mirror of
https://github.com/hwchase17/langchain.git
synced 2025-08-12 14:23:58 +00:00
update chat models
This commit is contained in:
philschmid
parent
7ecdac5240
commit
76ae2b203a
@ -1,35 +1,16 @@
|
||||
{
|
||||
"cells": [
|
||||
{
|
||||
"cell_type": "raw",
|
||||
"id": "afaf8039",
|
||||
"metadata": {},
|
||||
"source": [
|
||||
"---\n",
|
||||
"sidebar_label: Google AI\n",
|
||||
"---"
|
||||
]
|
||||
},
|
||||
{
|
||||
"cell_type": "markdown",
|
||||
"id": "e49f1e0d",
|
||||
"id": "56a6d990",
|
||||
"metadata": {},
|
||||
"source": [
|
||||
"# ChatGoogleGenerativeAI\n",
|
||||
"## ChatGoogleGenerativeAI (Gemini API & AI Studio)\n",
|
||||
"\n",
|
||||
"This docs will help you get started with Google AI [chat models](/docs/concepts/chat_models). For detailed documentation of all ChatGoogleGenerativeAI features and configurations head to the [API reference](https://python.langchain.com/api_reference/google_genai/chat_models/langchain_google_genai.chat_models.ChatGoogleGenerativeAI.html).\n",
|
||||
"Access Google's Generative AI models, including the Gemini family, directly via the Gemini API or experiment rapidly using Google AI Studio. The `langchain-google-genai` package provides the LangChain integration for these models. This is often the best starting point for individual developers.\n",
|
||||
"\n",
|
||||
"Google AI offers a number of different chat models. For information on the latest models, their features, context windows, etc. head to the [Google AI docs](https://ai.google.dev/gemini-api/docs/models/gemini).\n",
|
||||
"For information on the latest models, their features, context windows, etc. head to the [Google AI docs](https://ai.google.dev/gemini-api/docs/models/gemini).\n",
|
||||
"\n",
|
||||
":::info Google AI vs Google Cloud Vertex AI\n",
|
||||
"\n",
|
||||
"Google's Gemini models are accessible through Google AI and through Google Cloud Vertex AI. Using Google AI just requires a Google account and an API key. Using Google Cloud Vertex AI requires a Google Cloud account (with term agreements and billing) but offers enterprise features like customer encryption key, virtual private cloud, and more.\n",
|
||||
"\n",
|
||||
"To learn more about the key features of the two APIs see the [Google docs](https://cloud.google.com/vertex-ai/generative-ai/docs/migrate/migrate-google-ai#google-ai).\n",
|
||||
"\n",
|
||||
":::\n",
|
||||
"\n",
|
||||
"## Overview\n",
|
||||
"### Integration details\n",
|
||||
"\n",
|
||||
"| Class | Package | Local | Serializable | [JS support](https://js.langchain.com/docs/integrations/chat/google_generativeai) | Package downloads | Package latest |\n",
|
||||
@ -37,23 +18,46 @@
|
||||
"| [ChatGoogleGenerativeAI](https://python.langchain.com/api_reference/google_genai/chat_models/langchain_google_genai.chat_models.ChatGoogleGenerativeAI.html) | [langchain-google-genai](https://python.langchain.com/api_reference/google_genai/index.html) | ❌ | beta | ✅ |  |  |\n",
|
||||
"\n",
|
||||
"### Model features\n",
|
||||
"\n",
|
||||
"| [Tool calling](/docs/how_to/tool_calling) | [Structured output](/docs/how_to/structured_output/) | JSON mode | [Image input](/docs/how_to/multimodal_inputs/) | Audio input | Video input | [Token-level streaming](/docs/how_to/chat_streaming/) | Native async | [Token usage](/docs/how_to/chat_token_usage_tracking/) | [Logprobs](/docs/how_to/logprobs/) |\n",
|
||||
"| :---: | :---: | :---: | :---: | :---: | :---: | :---: | :---: | :---: | :---: |\n",
|
||||
"| ✅ | ✅ | ❌ | ✅ | ✅ | ✅ | ✅ | ✅ | ✅ | ❌ |\n",
|
||||
"\n",
|
||||
"## Setup\n",
|
||||
"### Setup\n",
|
||||
"\n",
|
||||
"To access Google AI models you'll need to create a Google Acount account, get a Google AI API key, and install the `langchain-google-genai` integration package.\n",
|
||||
"To access Google AI models you'll need to create a Google Account, get a Google AI API key, and install the `langchain-google-genai` integration package.\n",
|
||||
"\n",
|
||||
"### Credentials\n",
|
||||
"\n",
|
||||
"Head to https://ai.google.dev/gemini-api/docs/api-key to generate a Google AI API key. Once you've done this set the GOOGLE_API_KEY environment variable:"
|
||||
"**1. Installation:**"
|
||||
]
|
||||
},
|
||||
{
|
||||
"cell_type": "code",
|
||||
"execution_count": null,
|
||||
"id": "433e8d2b-9519-4b49-b2c4-7ab65b046c94",
|
||||
"id": "8d12ce35",
|
||||
"metadata": {},
|
||||
"outputs": [],
|
||||
"source": [
|
||||
"%pip install -U langchain-google-genai"
|
||||
]
|
||||
},
|
||||
{
|
||||
"cell_type": "markdown",
|
||||
"id": "60be0b38",
|
||||
"metadata": {},
|
||||
"source": [
|
||||
"**2. Credentials:**\n",
|
||||
"\n",
|
||||
"Head to [https://ai.google.dev/gemini-api/docs/api-key](https://ai.google.dev/gemini-api/docs/api-key) (or via Google AI Studio) to generate a Google AI API key.\n",
|
||||
"\n",
|
||||
"### Chat Models\n",
|
||||
"\n",
|
||||
"Use the `ChatGoogleGenerativeAI` class to interact with Google's chat models. See the [API reference](https://python.langchain.com/api_reference/google_genai/chat_models/langchain_google_genai.chat_models.ChatGoogleGenerativeAI.html) for full details.\n"
|
||||
]
|
||||
},
|
||||
{
|
||||
"cell_type": "code",
|
||||
"execution_count": 3,
|
||||
"id": "fb18c875",
|
||||
"metadata": {},
|
||||
"outputs": [],
|
||||
"source": [
|
||||
@ -61,12 +65,14 @@
|
||||
"import os\n",
|
||||
"\n",
|
||||
"if \"GOOGLE_API_KEY\" not in os.environ:\n",
|
||||
" os.environ[\"GOOGLE_API_KEY\"] = getpass.getpass(\"Enter your Google AI API key: \")"
|
||||
" os.environ[\"GOOGLE_API_KEY\"] = getpass.getpass(\"Enter your Google AI API key: \")\n",
|
||||
" \n",
|
||||
"os.environ[\"GOOGLE_API_KEY\"] = \"AIzaSyBJ51sF-xa4J5dGTNGw0TWjMKuiB5-ljuc\""
|
||||
]
|
||||
},
|
||||
{
|
||||
"cell_type": "markdown",
|
||||
"id": "72ee0c4b-9764-423a-9dbf-95129e185210",
|
||||
"id": "f050e8db",
|
||||
"metadata": {},
|
||||
"source": [
|
||||
"To enable automated tracing of your model calls, set your [LangSmith](https://docs.smith.langchain.com/) API key:"
|
||||
@ -75,7 +81,7 @@
|
||||
{
|
||||
"cell_type": "code",
|
||||
"execution_count": null,
|
||||
"id": "a15d341e-3e26-4ca3-830b-5aab30ed66de",
|
||||
"id": "82cb346f",
|
||||
"metadata": {},
|
||||
"outputs": [],
|
||||
"source": [
|
||||
@ -85,27 +91,7 @@
|
||||
},
|
||||
{
|
||||
"cell_type": "markdown",
|
||||
"id": "0730d6a1-c893-4840-9817-5e5251676d5d",
|
||||
"metadata": {},
|
||||
"source": [
|
||||
"### Installation\n",
|
||||
"\n",
|
||||
"The LangChain Google AI integration lives in the `langchain-google-genai` package:"
|
||||
]
|
||||
},
|
||||
{
|
||||
"cell_type": "code",
|
||||
"execution_count": null,
|
||||
"id": "652d6238-1f87-422a-b135-f5abbb8652fc",
|
||||
"metadata": {},
|
||||
"outputs": [],
|
||||
"source": [
|
||||
"%pip install -qU langchain-google-genai"
|
||||
]
|
||||
},
|
||||
{
|
||||
"cell_type": "markdown",
|
||||
"id": "a38cde65-254d-4219-a441-068766c0d4b5",
|
||||
"id": "273cefa0",
|
||||
"metadata": {},
|
||||
"source": [
|
||||
"## Instantiation\n",
|
||||
@ -115,15 +101,15 @@
|
||||
},
|
||||
{
|
||||
"cell_type": "code",
|
||||
"execution_count": 2,
|
||||
"id": "cb09c344-1836-4e0c-acf8-11d13ac1dbae",
|
||||
"execution_count": 4,
|
||||
"id": "7d3dc0b3",
|
||||
"metadata": {},
|
||||
"outputs": [],
|
||||
"source": [
|
||||
"from langchain_google_genai import ChatGoogleGenerativeAI\n",
|
||||
"\n",
|
||||
"llm = ChatGoogleGenerativeAI(\n",
|
||||
" model=\"gemini-2.0-flash-001\",\n",
|
||||
" model=\"gemini-2.0-flash\",\n",
|
||||
" temperature=0,\n",
|
||||
" max_tokens=None,\n",
|
||||
" timeout=None,\n",
|
||||
@ -134,7 +120,7 @@
|
||||
},
|
||||
{
|
||||
"cell_type": "markdown",
|
||||
"id": "2b4f3e15",
|
||||
"id": "343a8c13",
|
||||
"metadata": {},
|
||||
"source": [
|
||||
"## Invocation"
|
||||
@ -142,19 +128,17 @@
|
||||
},
|
||||
{
|
||||
"cell_type": "code",
|
||||
"execution_count": 3,
|
||||
"id": "62e0dbc3",
|
||||
"metadata": {
|
||||
"tags": []
|
||||
},
|
||||
"execution_count": 5,
|
||||
"id": "82c5708c",
|
||||
"metadata": {},
|
||||
"outputs": [
|
||||
{
|
||||
"data": {
|
||||
"text/plain": [
|
||||
"AIMessage(content=\"J'adore la programmation.\", additional_kwargs={}, response_metadata={'prompt_feedback': {'block_reason': 0, 'safety_ratings': []}, 'finish_reason': 'STOP', 'model_name': 'gemini-2.0-flash-001', 'safety_ratings': []}, id='run-61cff164-40be-4f88-a2df-cca58297502f-0', usage_metadata={'input_tokens': 20, 'output_tokens': 7, 'total_tokens': 27, 'input_token_details': {'cache_read': 0}})"
|
||||
"AIMessage(content=\"J'adore la programmation.\", additional_kwargs={}, response_metadata={'prompt_feedback': {'block_reason': 0, 'safety_ratings': []}, 'finish_reason': 'STOP', 'model_name': 'gemini-2.0-flash', 'safety_ratings': []}, id='run-3b28d4b8-8a62-4e6c-ad4e-b53e6e825749-0', usage_metadata={'input_tokens': 20, 'output_tokens': 7, 'total_tokens': 27, 'input_token_details': {'cache_read': 0}})"
|
||||
]
|
||||
},
|
||||
"execution_count": 3,
|
||||
"execution_count": 5,
|
||||
"metadata": {},
|
||||
"output_type": "execute_result"
|
||||
}
|
||||
@ -173,8 +157,8 @@
|
||||
},
|
||||
{
|
||||
"cell_type": "code",
|
||||
"execution_count": 4,
|
||||
"id": "d86145b3-bfef-46e8-b227-4dda5c9c2705",
|
||||
"execution_count": 6,
|
||||
"id": "49d2d0c2",
|
||||
"metadata": {},
|
||||
"outputs": [
|
||||
{
|
||||
@ -191,7 +175,7 @@
|
||||
},
|
||||
{
|
||||
"cell_type": "markdown",
|
||||
"id": "18e2bfc0-7e78-4528-a73f-499ac150dca8",
|
||||
"id": "ee3f6e1d",
|
||||
"metadata": {},
|
||||
"source": [
|
||||
"## Chaining\n",
|
||||
@ -201,17 +185,17 @@
|
||||
},
|
||||
{
|
||||
"cell_type": "code",
|
||||
"execution_count": 5,
|
||||
"id": "e197d1d7-a070-4c96-9f8a-a0e86d046e0b",
|
||||
"execution_count": 7,
|
||||
"id": "3c8407ee",
|
||||
"metadata": {},
|
||||
"outputs": [
|
||||
{
|
||||
"data": {
|
||||
"text/plain": [
|
||||
"AIMessage(content='Ich liebe Programmieren.', additional_kwargs={}, response_metadata={'prompt_feedback': {'block_reason': 0, 'safety_ratings': []}, 'finish_reason': 'STOP', 'model_name': 'gemini-2.0-flash-001', 'safety_ratings': []}, id='run-dd2f8fb9-62d9-4b84-9c97-ed9c34cda313-0', usage_metadata={'input_tokens': 15, 'output_tokens': 7, 'total_tokens': 22, 'input_token_details': {'cache_read': 0}})"
|
||||
"AIMessage(content='Ich liebe Programmieren.', additional_kwargs={}, response_metadata={'prompt_feedback': {'block_reason': 0, 'safety_ratings': []}, 'finish_reason': 'STOP', 'model_name': 'gemini-2.0-flash', 'safety_ratings': []}, id='run-e5561c6b-2beb-4411-9210-4796b576a7cd-0', usage_metadata={'input_tokens': 15, 'output_tokens': 7, 'total_tokens': 22, 'input_token_details': {'cache_read': 0}})"
|
||||
]
|
||||
},
|
||||
"execution_count": 5,
|
||||
"execution_count": 7,
|
||||
"metadata": {},
|
||||
"output_type": "execute_result"
|
||||
}
|
||||
@ -241,22 +225,174 @@
|
||||
},
|
||||
{
|
||||
"cell_type": "markdown",
|
||||
"id": "41c2ff10-a3ba-4f40-b3aa-7a395854849e",
|
||||
"id": "bdae9742",
|
||||
"metadata": {},
|
||||
"source": [
|
||||
"## Image generation\n",
|
||||
"## Multimodal Usage\n",
|
||||
"\n",
|
||||
"Some Gemini models (specifically `gemini-2.0-flash-exp`) support image generation capabilities.\n",
|
||||
"Gemini models can accept multimodal inputs (text, images, audio, video) and, for some models, generate multimodal outputs.\n",
|
||||
"\n",
|
||||
"### Text to image\n",
|
||||
"### Image Input\n",
|
||||
"\n",
|
||||
"See a simple usage example below:"
|
||||
"Provide image inputs along with text using a `HumanMessage` with a list content format. The `gemini-2.0-flash` model can handle images."
|
||||
]
|
||||
},
|
||||
{
|
||||
"cell_type": "code",
|
||||
"execution_count": 2,
|
||||
"id": "7589e14d-8d1b-4c82-965f-5558d80cb677",
|
||||
"execution_count": null,
|
||||
"id": "6833fe5d",
|
||||
"metadata": {},
|
||||
"outputs": [],
|
||||
"source": [
|
||||
"from langchain_core.messages import HumanMessage\n",
|
||||
"from langchain_google_genai import ChatGoogleGenerativeAI\n",
|
||||
"import base64\n",
|
||||
"\n",
|
||||
"# Example using a public URL (remains the same)\n",
|
||||
"message_url = HumanMessage(\n",
|
||||
" content=[\n",
|
||||
" {\n",
|
||||
" \"type\": \"text\",\n",
|
||||
" \"text\": \"Describe the image at the URL.\",\n",
|
||||
" },\n",
|
||||
" {\"type\": \"image_url\", \"image_url\": \"https://picsum.photos/seed/picsum/200/300\"},\n",
|
||||
" ]\n",
|
||||
")\n",
|
||||
"result_url = llm.invoke([message_url])\n",
|
||||
"print(f\"Response for URL image: {result_url.content}\")\n",
|
||||
"\n",
|
||||
"# Example using a local image file encoded in base64\n",
|
||||
"image_file_path = \"/Users/philschmid/projects/google-gemini/langchain/docs/static/img/agents_vs_chains.png\"\n",
|
||||
"\n",
|
||||
"with open(image_file_path, \"rb\") as image_file:\n",
|
||||
" encoded_image = base64.b64encode(image_file.read()).decode('utf-8')\n",
|
||||
"\n",
|
||||
"message_local = HumanMessage(\n",
|
||||
" content=[\n",
|
||||
" {\n",
|
||||
" \"type\": \"text\",\n",
|
||||
" \"text\": \"Describe the local image.\"\n",
|
||||
" },\n",
|
||||
" {\n",
|
||||
" \"type\": \"image_url\",\n",
|
||||
" \"image_url\": f\"data:image/png;base64,{encoded_image}\"\n",
|
||||
" }\n",
|
||||
" ]\n",
|
||||
")\n",
|
||||
"result_local = llm.invoke([message_local])\n",
|
||||
"print(f\"Response for local image: {result_local.content}\")"
|
||||
]
|
||||
},
|
||||
{
|
||||
"cell_type": "markdown",
|
||||
"id": "1b422382",
|
||||
"metadata": {},
|
||||
"source": [
|
||||
"Other supported `image_url` formats:\n",
|
||||
"- A Google Cloud Storage URI (`gs://...`). Ensure the service account has access.\n",
|
||||
"- A PIL Image object (the library handles encoding).\n",
|
||||
"\n",
|
||||
"### Audio Input\n",
|
||||
"\n",
|
||||
"Provide audio file inputs along with text. Use a model like `gemini-2.0-flash`."
|
||||
]
|
||||
},
|
||||
{
|
||||
"cell_type": "code",
|
||||
"execution_count": null,
|
||||
"id": "a3461836",
|
||||
"metadata": {},
|
||||
"outputs": [],
|
||||
"source": [
|
||||
"from langchain_core.messages import HumanMessage\n",
|
||||
"import base64\n",
|
||||
"\n",
|
||||
"# Ensure you have an audio file named 'example_audio.mp3' or provide the correct path.\n",
|
||||
"audio_file_path = \"example_audio.mp3\"\n",
|
||||
"audio_mime_type=\"audio/mpeg\"\n",
|
||||
"\n",
|
||||
"\n",
|
||||
"with open(audio_file_path, \"rb\") as audio_file:\n",
|
||||
" encoded_audio = base64.b64encode(audio_file.read()).decode('utf-8')\n",
|
||||
"\n",
|
||||
"message = HumanMessage(\n",
|
||||
" content=[\n",
|
||||
" {\n",
|
||||
" \"type\": \"text\",\n",
|
||||
" \"text\": \"Transcribe the audio.\"\n",
|
||||
" },\n",
|
||||
" {\n",
|
||||
" \"type\": \"media\",\n",
|
||||
" \"data\": encoded_audio, # Use base64 string directly\n",
|
||||
" \"mime_type\": audio_mime_type,\n",
|
||||
" }\n",
|
||||
" ]\n",
|
||||
")\n",
|
||||
"response = llm.invoke([message]) # Uncomment to run\n",
|
||||
"print(f\"Response for audio: {response.content}\")\n"
|
||||
]
|
||||
},
|
||||
{
|
||||
"cell_type": "markdown",
|
||||
"id": "0d898e27",
|
||||
"metadata": {},
|
||||
"source": [
|
||||
"### Video Input\n",
|
||||
"\n",
|
||||
"Provide video file inputs along with text. Use a model like `gemini-2.0-flash`."
|
||||
]
|
||||
},
|
||||
{
|
||||
"cell_type": "code",
|
||||
"execution_count": null,
|
||||
"id": "3046e74b",
|
||||
"metadata": {},
|
||||
"outputs": [],
|
||||
"source": [
|
||||
"from langchain_core.messages import HumanMessage\n",
|
||||
"from langchain_google_genai import ChatGoogleGenerativeAI\n",
|
||||
"import base64\n",
|
||||
"\n",
|
||||
"# Ensure you have a video file named 'example_video.mp4' or provide the correct path.\n",
|
||||
"video_file_path = \"example_video.mp4\"\n",
|
||||
"video_mime_type=\"video/mp4\"\n",
|
||||
"\n",
|
||||
"\n",
|
||||
"with open(video_file_path, \"rb\") as video_file:\n",
|
||||
" encoded_video = base64.b64encode(video_file.read()).decode('utf-8')\n",
|
||||
"\n",
|
||||
"message = HumanMessage(\n",
|
||||
" content=[\n",
|
||||
" {\n",
|
||||
" \"type\": \"text\",\n",
|
||||
" \"text\": \"Describe the first few frames of the video.\"\n",
|
||||
" },\n",
|
||||
" {\n",
|
||||
" \"type\": \"media\",\n",
|
||||
" \"data\": encoded_video, # Use base64 string directly\n",
|
||||
" \"mime_type\": video_mime_type,\n",
|
||||
" }\n",
|
||||
" ]\n",
|
||||
")\n",
|
||||
"response = llm.invoke([message]) # Uncomment to run\n",
|
||||
"print(f\"Response for video: {response.content}\")\n",
|
||||
" "
|
||||
]
|
||||
},
|
||||
{
|
||||
"cell_type": "markdown",
|
||||
"id": "2df11d89",
|
||||
"metadata": {},
|
||||
"source": [
|
||||
"### Image Generation (Multimodal Output)\n",
|
||||
"\n",
|
||||
"The `gemini-2.0-flash` model can generate text and images inline (image generation is experimental). You need to specify the desired `response_modalities`."
|
||||
]
|
||||
},
|
||||
{
|
||||
"cell_type": "code",
|
||||
"execution_count": null,
|
||||
"id": "c0b7180f",
|
||||
"metadata": {},
|
||||
"outputs": [
|
||||
{
|
||||
@ -266,17 +402,12 @@
|
||||
"<IPython.core.display.Image object>"
|
||||
]
|
||||
},
|
||||
"metadata": {
|
||||
"image/png": {
|
||||
"width": 300
|
||||
}
|
||||
},
|
||||
"metadata": {},
|
||||
"output_type": "display_data"
|
||||
}
|
||||
],
|
||||
"source": [
|
||||
"import base64\n",
|
||||
"from io import BytesIO\n",
|
||||
"\n",
|
||||
"from IPython.display import Image, display\n",
|
||||
"from langchain_google_genai import ChatGoogleGenerativeAI\n",
|
||||
@ -301,7 +432,7 @@
|
||||
},
|
||||
{
|
||||
"cell_type": "markdown",
|
||||
"id": "b14c0d87-cf7e-4d88-bda1-2ab40ec0350a",
|
||||
"id": "14bf00f1",
|
||||
"metadata": {},
|
||||
"source": [
|
||||
"### Image and text to image\n",
|
||||
@ -311,8 +442,8 @@
|
||||
},
|
||||
{
|
||||
"cell_type": "code",
|
||||
"execution_count": 3,
|
||||
"id": "0f4ed7a5-980c-4b54-b743-0b988909744c",
|
||||
"execution_count": null,
|
||||
"id": "d65e195c",
|
||||
"metadata": {},
|
||||
"outputs": [
|
||||
{
|
||||
@ -322,11 +453,7 @@
|
||||
"<IPython.core.display.Image object>"
|
||||
]
|
||||
},
|
||||
"metadata": {
|
||||
"image/png": {
|
||||
"width": 300
|
||||
}
|
||||
},
|
||||
"metadata": {},
|
||||
"output_type": "display_data"
|
||||
}
|
||||
],
|
||||
@ -349,7 +476,7 @@
|
||||
},
|
||||
{
|
||||
"cell_type": "markdown",
|
||||
"id": "a62669d8-becd-495f-8f4a-82d7c5d87969",
|
||||
"id": "43b54d3f",
|
||||
"metadata": {},
|
||||
"source": [
|
||||
"You can also represent an input image and query in a single message by encoding the base64 data in the [data URI scheme](https://en.wikipedia.org/wiki/Data_URI_scheme):"
|
||||
@ -357,8 +484,8 @@
|
||||
},
|
||||
{
|
||||
"cell_type": "code",
|
||||
"execution_count": 9,
|
||||
"id": "6241da43-e210-43bc-89af-b3c480ea06e9",
|
||||
"execution_count": null,
|
||||
"id": "0dfc7e1e",
|
||||
"metadata": {},
|
||||
"outputs": [
|
||||
{
|
||||
@ -368,11 +495,7 @@
|
||||
"<IPython.core.display.Image object>"
|
||||
]
|
||||
},
|
||||
"metadata": {
|
||||
"image/png": {
|
||||
"width": 300
|
||||
}
|
||||
},
|
||||
"metadata": {},
|
||||
"output_type": "display_data"
|
||||
}
|
||||
],
|
||||
@ -403,7 +526,7 @@
|
||||
},
|
||||
{
|
||||
"cell_type": "markdown",
|
||||
"id": "cfe228d3-6773-4283-9788-87bdf6912b1c",
|
||||
"id": "789818d7",
|
||||
"metadata": {},
|
||||
"source": [
|
||||
"You can also use LangGraph to manage the conversation history for you as in [this tutorial](/docs/tutorials/chatbot/)."
|
||||
@ -411,7 +534,298 @@
|
||||
},
|
||||
{
|
||||
"cell_type": "markdown",
|
||||
"id": "d1ee55bc-ffc8-4cfa-801c-993953a08cfd",
|
||||
"id": "b037e2dc",
|
||||
"metadata": {},
|
||||
"source": [
|
||||
"## Tool Calling\n",
|
||||
"\n",
|
||||
"You can equip the model with tools to call."
|
||||
]
|
||||
},
|
||||
{
|
||||
"cell_type": "code",
|
||||
"execution_count": null,
|
||||
"id": "b0d759f9",
|
||||
"metadata": {},
|
||||
"outputs": [
|
||||
{
|
||||
"name": "stdout",
|
||||
"output_type": "stream",
|
||||
"text": [
|
||||
"[{'name': 'get_weather', 'args': {'location': 'San Francisco'}, 'id': 'a6248087-74c5-4b7c-9250-f335e642927c', 'type': 'tool_call'}]\n"
|
||||
]
|
||||
},
|
||||
{
|
||||
"data": {
|
||||
"text/plain": [
|
||||
"AIMessage(content=\"OK. It's sunny in San Francisco.\", additional_kwargs={}, response_metadata={'prompt_feedback': {'block_reason': 0, 'safety_ratings': []}, 'finish_reason': 'STOP', 'model_name': 'gemini-2.0-flash', 'safety_ratings': []}, id='run-ac5bb52c-e244-4c72-9fbc-fb2a9cd7a72e-0', usage_metadata={'input_tokens': 29, 'output_tokens': 11, 'total_tokens': 40, 'input_token_details': {'cache_read': 0}})"
|
||||
]
|
||||
},
|
||||
"execution_count": 28,
|
||||
"metadata": {},
|
||||
"output_type": "execute_result"
|
||||
}
|
||||
],
|
||||
"source": [
|
||||
"from langchain_google_genai import ChatGoogleGenerativeAI\n",
|
||||
"from langchain_core.tools import tool\n",
|
||||
"\n",
|
||||
"# Define the tool\n",
|
||||
"@tool(description=\"Get the current weather in a given location\")\n",
|
||||
"def get_weather(location: str) -> str:\n",
|
||||
" return \"It's sunny.\"\n",
|
||||
"\n",
|
||||
"# Initialize the model and bind the tool\n",
|
||||
"llm = ChatGoogleGenerativeAI(model=\"gemini-2.0-flash\")\n",
|
||||
"llm_with_tools = llm.bind_tools([get_weather])\n",
|
||||
"\n",
|
||||
"# Invoke the model with a query that should trigger the tool\n",
|
||||
"query = \"What's the weather in San Francisco?\"\n",
|
||||
"ai_msg = llm_with_tools.invoke(query)\n",
|
||||
"\n",
|
||||
"# Check the tool calls in the response\n",
|
||||
"print(ai_msg.tool_calls)\n",
|
||||
"\n",
|
||||
"# Example tool call message would be needed here if you were actually running the tool\n",
|
||||
"from langchain_core.messages import ToolMessage\n",
|
||||
"tool_message = ToolMessage(content=get_weather(*ai_msg.tool_calls[0]['args']), tool_call_id=ai_msg.tool_calls[0]['id'])\n",
|
||||
"llm_with_tools.invoke([ai_msg, tool_message]) # Example of passing tool result back\n"
|
||||
]
|
||||
},
|
||||
{
|
||||
"cell_type": "markdown",
|
||||
"id": "91d42b86",
|
||||
"metadata": {},
|
||||
"source": [
|
||||
"## Structured Output\n",
|
||||
"\n",
|
||||
"Force the model to respond with a specific structure using Pydantic models."
|
||||
]
|
||||
},
|
||||
{
|
||||
"cell_type": "code",
|
||||
"execution_count": 14,
|
||||
"id": "7457dbe4",
|
||||
"metadata": {},
|
||||
"outputs": [
|
||||
{
|
||||
"name": "stdout",
|
||||
"output_type": "stream",
|
||||
"text": [
|
||||
"name='Abraham Lincoln' height_m=1.93\n"
|
||||
]
|
||||
}
|
||||
],
|
||||
"source": [
|
||||
"from langchain_core.pydantic_v1 import BaseModel, Field\n",
|
||||
"from langchain_google_genai import ChatGoogleGenerativeAI\n",
|
||||
"\n",
|
||||
"# Define the desired structure\n",
|
||||
"class Person(BaseModel):\n",
|
||||
" '''Information about a person.'''\n",
|
||||
" name: str = Field(..., description=\"The person's name\")\n",
|
||||
" height_m: float = Field(..., description=\"The person's height in meters\")\n",
|
||||
"\n",
|
||||
"# Initialize the model\n",
|
||||
"llm = ChatGoogleGenerativeAI(model=\"gemini-2.0-flash\", temperature=0)\n",
|
||||
"structured_llm = llm.with_structured_output(Person)\n",
|
||||
"\n",
|
||||
"# Invoke the model with a query asking for structured information\n",
|
||||
"result = structured_llm.invoke(\"Who was the 16th president of the USA, and how tall was he in meters?\")\n",
|
||||
"print(result)\n"
|
||||
]
|
||||
},
|
||||
{
|
||||
"cell_type": "markdown",
|
||||
"id": "90d4725e",
|
||||
"metadata": {},
|
||||
"source": [
|
||||
"\n",
|
||||
"\n",
|
||||
"## Token Usage Tracking\n",
|
||||
"\n",
|
||||
"Access token usage information from the response metadata."
|
||||
]
|
||||
},
|
||||
{
|
||||
"cell_type": "code",
|
||||
"execution_count": 18,
|
||||
"id": "edcc003e",
|
||||
"metadata": {},
|
||||
"outputs": [
|
||||
{
|
||||
"name": "stdout",
|
||||
"output_type": "stream",
|
||||
"text": [
|
||||
"Prompt engineering is the art and science of crafting effective text prompts to elicit desired and accurate responses from large language models.\n",
|
||||
"\n",
|
||||
"Usage Metadata:\n",
|
||||
"{'input_tokens': 10, 'output_tokens': 24, 'total_tokens': 34, 'input_token_details': {'cache_read': 0}}\n"
|
||||
]
|
||||
}
|
||||
],
|
||||
"source": [
|
||||
"from langchain_google_genai import ChatGoogleGenerativeAI\n",
|
||||
"\n",
|
||||
"llm = ChatGoogleGenerativeAI(model=\"gemini-2.0-flash\")\n",
|
||||
"\n",
|
||||
"result = llm.invoke(\"Explain the concept of prompt engineering in one sentence.\")\n",
|
||||
"\n",
|
||||
"print(result.content)\n",
|
||||
"print(\"\\nUsage Metadata:\")\n",
|
||||
"print(result.usage_metadata)"
|
||||
]
|
||||
},
|
||||
{
|
||||
"cell_type": "markdown",
|
||||
"id": "28950dbc",
|
||||
"metadata": {},
|
||||
"source": [
|
||||
"## Built-in tools\n",
|
||||
"\n",
|
||||
"Google Gemini supports a variety of built-in tools ([google search](https://ai.google.dev/gemini-api/docs/grounding/search-suggestions), [code execution](https://ai.google.dev/gemini-api/docs/code-execution?lang=python)), which can be bound to the model in the usual way."
|
||||
]
|
||||
},
|
||||
{
|
||||
"cell_type": "code",
|
||||
"execution_count": null,
|
||||
"id": "dd074816",
|
||||
"metadata": {},
|
||||
"outputs": [
|
||||
{
|
||||
"name": "stdout",
|
||||
"output_type": "stream",
|
||||
"text": [
|
||||
"The next total solar eclipse visible in the United States will occur on August 23, 2044. However, the path of totality will only pass through Montana, North Dakota, and South Dakota.\n",
|
||||
"\n",
|
||||
"For a total solar eclipse that crosses a significant portion of the continental U.S., you'll have to wait until August 12, 2045. This eclipse will start in California and end in Florida.\n"
|
||||
]
|
||||
}
|
||||
],
|
||||
"source": [
|
||||
"from google.ai.generativelanguage_v1beta.types import Tool as GenAITool\n",
|
||||
"\n",
|
||||
"resp = llm.invoke(\n",
|
||||
" \"When is the next total solar eclipse in US?\",\n",
|
||||
" tools=[GenAITool(google_search={})],\n",
|
||||
")\n",
|
||||
"\n",
|
||||
"print(resp.content)"
|
||||
]
|
||||
},
|
||||
{
|
||||
"cell_type": "code",
|
||||
"execution_count": 43,
|
||||
"id": "6964be2d",
|
||||
"metadata": {},
|
||||
"outputs": [
|
||||
{
|
||||
"name": "stdout",
|
||||
"output_type": "stream",
|
||||
"text": [
|
||||
"Executable code: print(2*2)\n",
|
||||
"\n",
|
||||
"Code execution result: 4\n",
|
||||
"\n",
|
||||
"2*2 is 4.\n"
|
||||
]
|
||||
},
|
||||
{
|
||||
"name": "stderr",
|
||||
"output_type": "stream",
|
||||
"text": [
|
||||
"/Users/philschmid/projects/google-gemini/langchain/.venv/lib/python3.9/site-packages/langchain_google_genai/chat_models.py:580: UserWarning: \n",
|
||||
" ⚠️ Warning: Output may vary each run. \n",
|
||||
" - 'executable_code': Always present. \n",
|
||||
" - 'execution_result' & 'image_url': May be absent for some queries. \n",
|
||||
"\n",
|
||||
" Validate before using in production.\n",
|
||||
"\n",
|
||||
" warnings.warn(\n"
|
||||
]
|
||||
}
|
||||
],
|
||||
"source": [
|
||||
"from google.ai.generativelanguage_v1beta.types import Tool as GenAITool\n",
|
||||
"\n",
|
||||
"resp = llm.invoke(\n",
|
||||
" \"What is 2*2, use python\",\n",
|
||||
" tools=[GenAITool(code_execution={})],\n",
|
||||
")\n",
|
||||
"\n",
|
||||
"for c in resp.content:\n",
|
||||
" if isinstance(c, dict):\n",
|
||||
" if c[\"type\"] == 'code_execution_result':\n",
|
||||
" print(f\"Code execution result: {c['code_execution_result']}\")\n",
|
||||
" elif c[\"type\"] == 'executable_code':\n",
|
||||
" print(f\"Executable code: {c['executable_code']}\")\n",
|
||||
" else:\n",
|
||||
" print(c)"
|
||||
]
|
||||
},
|
||||
{
|
||||
"cell_type": "markdown",
|
||||
"id": "a27e6ff4",
|
||||
"metadata": {},
|
||||
"source": [
|
||||
"## Native Async\n",
|
||||
"\n",
|
||||
"Use asynchronous methods for non-blocking calls."
|
||||
]
|
||||
},
|
||||
{
|
||||
"cell_type": "code",
|
||||
"execution_count": 17,
|
||||
"id": "c6803e57",
|
||||
"metadata": {},
|
||||
"outputs": [
|
||||
{
|
||||
"name": "stdout",
|
||||
"output_type": "stream",
|
||||
"text": [
|
||||
"Async Invoke Result: The sky is blue due to a phenomenon called **Rayle...\n",
|
||||
"\n",
|
||||
"Async Stream Result:\n",
|
||||
"The thread is free, it does not wait,\n",
|
||||
"For answers slow, or tasks of fate.\n",
|
||||
"A promise made, a future bright,\n",
|
||||
"It moves ahead, with all its might.\n",
|
||||
"\n",
|
||||
"A callback waits, a signal sent,\n",
|
||||
"When data's read, or job is spent.\n",
|
||||
"Non-blocking code, a graceful dance,\n",
|
||||
"Responsive apps, a fleeting glance.\n",
|
||||
"\n",
|
||||
"Async Batch Results: ['1 + 1 = 2', '2 + 2 = 4']\n"
|
||||
]
|
||||
}
|
||||
],
|
||||
"source": [
|
||||
"from langchain_google_genai import ChatGoogleGenerativeAI\n",
|
||||
"\n",
|
||||
"llm = ChatGoogleGenerativeAI(model=\"gemini-2.0-flash\")\n",
|
||||
"\n",
|
||||
"async def run_async_calls():\n",
|
||||
" # Async invoke\n",
|
||||
" result_ainvoke = await llm.ainvoke(\"Why is the sky blue?\")\n",
|
||||
" print(\"Async Invoke Result:\", result_ainvoke.content[:50] + \"...\")\n",
|
||||
"\n",
|
||||
" # Async stream\n",
|
||||
" print(\"\\nAsync Stream Result:\")\n",
|
||||
" async for chunk in llm.astream(\"Write a short poem about asynchronous programming.\"):\n",
|
||||
" print(chunk.content, end=\"\", flush=True)\n",
|
||||
" print(\"\\n\")\n",
|
||||
"\n",
|
||||
" # Async batch\n",
|
||||
" results_abatch = await llm.abatch([\"What is 1+1?\", \"What is 2+2?\"])\n",
|
||||
" print(\"Async Batch Results:\", [res.content for res in results_abatch])\n",
|
||||
"\n",
|
||||
"await run_async_calls()"
|
||||
]
|
||||
},
|
||||
{
|
||||
"cell_type": "markdown",
|
||||
"id": "99204b32",
|
||||
"metadata": {},
|
||||
"source": [
|
||||
"## Safety Settings\n",
|
||||
@ -421,8 +835,8 @@
|
||||
},
|
||||
{
|
||||
"cell_type": "code",
|
||||
"execution_count": 14,
|
||||
"id": "238b2f96-e573-4fac-bbf2-7e52ad926833",
|
||||
"execution_count": null,
|
||||
"id": "d4c14039",
|
||||
"metadata": {},
|
||||
"outputs": [],
|
||||
"source": [
|
||||
@ -442,7 +856,7 @@
|
||||
},
|
||||
{
|
||||
"cell_type": "markdown",
|
||||
"id": "5805d40c-deb8-4924-8e72-a294a0482fc9",
|
||||
"id": "dea38fb1",
|
||||
"metadata": {},
|
||||
"source": [
|
||||
"For an enumeration of the categories and thresholds available, see Google's [safety setting types](https://ai.google.dev/api/python/google/generativeai/types/SafetySettingDict)."
|
||||
@ -450,7 +864,7 @@
|
||||
},
|
||||
{
|
||||
"cell_type": "markdown",
|
||||
"id": "3a5bb5ca-c3ae-4a58-be67-2cd18574b9a3",
|
||||
"id": "d6d0e853",
|
||||
"metadata": {},
|
||||
"source": [
|
||||
"## API reference\n",
|
||||
@ -461,7 +875,7 @@
|
||||
],
|
||||
"metadata": {
|
||||
"kernelspec": {
|
||||
"display_name": "Python 3 (ipykernel)",
|
||||
"display_name": ".venv",
|
||||
"language": "python",
|
||||
"name": "python3"
|
||||
},
|
||||
@ -475,7 +889,7 @@
|
||||
"name": "python",
|
||||
"nbconvert_exporter": "python",
|
||||
"pygments_lexer": "ipython3",
|
||||
"version": "3.10.4"
|
||||
"version": "3.9.6"
|
||||
}
|
||||
},
|
||||
"nbformat": 4,
|
||||
|
||||
Loading…
Reference in New Issue
Block a user