Add OpenLM LLM multi-provider (langchain-ai#4993)

OpenLM is a zero-dependency OpenAI-compatible LLM provider that can call
different inference endpoints directly via HTTP. It implements the
OpenAI Completion class so that it can be used as a drop-in replacement
for the OpenAI API. This changeset utilizes BaseOpenAI for minimal added
code.

---------

Co-authored-by: Dev 2049 <[email protected]>

docs/modules/models/llms/integrations/openlm.ipynb

@@ -0,0 +1,133 @@
+{
+ "cells": [
+  {
+   "cell_type": "markdown",
+   "metadata": {},
+   "source": [
+    "# OpenLM\n",
+    "[OpenLM](https://github.com/r2d4/openlm) is a zero-dependency OpenAI-compatible LLM provider that can call different inference endpoints directly via HTTP. \n",
+    "\n",
+    "\n",
+    "It implements the OpenAI Completion class so that it can be used as a drop-in replacement for the OpenAI API. This changeset utilizes BaseOpenAI for minimal added code.\n",
+    "\n",
+    "This examples goes over how to use LangChain to interact with both OpenAI and HuggingFace. You'll need API keys from both."
+   ]
+  },
+  {
+   "cell_type": "markdown",
+   "metadata": {},
+   "source": [
+    "### Setup\n",
+    "Install dependencies and set API keys."
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": 1,
+   "metadata": {},
+   "outputs": [],
+   "source": [
+    "# Uncomment to install openlm and openai if you haven't already\n",
+    "\n",
+    "# !pip install openlm\n",
+    "# !pip install openai"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": 2,
+   "metadata": {},
+   "outputs": [],
+   "source": [
+    "from getpass import getpass\n",
+    "import os\n",
+    "import subprocess\n",
+    "\n",
+    "\n",
+    "# Check if OPENAI_API_KEY environment variable is set\n",
+    "if \"OPENAI_API_KEY\" not in os.environ:\n",
+    "    print(\"Enter your OpenAI API key:\")\n",
+    "    os.environ[\"OPENAI_API_KEY\"] = getpass()\n",
+    "\n",
+    "# Check if HF_API_TOKEN environment variable is set\n",
+    "if \"HF_API_TOKEN\" not in os.environ:\n",
+    "    print(\"Enter your HuggingFace Hub API key:\")\n",
+    "    os.environ[\"HF_API_TOKEN\"] = getpass()\n"
+   ]
+  },
+  {
+   "cell_type": "markdown",
+   "metadata": {},
+   "source": [
+    "### Using LangChain with OpenLM\n",
+    "\n",
+    "Here we're going to call two models in an LLMChain, `text-davinci-003` from OpenAI and `gpt2` on HuggingFace."
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": 4,
+   "metadata": {},
+   "outputs": [],
+   "source": [
+    "from langchain.llms import OpenLM\n",
+    "from langchain import PromptTemplate, LLMChain"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": 5,
+   "metadata": {},
+   "outputs": [
+    {
+     "name": "stdout",
+     "output_type": "stream",
+     "text": [
+      "Model: text-davinci-003\n",
+      "Result:  France is a country in Europe. The capital of France is Paris.\n",
+      "Model: huggingface.co/gpt2\n",
+      "Result: Question: What is the capital of France?\n",
+      "\n",
+      "Answer: Let's think step by step. I am not going to lie, this is a complicated issue, and I don't see any solutions to all this, but it is still far more\n"
+     ]
+    }
+   ],
+   "source": [
+    "question = \"What is the capital of France?\"\n",
+    "template = \"\"\"Question: {question}\n",
+    "\n",
+    "Answer: Let's think step by step.\"\"\"\n",
+    "\n",
+    "prompt = PromptTemplate(template=template, input_variables=[\"question\"])\n",
+    "\n",
+    "for model in [\"text-davinci-003\", \"huggingface.co/gpt2\"]:\n",
+    "    llm = OpenLM(model=model)\n",
+    "    llm_chain = LLMChain(prompt=prompt, llm=llm)\n",
+    "    result = llm_chain.run(question)\n",
+    "    print(\"\"\"Model: {}\n",
+    "Result: {}\"\"\".format(model, result))"
+   ]
+  }
+ ],
+ "metadata": {
+  "kernelspec": {
+   "display_name": "Python 3 (ipykernel)",
+   "language": "python",
+   "name": "python3"
+  },
+  "language_info": {
+   "codemirror_mode": {
+    "name": "ipython",
+    "version": 3
+   },
+   "file_extension": ".py",
+   "mimetype": "text/x-python",
+   "name": "python",
+   "nbconvert_exporter": "python",
+   "pygments_lexer": "ipython3",
+   "version": "3.11.3"
+  }
+ },
+ "nbformat": 4,
+ "nbformat_minor": 2
+}

langchain/llms/__init__.py

@@ -24,6 +24,7 @@
 from langchain.llms.modal import Modal
 from langchain.llms.nlpcloud import NLPCloud
 from langchain.llms.openai import AzureOpenAI, OpenAI, OpenAIChat
+from langchain.llms.openlm import OpenLM
 from langchain.llms.petals import Petals
 from langchain.llms.pipelineai import PipelineAI
 from langchain.llms.predictionguard import PredictionGuard
@@ -53,6 +54,7 @@
     "NLPCloud",
     "OpenAI",
     "OpenAIChat",
+    "OpenLM",
     "Petals",
     "PipelineAI",
     "HuggingFaceEndpoint",
@@ -96,6 +98,7 @@
     "nlpcloud": NLPCloud,
     "human-input": HumanInputLLM,
     "openai": OpenAI,
+    "openlm": OpenLM,
     "petals": Petals,
     "pipelineai": PipelineAI,
     "huggingface_pipeline": HuggingFacePipeline,

langchain/llms/openlm.py

@@ -0,0 +1,26 @@
+from typing import Any, Dict
+
+from pydantic import root_validator
+
+from langchain.llms.openai import BaseOpenAI
+
+
+class OpenLM(BaseOpenAI):
+    @property
+    def _invocation_params(self) -> Dict[str, Any]:
+        return {**{"model": self.model_name}, **super()._invocation_params}
+
+    @root_validator()
+    def validate_environment(cls, values: Dict) -> Dict:
+        try:
+            import openlm
+
+            values["client"] = openlm.Completion
+        except ImportError:
+            raise ValueError(
+                "Could not import openlm python package. "
+                "Please install it with `pip install openlm`."
+            )
+        if values["streaming"]:
+            raise ValueError("Streaming not supported with openlm")
+        return values

pyproject.toml

@@ -91,6 +91,7 @@ psychicapi = {version = "^0.2", optional = true}
 zep-python = {version="^0.25", optional=true}
 chardet = {version="^5.1.0", optional=true}
 requests-toolbelt = {version = "^1.0.0", optional = true}
+openlm = {version = "^0.0.5", optional = true}
 
 [tool.poetry.group.docs.dependencies]
 autodoc_pydantic = "^1.8.0"
@@ -175,7 +176,7 @@ playwright = "^1.28.0"
 setuptools = "^67.6.1"
 
 [tool.poetry.extras]
-llms = ["anthropic", "cohere", "openai", "nlpcloud", "huggingface_hub", "manifest-ml", "torch", "transformers"]
+llms = ["anthropic", "cohere", "openai", "openlm", "nlpcloud", "huggingface_hub", "manifest-ml", "torch", "transformers"]
 qdrant = ["qdrant-client"]
 openai = ["openai", "tiktoken"]
 text_helpers = ["chardet"]
@@ -241,6 +242,7 @@ all = [
     "lxml",
     "requests-toolbelt",
     "neo4j",
+    "openlm"
 ]
 
 # An extra used to be able to add extended testing.

tests/integration_tests/llms/test_openlm.py

@@ -0,0 +1,8 @@
+from langchain.llms.openlm import OpenLM
+
+
+def test_openlm_call() -> None:
+    """Test valid call to openlm."""
+    llm = OpenLM(model_name="dolly-v2-7b", max_tokens=10)
+    output = llm(prompt="Say foo:")
+    assert isinstance(output, str)