Read OAI Ft Dataset Method (#245)

Useful for a chat loader

js/package.json

@@ -1,6 +1,6 @@
 {
   "name": "langsmith",
-  "version": "0.0.41",
+  "version": "0.0.42",
   "description": "Client library to connect to the LangSmith LLM Tracing and Evaluation Platform.",
   "files": [
     "dist/",

js/src/client.ts

@@ -1,27 +1,29 @@
 import * as uuid from "uuid";
+
 import { AsyncCaller, AsyncCallerParams } from "./utils/async_caller.js";
 import {
+  DataType,
   Dataset,
   Example,
   ExampleCreate,
   ExampleUpdate,
   Feedback,
   KVMap,
+  LangChainBaseMessage,
   Run,
   RunCreate,
   RunUpdate,
   ScoreType,
   TracerSession,
   TracerSessionResult,
   ValueType,
-  DataType,
-  LangChainBaseMessage,
 } from "./schemas.js";
 import {
   convertLangChainMessageToExample,
   isLangChainMessage,
 } from "./utils/messages.js";
 import { getEnvironmentVariable, getRuntimeEnvironment } from "./utils/env.js";
+
 import { RunEvaluator } from "./evaluation/evaluator.js";
 
 interface ClientConfig {
@@ -235,10 +237,10 @@ export class Client {
     return headers;
   }
 
-  private async _get<T>(
+  private async _getResponse(
     path: string,
     queryParams?: URLSearchParams
-  ): Promise<T> {
+  ): Promise<Response> {
     const paramsString = queryParams?.toString() ?? "";
     const url = `${this.apiUrl}${path}?${paramsString}`;
     const response = await this.caller.call(fetch, url, {
@@ -251,6 +253,14 @@ export class Client {
         `Failed to fetch ${path}: ${response.status} ${response.statusText}`
       );
     }
+    return response;
+  }
+
+  private async _get<T>(
+    path: string,
+    queryParams?: URLSearchParams
+  ): Promise<T> {
+    const response = await this._getResponse(path, queryParams);
     return response.json() as T;
   }
   private async *_getPaginated<T>(
@@ -860,6 +870,30 @@ export class Client {
     return result;
   }
 
+  public async readDatasetOpenaiFinetuning({
+    datasetId,
+    datasetName,
+  }: {
+    datasetId?: string;
+    datasetName?: string;
+  }): Promise<any[]> {
+    let path = "/datasets";
+    if (datasetId !== undefined) {
+      // do nothing
+    } else if (datasetName !== undefined) {
+      datasetId = (await this.readDataset({ datasetName })).id;
+    } else {
+      throw new Error("Must provide datasetName or datasetId");
+    }
+    const response = await this._getResponse(`${path}/${datasetId}/openai_ft`);
+    const datasetText = await response.text();
+    const dataset = datasetText
+      .trim()
+      .split("\n")
+      .map((line: string) => JSON.parse(line));
+    return dataset;
+  }
+
   public async *listDatasets({
     limit = 100,
     offset = 0,

python/langsmith/client.py

@@ -1212,6 +1212,37 @@ def read_dataset(
             return ls_schemas.Dataset(**result[0], _host_url=self._host_url)
         return ls_schemas.Dataset(**result, _host_url=self._host_url)
 
+    def read_dataset_openai_finetuning(
+        self, dataset_id: Optional[str] = None, *, dataset_name: Optional[str] = None
+    ) -> list:
+        """
+        Download a dataset in OpenAI Jsonl format and load it as a list of dicts.
+
+        Parameters
+        ----------
+        dataset_id : str
+            The ID of the dataset to download.
+        dataset_name : str
+            The name of the dataset to download.
+
+        Returns
+        -------
+        list
+            The dataset loaded as a list of dicts.
+        """
+        path = "/datasets"
+        if dataset_id is not None:
+            pass
+        elif dataset_name is not None:
+            dataset_id = self.read_dataset(dataset_name=dataset_name).id
+        else:
+            raise ValueError("Must provide dataset_name or dataset_id")
+        response = self._get_with_retries(
+            f"{path}/{dataset_id}/openai_ft",
+        )
+        dataset = [json.loads(line) for line in response.text.strip().split("\n")]
+        return dataset
+
     def list_datasets(
         self,
         *,

python/pyproject.toml

@@ -1,6 +1,6 @@
 [tool.poetry]
 name = "langsmith"
-version = "0.0.42"
+version = "0.0.43"
 description = "Client library to connect to the LangSmith LLM Tracing and Evaluation Platform."
 authors = ["LangChain <support@langchain.dev>"]
 license = "MIT"