node.d.ts",
+          "module": "./dist/node.d.ts",
+          "require": "./dist/node.d.cts"
+        },
+        "import": "./dist/builds/node.js",
+        "module": "./dist/builds/node.js",
+        "require": "./dist/builds/node.cjs"
+      },
+      "default": {
+        "types": "./dist/browser.d.ts",
+        "module": "./dist/builds/browser.js",
+        "import": "./dist/builds/browser.js",
+        "default": "./dist/builds/browser.umd.js"
+      }
+    },
+    "./dist/builds/*": "./dist/builds/*.js"
+  },
+  "jsdelivr": "./dist/builds/browser.umd.js",
+  "unpkg": "./dist/builds/browser.umd.js",
+  "react-native": "./dist/builds/browser.js",
+  "files": [
+    "dist",
+    "index.js",
+    "index.d.ts"
+  ],
+  "scripts": {
+    "build": "yarn clean && yarn tsup",
+    "clean": "rm -rf ./dist || true",
+    "test": "tsc --noEmit && vitest --run",
+    "test:bundle": "publint . && attw --pack ."
+  },
+  "dependencies": {
+    "@clinia/models-client-common": "1.0.0",
+    "@clinia/models-requester-grpc": "1.0.0"
+  },
+  "devDependencies": {
+    "@arethetypeswrong/cli": "0.17.3",
+    "@types/node": "22.10.7",
+    "publint": "0.3.2",
+    "rollup": "4.30.1",
+    "tsup": "8.3.5",
+    "typescript": "5.7.3",
+    "vitest": "^3.0.5"
+  },
+  "engines": {
+    "node": ">= 14.0.0"
+  }
+}

packages/models-client-chunker/src/chunker.ts

@@ -0,0 +1,96 @@
+import {
+  type ClientOptions,
+  type Requester,
+  type Input,
+  type Datatype,
+  type Output,
+  getOutputStringContents,
+} from '@clinia/models-client-common';
+
+export type Chunk = {
+  id: string;
+  text: string;
+  startIndex: number;
+  endIndex: number;
+  tokenCount: number;
+};
+
+export type ChunkRequest = {
+  id: string;
+  texts: string[];
+};
+
+export type ChunkResponse = {
+  id: string;
+  chunks: Chunk[][];
+};
+
+const CHUNK_INPUT_KEY = 'text';
+const CHUNK_OUTPUT_KEY = 'chunk';
+const CHUNK_INPUT_DATATYPE: Datatype = 'BYTES';
+
+export class Chunker {
+  private _requester: Requester;
+
+  constructor(options: ClientOptions) {
+    this._requester = options.requester;
+  }
+
+  private processOutput(output: Output): Chunk[][] {
+    let textChunks = getOutputStringContents(output);
+    let formattedChunks = [];
+    for (let text_chunk of textChunks) {
+      // Filter out "pad" values and then map
+      let filtered = text_chunk.filter((chunk) => chunk !== 'pad');
+      let formatted_text_chunks = filtered.map((chunk) => {
+        try {
+          return JSON.parse(chunk) as Chunk;
+        } catch (e) {
+          throw new Error(`Invalid JSON: ${chunk}`);
+        }
+      });
+      formattedChunks.push(formatted_text_chunks);
+    }
+
+    return formattedChunks;
+  }
+
+  async chunk(
+    modelName: string,
+    modelVersion: string,
+    request: ChunkRequest,
+  ): Promise<ChunkResponse> {
+    const inputs: Input[] = [
+      {
+        name: CHUNK_INPUT_KEY,
+        shape: [request.texts.length],
+        datatype: CHUNK_INPUT_DATATYPE,
+        contents: [
+          {
+            stringContents: request.texts,
+          },
+        ],
+      },
+    ];
+
+    // The chunker model has only one input and one output
+    const outputKeys = [CHUNK_OUTPUT_KEY];
+
+    const outputs = await this._requester.infer(
+      modelName,
+      modelVersion,
+      inputs,
+      outputKeys,
+      request.id,
+    );
+
+    // Since we have only one output, we can directly access the first output.
+    // We already check the size of the output in the infer function therefore we can "safely" access the element 0.
+    const chunks = this.processOutput(outputs[0]);
+
+    return {
+      id: request.id,
+      chunks,
+    };
+  }
+}

packages/models-client-chunker/tsconfig.json

@@ -0,0 +1,10 @@
+{
+  "extends": "../../tsconfig.json",
+  "compilerOptions": {
+    "types": ["node", "vitest/globals"],
+    "outDir": "dist",
+    "skipLibCheck": true
+  },
+  "include": ["src", "model", "builds"],
+  "exclude": ["dist", "node_modules"]
+}

packages/models-client-chunker/tsup.config.ts

@@ -0,0 +1,42 @@
+import type { Options } from 'tsup';
+import { defineConfig } from 'tsup';
+
+import { getBaseBrowserOptions, getBaseNodeOptions } from '../../base.tsup.config';
+
+import pkg from './package.json' with { type: 'json' };
+
+const nodeOptions: Options = {
+  ...getBaseNodeOptions(pkg, __dirname),
+  dts: { entry: { node: 'builds/node.ts' } },
+  entry: ['builds/node.ts', 'src/*.ts'],
+};
+
+const nodeConfigs: Options[] = [
+  {
+    ...nodeOptions,
+    format: 'cjs',
+    name: `node ${pkg.name} cjs`,
+  },
+  {
+    ...nodeOptions,
+    format: 'esm',
+    name: `node ${pkg.name} esm`,
+  },
+];
+
+const browserOptions: Options = {
+  ...getBaseBrowserOptions(pkg, __dirname),
+  globalName: 'chunker',
+};
+
+const browserConfigs: Options[] = [
+  {
+    ...browserOptions,
+    minify: false,
+    name: `browser ${pkg.name} esm`,
+    dts: { entry: { browser: 'builds/browser.ts' } },
+    entry: ['builds/browser.ts', 'src/*.ts'],
+  },
+];
+
+export default defineConfig([...nodeConfigs, ...browserConfigs]);

packages/models-client-chunker/vitest.config.ts

@@ -0,0 +1,5 @@
+import { defineConfig } from 'vitest/config';
+
+export default defineConfig({
+  test: { environment: 'jsdom', globals: true },
+});

packages/models-client-common/src/output.ts

@@ -20,3 +20,16 @@ export const getOutputFp32Contents = (output: Output): Float32Array[] => {
 
   return contents;
 };
+
+export const getOutputStringContents = (output: Output): string[][] => {
+  if (output.datatype !== 'BYTES') {
+    throw new Error('Data type not supported');
+  }
+
+  const contents: string[][] = [];
+  for (const content of output.contents) {
+    if (content.stringContents) contents.push(content.stringContents);
+  }
+
+  return contents;
+};