feat(generator): extend construction to any langchain LLM and Embeddings (explodinggradients#670)

## **User description**
The current version of `with_openai` contains a hardcoded instantiation
of `langchain_openai.chat_models.ChatOpenAI`, which makes
`TestsetGenerator` very limited and not compatible with completion
models, Azure OpenAI models, and open-source models.

This PR extends `TestsetGenerator` to any `BaseLanguageModel` and
`Embeddings` from langchain for versatility, addressing explodinggradients#230, explodinggradients#342,
explodinggradients#635, and explodinggradients#636.

Lastly, I've removed all the occurrences of mutable default arguments
(bad antipattern, read
[here](https://docs.python-guide.org/writing/gotchas/#mutable-default-arguments)).

---------

Co-authored-by: Shahules786 <[email protected]>
Co-authored-by: jjmachan <[email protected]>

Author: 3 people

.gitignore

@@ -162,7 +162,7 @@ cython_debug/
 #  be found at https://github.com/github/gitignore/blob/main/Global/JetBrains.gitignore
 #  and can be added to the global gitignore or merged into this file.  For a more nuclear
 #  option (not recommended) you can uncomment the following to ignore the entire idea folder.
-#.idea/
+.idea/
 
 # Ragas specific
 experiments/

docs/alfred.py

@@ -1,14 +1,15 @@
 from __future__ import annotations
 
-import os
-from collections import namedtuple
 import argparse
 import asyncio
-from tqdm.asyncio import tqdm
+import os
 import typing as t
-from langchain_openai.chat_models import ChatOpenAI
-from langchain_core.language_models.chat_models import BaseChatModel
+from collections import namedtuple
+
 from langchain.prompts import ChatPromptTemplate
+from langchain_core.language_models.chat_models import BaseChatModel
+from langchain_openai.chat_models import ChatOpenAI
+from tqdm.asyncio import tqdm
 
 File = namedtuple("File", "name content")
 

docs/concepts/testset_generation.md

@@ -60,11 +60,20 @@ Checkout [llama-index](https://gpt-index.readthedocs.io/en/stable/core_modules/d
 :caption: Customising test data distribution 
 from ragas.testset.generator import TestsetGenerator
 from ragas.testset.evolutions import simple, reasoning, multi_context
+from langchain_openai import ChatOpenAI, OpenAIEmbeddings
 
 # documents = load your documents
 
 # generator with openai models
-generator = TestsetGenerator.with_openai()
+generator_llm = ChatOpenAI(model="gpt-3.5-turbo-16k")
+critic_llm = ChatOpenAI(model="gpt-4")
+embeddings = OpenAIEmbeddings()
+
+generator = TestsetGenerator.from_langchain(
+    generator_llm,
+    critic_llm,
+    embeddings
+)
 
 # Change resulting question type distribution
 distributions = {

docs/getstarted/testset_generation.md

@@ -41,9 +41,18 @@ Now, we'll import and use Ragas' `TestsetGenerator` to quickly generate a synthe
 :caption: Create 10 samples using default configuration
 from ragas.testset.generator import TestsetGenerator
 from ragas.testset.evolutions import simple, reasoning, multi_context
+from langchain_openai import ChatOpenAI, OpenAIEmbeddings
 
 # generator with openai models
-generator = TestsetGenerator.with_openai()
+generator_llm = ChatOpenAI(model="gpt-3.5-turbo-16k")
+critic_llm = ChatOpenAI(model="gpt-4")
+embeddings = OpenAIEmbeddings()
+
+generator = TestsetGenerator.from_langchain(
+    generator_llm,
+    critic_llm,
+    embeddings
+)
 
 # generate testset
 testset = generator.generate_with_langchain_docs(documents, test_size=10, distributions={simple: 0.5, reasoning: 0.25, multi_context: 0.25})

docs/howtos/applications/compare_embeddings.md

@@ -29,14 +29,24 @@ For this tutorial notebook, I am using papers from Semantic Scholar that is rela
 :caption: load documents using llama-hub and create test data
 from llama_index import download_loader
 from ragas.testset.evolutions import simple, reasoning, multi_context
+from langchain_openai import ChatOpenAI, OpenAIEmbeddings
 
 SemanticScholarReader = download_loader("SemanticScholarReader")
 loader = SemanticScholarReader()
 query_space = "large language models"
 documents = loader.load_data(query=query_space, limit=100)
 
 # generator with openai models
-generator = TestsetGenerator.with_openai()
+generator_llm = ChatOpenAI(model="gpt-3.5-turbo-16k")
+critic_llm = ChatOpenAI(model="gpt-4")
+embeddings = OpenAIEmbeddings()
+
+generator = TestsetGenerator.from_langchain(
+    generator_llm,
+    critic_llm,
+    embeddings
+)
+
 
 distributions = {
     simple: 0.5,

docs/howtos/applications/compare_llms.md

@@ -35,6 +35,7 @@ from llama_index import download_loader, SimpleDirectoryReader
 from ragas.testset import TestsetGenerator
 from ragas.testset.generator import TestsetGenerator
 from ragas.testset.evolutions import simple, reasoning, multi_context
+from langchain_openai import ChatOpenAI, OpenAIEmbeddings
 
 os.environ['OPENAI_API_KEY'] = 'Your OPEN AI key'
 
@@ -43,7 +44,15 @@ reader = SimpleDirectoryReader("./arxiv-papers/",num_files_limit=30)
 documents = reader.load_data()
 
 # generator with openai models
-generator = TestsetGenerator.with_openai()
+generator_llm = ChatOpenAI(model="gpt-3.5-turbo-16k")
+critic_llm = ChatOpenAI(model="gpt-4")
+embeddings = OpenAIEmbeddings()
+
+generator = TestsetGenerator.from_langchain(
+    generator_llm,
+    critic_llm,
+    embeddings
+)
 
 distributions = {
     simple: 0.5,

docs/howtos/applications/use_prompt_adaptation.md

@@ -125,9 +125,18 @@ Now we can import all the required evolutions and adapt it using `generator.adap
 
 from ragas.testset.generator import TestsetGenerator
 from ragas.testset.evolutions import simple, reasoning, multi_context,conditional
+from langchain_openai import ChatOpenAI, OpenAIEmbeddings
 
 # generator with openai models
-generator = TestsetGenerator.with_openai()
+generator_llm = ChatOpenAI(model="gpt-3.5-turbo-16k")
+critic_llm = ChatOpenAI(model="gpt-4")
+embeddings = OpenAIEmbeddings()
+
+generator = TestsetGenerator.from_langchain(
+    generator_llm,
+    critic_llm,
+    embeddings
+)
 
 # adapt to language
 language = "hindi"

docs/howtos/customisations/azure-openai.ipynb

@@ -7,7 +7,11 @@
    "source": [
     "# Using Azure OpenAI\n",
     "\n",
-    "This tutorial will show you how to use Azure OpenAI endpoints instead of OpenAI endpoints."
+    "This tutorial will show you how to use Azure OpenAI endpoints instead of OpenAI endpoints.\n",
+    "\n",
+    "\n",
+    "- [Evaluation](#load-sample-dataset)\n",
+    "- [Test set generation](#test-set-generation)"
    ]
   },
   {
@@ -416,6 +420,75 @@
     "\n",
     "if you have any suggestion/feedbacks/things your not happy about, please do share it in the [issue section](https://github.com/explodinggradients/ragas/issues). We love hearing from you 😁"
    ]
+  },
+  {
+   "cell_type": "markdown",
+   "id": "3cee41e9",
+   "metadata": {},
+   "source": [
+    "### Test set generation\n",
+    "\n",
+    "Here you will learn how to generate a test set from your dataset using the Azure OpenAI endpoints."
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": null,
+   "id": "aa9ff398",
+   "metadata": {},
+   "outputs": [],
+   "source": [
+    "! git clone https://huggingface.co/datasets/explodinggradients/2023-llm-papers"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": 2,
+   "id": "d935a561",
+   "metadata": {},
+   "outputs": [],
+   "source": [
+    "from langchain.document_loaders import DirectoryLoader\n",
+    "from ragas.testset.generator import TestsetGenerator\n",
+    "from ragas.testset.evolutions import simple, reasoning, multi_context\n",
+    "\n",
+    "\n",
+    "loader = DirectoryLoader(\"./2023-llm-papers/\", use_multithreading=True, silent_errors=True,sample_size=1)\n",
+    "documents = loader.load()\n",
+    "\n",
+    "for document in documents:\n",
+    "    document.metadata['filename'] = document.metadata['source']"
+   ]
+  },
+  {
+   "cell_type": "markdown",
+   "id": "c8f735a7",
+   "metadata": {},
+   "source": [
+    "Use the `azure_model` and `azure_embedding` that we initialized in above section to generate the test set"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": null,
+   "id": "04abc4b1",
+   "metadata": {},
+   "outputs": [],
+   "source": [
+    "generator = TestsetGenerator.from_langchain(generator_llm=azure_model,critic_llm=azure_model,embeddings=azure_embeddings)\n",
+    "\n",
+    "testset = generator.generate_with_langchain_docs(documents, test_size=10, \n",
+    "                                                 raise_exceptions=False, with_debugging_logs=False,\n",
+    "                                                 distributions={simple: 0.5, reasoning: 0.25, multi_context: 0.25})    "
+   ]
+  },
+  {
+   "cell_type": "markdown",
+   "id": "d2f5a7f7",
+   "metadata": {},
+   "source": [
+    "testset.to_pandas()"
+   ]
   }
  ],
  "metadata": {
@@ -434,7 +507,7 @@
    "name": "python",
    "nbconvert_exporter": "python",
    "pygments_lexer": "ipython3",
-   "version": "3.10.12"
+   "version": "3.10.8"
   }
  },
  "nbformat": 4,