github/langchain
1
0
Fork 0
mirror of https://github.com/hwchase17/langchain.git synced 2025-09-28 23:07:11 +00:00
Code Issues Packages Projects Releases Wiki Activity
Files
aa9f2478035b079da8ffb7bd477e761615405236
langchain/docs/docs/how_to/chat_models_universal_init.ipynb

445 lines
15 KiB
Plaintext
Raw Blame History

{
"cells": [
{
"cell_type": "markdown",
"id": "cfdf4f09-8125-4ed1-8063-6feed57da8a3",
"metadata": {},
"source": [
"# How to init any model in one line\n",
"\n",
"Many LLM applications let end users specify what model provider and model they want the application to be powered by. This requires writing some logic to initialize different ChatModels based on some user configuration. The `init_chat_model()` helper method makes it easy to initialize a number of different model integrations without having to worry about import paths and class names.\n",
"\n",
":::tip Supported models\n",
"\n",
"See the [init_chat_model()](https://python.langchain.com/v0.2/api_reference/langchain/chat_models/langchain.chat_models.base.init_chat_model.html) API reference for a full list of supported integrations.\n",
"\n",
"Make sure you have the integration packages installed for any model providers you want to support. E.g. you should have `langchain-openai` installed to init an OpenAI model.\n",
"\n",
":::\n",
"\n",
":::info Requires ``langchain >= 0.2.8``\n",
"\n",
"This functionality was added in ``langchain-core == 0.2.8``. Please make sure your package is up to date.\n",
"\n",
":::"
]
},
{
"cell_type": "code",
"execution_count": 1,
"id": "165b0de6-9ae3-4e3d-aa98-4fc8a97c4a06",
"metadata": {
"execution": {
"iopub.execute_input": "2024-09-10T20:22:32.858670Z",
"iopub.status.busy": "2024-09-10T20:22:32.858278Z",
"iopub.status.idle": "2024-09-10T20:22:33.009452Z",
"shell.execute_reply": "2024-09-10T20:22:33.007022Z"
}
},
"outputs": [
{
"name": "stdout",
"output_type": "stream",
"text": [
"zsh:1: 0.2.8 not found\r\n"
]
},
{
"name": "stdout",
"output_type": "stream",
"text": [
"Note: you may need to restart the kernel to use updated packages.\n"
]
}
],
"source": [
"%pip install -qU langchain>=0.2.8 langchain-openai langchain-anthropic langchain-google-vertexai"
]
},
{
"cell_type": "markdown",
"id": "ea2c9f57-a796-45f8-b6f4-3efd3f361a9b",
"metadata": {},
"source": [
"## Basic usage"
]
},
{
"cell_type": "code",
"execution_count": 2,
"id": "79e14913-803c-4382-9009-5c6af3d75d35",
"metadata": {
"execution": {
"iopub.execute_input": "2024-09-10T20:22:33.015729Z",
"iopub.status.busy": "2024-09-10T20:22:33.015241Z",
"iopub.status.idle": "2024-09-10T20:22:39.391716Z",
"shell.execute_reply": "2024-09-10T20:22:39.390438Z"
}
},
"outputs": [
{
"name": "stderr",
"output_type": "stream",
"text": [
"/var/folders/4j/2rz3865x6qg07tx43146py8h0000gn/T/ipykernel_95293/571506279.py:4: LangChainBetaWarning: The function `init_chat_model` is in beta. It is actively being worked on, so the API may change.\n",
" gpt_4o = init_chat_model(\"gpt-4o\", model_provider=\"openai\", temperature=0)\n"
]
},
{
"name": "stdout",
"output_type": "stream",
"text": [
"GPT-4o: I'm an AI created by OpenAI, and I don't have a personal name. How can I assist you today?\n",
"\n"
]
},
{
"name": "stdout",
"output_type": "stream",
"text": [
"Claude Opus: My name is Claude. It's nice to meet you!\n",
"\n"
]
},
{
"name": "stdout",
"output_type": "stream",
"text": [
"Gemini 1.5: I am a large language model, trained by Google. \n",
"\n",
"I don't have a name like a person does. You can call me Bard if you like! 😊 \n",
"\n",
"\n"
]
}
],
"source": [
"from langchain.chat_models import init_chat_model\n",
"\n",
"# Returns a langchain_openai.ChatOpenAI instance.\n",
"gpt_4o = init_chat_model(\"gpt-4o\", model_provider=\"openai\", temperature=0)\n",
"# Returns a langchain_anthropic.ChatAnthropic instance.\n",
"claude_opus = init_chat_model(\n",
" \"claude-3-opus-20240229\", model_provider=\"anthropic\", temperature=0\n",
")\n",
"# Returns a langchain_google_vertexai.ChatVertexAI instance.\n",
"gemini_15 = init_chat_model(\n",
" \"gemini-1.5-pro\", model_provider=\"google_vertexai\", temperature=0\n",
")\n",
"\n",
"# Since all model integrations implement the ChatModel interface, you can use them in the same way.\n",
"print(\"GPT-4o: \" + gpt_4o.invoke(\"what's your name\").content + \"\\n\")\n",
"print(\"Claude Opus: \" + claude_opus.invoke(\"what's your name\").content + \"\\n\")\n",
"print(\"Gemini 1.5: \" + gemini_15.invoke(\"what's your name\").content + \"\\n\")"
]
},
{
"cell_type": "markdown",
"id": "f811f219-5e78-4b62-b495-915d52a22532",
"metadata": {},
"source": [
"## Inferring model provider\n",
"\n",
"For common and distinct model names `init_chat_model()` will attempt to infer the model provider. See the [API reference](https://python.langchain.com/v0.2/api_reference/langchain/chat_models/langchain.chat_models.base.init_chat_model.html) for a full list of inference behavior. E.g. any model that starts with `gpt-3...` or `gpt-4...` will be inferred as using model provider `openai`."
]
},
{
"cell_type": "code",
"execution_count": 3,
"id": "0378ccc6-95bc-4d50-be50-fccc193f0a71",
"metadata": {
"execution": {
"iopub.execute_input": "2024-09-10T20:22:39.396908Z",
"iopub.status.busy": "2024-09-10T20:22:39.396563Z",
"iopub.status.idle": "2024-09-10T20:22:39.444959Z",
"shell.execute_reply": "2024-09-10T20:22:39.444646Z"
}
},
"outputs": [],
"source": [
"gpt_4o = init_chat_model(\"gpt-4o\", temperature=0)\n",
"claude_opus = init_chat_model(\"claude-3-opus-20240229\", temperature=0)\n",
"gemini_15 = init_chat_model(\"gemini-1.5-pro\", temperature=0)"
]
},
{
"cell_type": "markdown",
"id": "476a44db-c50d-4846-951d-0f1c9ba8bbaa",
"metadata": {},
"source": [
"## Creating a configurable model\n",
"\n",
"You can also create a runtime-configurable model by specifying `configurable_fields`. If you don't specify a `model` value, then \"model\" and \"model_provider\" be configurable by default."
]
},
{
"cell_type": "code",
"execution_count": 4,
"id": "6c037f27-12d7-4e83-811e-4245c0e3ba58",
"metadata": {
"execution": {
"iopub.execute_input": "2024-09-10T20:22:39.446901Z",
"iopub.status.busy": "2024-09-10T20:22:39.446773Z",
"iopub.status.idle": "2024-09-10T20:22:40.301906Z",
"shell.execute_reply": "2024-09-10T20:22:40.300918Z"
}
},
"outputs": [
{
"data": {
"text/plain": [
"AIMessage(content=\"I'm an AI created by OpenAI, and I don't have a personal name. How can I assist you today?\", additional_kwargs={'refusal': None}, response_metadata={'token_usage': {'completion_tokens': 23, 'prompt_tokens': 11, 'total_tokens': 34}, 'model_name': 'gpt-4o-2024-05-13', 'system_fingerprint': 'fp_25624ae3a5', 'finish_reason': 'stop', 'logprobs': None}, id='run-b41df187-4627-490d-af3c-1c96282d3eb0-0', usage_metadata={'input_tokens': 11, 'output_tokens': 23, 'total_tokens': 34})"
]
},
"execution_count": 4,
"metadata": {},
"output_type": "execute_result"
}
],
"source": [
"configurable_model = init_chat_model(temperature=0)\n",
"\n",
"configurable_model.invoke(\n",
" \"what's your name\", config={\"configurable\": {\"model\": \"gpt-4o\"}}\n",
")"
]
},
{
"cell_type": "code",
"execution_count": 5,
"id": "321e3036-abd2-4e1f-bcc6-606efd036954",
"metadata": {
"execution": {
"iopub.execute_input": "2024-09-10T20:22:40.316030Z",
"iopub.status.busy": "2024-09-10T20:22:40.315628Z",
"iopub.status.idle": "2024-09-10T20:22:41.199134Z",
"shell.execute_reply": "2024-09-10T20:22:41.198173Z"
}
},
"outputs": [
{
"data": {
"text/plain": [
"AIMessage(content=\"My name is Claude. It's nice to meet you!\", additional_kwargs={}, response_metadata={'id': 'msg_01Fx9P74A7syoFkwE73CdMMY', 'model': 'claude-3-5-sonnet-20240620', 'stop_reason': 'end_turn', 'stop_sequence': None, 'usage': {'input_tokens': 11, 'output_tokens': 15}}, id='run-a0fd2bbd-3b7e-46bf-8d69-a48c7e60b03c-0', usage_metadata={'input_tokens': 11, 'output_tokens': 15, 'total_tokens': 26})"
]
},
"execution_count": 5,
"metadata": {},
"output_type": "execute_result"
}
],
"source": [
"configurable_model.invoke(\n",
" \"what's your name\", config={\"configurable\": {\"model\": \"claude-3-5-sonnet-20240620\"}}\n",
")"
]
},
{
"cell_type": "markdown",
"id": "7f3b3d4a-4066-45e4-8297-ea81ac8e70b7",
"metadata": {},
"source": [
"### Configurable model with default values\n",
"\n",
"We can create a configurable model with default model values, specify which parameters are configurable, and add prefixes to configurable params:"
]
},
{
"cell_type": "code",
"execution_count": 6,
"id": "814a2289-d0db-401e-b555-d5116112b413",
"metadata": {
"execution": {
"iopub.execute_input": "2024-09-10T20:22:41.203346Z",
"iopub.status.busy": "2024-09-10T20:22:41.203004Z",
"iopub.status.idle": "2024-09-10T20:22:41.891450Z",
"shell.execute_reply": "2024-09-10T20:22:41.890539Z"
}
},
"outputs": [
{
"data": {
"text/plain": [
"AIMessage(content=\"I'm an AI created by OpenAI, and I don't have a personal name. How can I assist you today?\", additional_kwargs={'refusal': None}, response_metadata={'token_usage': {'completion_tokens': 23, 'prompt_tokens': 11, 'total_tokens': 34}, 'model_name': 'gpt-4o-2024-05-13', 'system_fingerprint': 'fp_25624ae3a5', 'finish_reason': 'stop', 'logprobs': None}, id='run-3380f977-4b89-4f44-bc02-b64043b3166f-0', usage_metadata={'input_tokens': 11, 'output_tokens': 23, 'total_tokens': 34})"
]
},
"execution_count": 6,
"metadata": {},
"output_type": "execute_result"
}
],
"source": [
"first_llm = init_chat_model(\n",
" model=\"gpt-4o\",\n",
" temperature=0,\n",
" configurable_fields=(\"model\", \"model_provider\", \"temperature\", \"max_tokens\"),\n",
" config_prefix=\"first\", # useful when you have a chain with multiple models\n",
")\n",
"\n",
"first_llm.invoke(\"what's your name\")"
]
},
{
"cell_type": "code",
"execution_count": 7,
"id": "6c8755ba-c001-4f5a-a497-be3f1db83244",
"metadata": {
"execution": {
"iopub.execute_input": "2024-09-10T20:22:41.896413Z",
"iopub.status.busy": "2024-09-10T20:22:41.895967Z",
"iopub.status.idle": "2024-09-10T20:22:42.767565Z",
"shell.execute_reply": "2024-09-10T20:22:42.766619Z"
}
},
"outputs": [
{
"data": {
"text/plain": [
"AIMessage(content=\"My name is Claude. It's nice to meet you!\", additional_kwargs={}, response_metadata={'id': 'msg_01EFKSWpmsn2PSYPQa4cNHWb', 'model': 'claude-3-5-sonnet-20240620', 'stop_reason': 'end_turn', 'stop_sequence': None, 'usage': {'input_tokens': 11, 'output_tokens': 15}}, id='run-3c58f47c-41b9-4e56-92e7-fb9602e3787c-0', usage_metadata={'input_tokens': 11, 'output_tokens': 15, 'total_tokens': 26})"
]
},
"execution_count": 7,
"metadata": {},
"output_type": "execute_result"
}
],
"source": [
"first_llm.invoke(\n",
" \"what's your name\",\n",
" config={\n",
" \"configurable\": {\n",
" \"first_model\": \"claude-3-5-sonnet-20240620\",\n",
" \"first_temperature\": 0.5,\n",
" \"first_max_tokens\": 100,\n",
" }\n",
" },\n",
")"
]
},
{
"cell_type": "markdown",
"id": "0072b1a3-7e44-4b4e-8b07-efe1ba91a689",
"metadata": {},
"source": [
"### Using a configurable model declaratively\n",
"\n",
"We can call declarative operations like `bind_tools`, `with_structured_output`, `with_configurable`, etc. on a configurable model and chain a configurable model in the same way that we would a regularly instantiated chat model object."
]
},
{
"cell_type": "code",
"execution_count": 8,
"id": "067dabee-1050-4110-ae24-c48eba01e13b",
"metadata": {
"execution": {
"iopub.execute_input": "2024-09-10T20:22:42.771941Z",
"iopub.status.busy": "2024-09-10T20:22:42.771606Z",
"iopub.status.idle": "2024-09-10T20:22:43.909206Z",
"shell.execute_reply": "2024-09-10T20:22:43.908496Z"
}
},
"outputs": [
{
"data": {
"text/plain": [
"[{'name': 'GetPopulation',\n",
" 'args': {'location': 'Los Angeles, CA'},\n",
" 'id': 'call_Ga9m8FAArIyEjItHmztPYA22',\n",
" 'type': 'tool_call'},\n",
" {'name': 'GetPopulation',\n",
" 'args': {'location': 'New York, NY'},\n",
" 'id': 'call_jh2dEvBaAHRaw5JUDthOs7rt',\n",
" 'type': 'tool_call'}]"
]
},
"execution_count": 8,
"metadata": {},
"output_type": "execute_result"
}
],
"source": [
"from pydantic import BaseModel, Field\n",
"\n",
"\n",
"class GetWeather(BaseModel):\n",
" \"\"\"Get the current weather in a given location\"\"\"\n",
"\n",
" location: str = Field(..., description=\"The city and state, e.g. San Francisco, CA\")\n",
"\n",
"\n",
"class GetPopulation(BaseModel):\n",
" \"\"\"Get the current population in a given location\"\"\"\n",
"\n",
" location: str = Field(..., description=\"The city and state, e.g. San Francisco, CA\")\n",
"\n",
"\n",
"llm = init_chat_model(temperature=0)\n",
"llm_with_tools = llm.bind_tools([GetWeather, GetPopulation])\n",
"\n",
"llm_with_tools.invoke(\n",
" \"what's bigger in 2024 LA or NYC\", config={\"configurable\": {\"model\": \"gpt-4o\"}}\n",
").tool_calls"
]
},
{
"cell_type": "code",
"execution_count": 9,
"id": "e57dfe9f-cd24-4e37-9ce9-ccf8daf78f89",
"metadata": {
"execution": {
"iopub.execute_input": "2024-09-10T20:22:43.912746Z",
"iopub.status.busy": "2024-09-10T20:22:43.912447Z",
"iopub.status.idle": "2024-09-10T20:22:46.437049Z",
"shell.execute_reply": "2024-09-10T20:22:46.436093Z"
}
},
"outputs": [
{
"data": {
"text/plain": [
"[{'name': 'GetPopulation',\n",
" 'args': {'location': 'Los Angeles, CA'},\n",
" 'id': 'toolu_01JMufPf4F4t2zLj7miFeqXp',\n",
" 'type': 'tool_call'},\n",
" {'name': 'GetPopulation',\n",
" 'args': {'location': 'New York City, NY'},\n",
" 'id': 'toolu_01RQBHcE8kEEbYTuuS8WqY1u',\n",
" 'type': 'tool_call'}]"
]
},
"execution_count": 9,
"metadata": {},
"output_type": "execute_result"
}
],
"source": [
"llm_with_tools.invoke(\n",
" \"what's bigger in 2024 LA or NYC\",\n",
" config={\"configurable\": {\"model\": \"claude-3-5-sonnet-20240620\"}},\n",
").tool_calls"
]
}
],
"metadata": {
"kernelspec": {
"display_name": "poetry-venv-2",
"language": "python",
"name": "poetry-venv-2"
},
"language_info": {
"codemirror_mode": {
"name": "ipython",
"version": 3
},
"file_extension": ".py",
"mimetype": "text/x-python",
"name": "python",
"nbconvert_exporter": "python",
"pygments_lexer": "ipython3",
"version": "3.11.9"
}
},
"nbformat": 4,
"nbformat_minor": 5
}
Reference in New Issue View Git Blame Copy Permalink