标签：multi url image agent langchain modal import tools support

How to pass multimodal data directly to models

https://python.langchain.com/v0.2/docs/how_to/multimodal_inputs/

message = HumanMessage(
    content=[
        {"type": "text", "text": "describe the weather in this image"},
        {"type": "image_url", "image_url": {"url": image_url}},
    ],
)
response = model.invoke([message])
print(response.content)

message = HumanMessage(
    content=[
        {"type": "text", "text": "describe the weather in this image"},
        {
            "type": "image_url",
            "image_url": {"url": f"data:image/jpeg;base64,{image_data}"},
        },
    ],
)
response = model.invoke([message])
print(response.content)

from typing import Literal

from langchain_core.tools import tool


@tool
def weather_tool(weather: Literal["sunny", "cloudy", "rainy"]) -> None:
    """Describe the weather"""
    pass


model_with_tools = model.bind_tools([weather_tool])

message = HumanMessage(
    content=[
        {"type": "text", "text": "describe the weather in this image"},
        {"type": "image_url", "image_url": {"url": image_url}},
    ],
)
response = model_with_tools.invoke([message])
print(response.tool_calls)

How to use ImagePromptTemplate

https://github.com/langchain-ai/langchain/discussions/20820

from langchain.agents import AgentExecutor, create_openai_tools_agent
from langchain.schema import SystemMessage
from langchain_community.tools import CopyFileTool
from langchain_core.prompts import (
    ChatPromptTemplate,
    HumanMessagePromptTemplate,
    MessagesPlaceholder,
)
from langchain_openai import ChatOpenAI

# Define the prompt template with the image_url
prompt_messages = [
    SystemMessage(content="Describe the image provided"),
    HumanMessagePromptTemplate.from_template(
        template=[
            {"type": "image_url", "image_url": {"url": "{image_url}"}},
        ]
    ),
    MessagesPlaceholder("agent_scratchpad"),
]

prompt = ChatPromptTemplate(messages=prompt_messages)

# Define the LLM model
llm = ChatOpenAI(model="gpt-4o")

# Define a dummy tool
tools = [CopyFileTool()]

# Create an Agent that can be used to call the tools we defined
agent = create_openai_tools_agent(llm, tools, prompt)

# Define the AgentExecutor properly
agent_executor = AgentExecutor(
    agent=agent,
    tools=tools,
)

# Example image URL you want to pass in at runtime
image_url = "https://fastly.picsum.photos/id/450/200/200.jpg?hmac=DluUYibC-zBoNHLOHsO6aHIuiA3pDhholFjiR5KcwR0"

# Execute the Agent with the image URL
response = agent_executor.invoke({"image_url": image_url, "agent_scratchpad": []})

Image-caption-with-LLM-and-Langchain

https://github.com/stemgene/Image-caption-with-LLM-and-Langchain/blob/main/main.py

import streamlit as st
import os
import requests
import torch
from PIL import Image
from langchain.agents import initialize_agent
from langchain.chat_models import ChatOpenAI
from langchain.chains.conversation.memory import ConversationBufferWindowMemory
from tools import ImageCaptionTool, ObjectDetectionTool
from tempfile import NamedTemporaryFile

########################
### Initialize agent ###
########################

tools = [ImageCaptionTool(), ObjectDetectionTool()]
conversational_memory = ConversationBufferWindowMemory(
    memory_key='chat_history',
    k=5,
    return_messages=True
)
llm = ChatOpenAI(
    # openai_api_key=os.environ.get("OPENAI_API_KEY"), # for local usage
    openai_api_key=st.secrets['auth_key'],
    temperature=0,
    model_name='gpt-3.5-turbo'
)
agent = initialize_agent(
    agent='chat-conversational-react-description',
    tools=tools,
    llm=llm,
    max_iterations=5,
    verbose=True,
    memory=conversational_memory,
    early_stoppy_method='generate'
)
# Set title
st.title("Ask a question to an image")
# set header
st.header("Please upload an image")

# upload file
# file = st.file_uploader("", type=['jpg', 'png', 'jpeg'])

# input image url
image_url = st.text_input('Please input the image URL')


if image_url and image_url != "":
    image_object = Image.open(requests.get(image_url, stream=True).raw).convert("RGB")

    # display image
    
    #st.image(file, use_column_width=True)
    st.image(image_object, use_column_width=True)

    # text input
    user_question = st.text_input('Ask a question your image:')

    ##############################
    ### Compute agent response ###
    ##############################

    # with NamedTemporaryFile(dir='.', mode='w+b') as f:
    #     f.write(file.getbuffer())
    #     image_path = f.name

    #     # write agent response
    #     if user_question and user_question != "":
    #         with st.spinner(text='In progress...'):
    #             response = agent.run('{}, this is the image path: {}'.format(user_question, image_path))
    #             st.write(response)

    #write agent response
    if user_question and user_question != "":
        with st.spinner(text='In progress...'):
            response = agent.run('{}, this is the image path: {}'.format(user_question, image_url))
            st.write(response)

how_to_use_crew_ai_for_image_understanding

https://github.com/mdwoicke/crewai_multi_agent/blob/main/01_how_to_use_crew_ai_for_image_understanding.py

"""
Author: Rajib Deb
Date: 02/10/2024
Description: This is the driver program that starts the MasterChef crew.
"""
from crew_ai_crews.master_chef import MasterChefCrew

if __name__ == "__main__":
    # url="https://rumkisgoldenspoon.com/wp-content/uploads/2022/02/Aar-macher-jhol.jpg"
    url="https://m.media-amazon.com/images/I/51dFvTRE3iL.__AC_SX300_SY300_QL70_FMwebp_.jpg"
    crew = MasterChefCrew(url=url)
    print(crew.kickoff())

分析菜肴成分，并给出做法

"""
Author: Rajib Deb
Date: 02/10/2024
Description: This module creates the crew objects with all the required agents and tasks
"""
import os

from crewai import Crew
from dotenv import load_dotenv

from crew_ai_crews.agents import MasterChef
from crew_ai_crews.tasks import ExtractIngredientsFromImage

load_dotenv()
OPENAI_API_KEY = os.environ.get('OPENAI_API_KEY')


class MasterChefCrew():
    """
    The master chef crew consisting of the food connoisseur and the chef
    """
    def __init__(self, url):
        agents = MasterChef(url=url)
        self.extract_agent = agents.get_recipe_from_image()
        self.instruct_agent = agents.instruct_to_cook_dish()

    def kickoff(self):
        """
        Creating the crew with the two agents
        :return:
        """
        tasks = ExtractIngredientsFromImage()
        crew = Crew(
            agents=[self.extract_agent, self.instruct_agent],
            tasks=[tasks.get_ingredients(self.extract_agent), tasks.cooking_instruction(self.instruct_agent)],
            verbose=True
        )
        result = crew.kickoff()
        return result

LangChain vs. CrewAI: Comparing AI Development Platforms

https://smythos.com/ai-agents/ai-agent-builders/langchain-vs-crewai/

crewai不支持多模态，其本身是workflow定义

Feature Comparison Table

LangChain CrewAI SmythOS

CORE FEATURES

Hosted Agents (Dev, Production) ✅ ❌ ✅

Environments (Dev, Production) ✅ ❌ ✅

Visual Builder ❌ ❌ ✅

No-Code Options ❌ ❌ ✅

Explainability & Transparency ✅ ❌ ✅

Debug Tools ✅ ❌ ✅

Multimodal ✅ ❌ ✅

Audit Logs for Analytics ✅ ❌ ✅

Agent Work Scheduler ❌ ✅ ✅

SECURITY

Constrained Alignment ❌ ❌ ✅

Data Encryption ✅ ❌ ✅

OAuth ✅ ❌ ✅

IP Control ❌ ❌ ✅

COMPONENTS

Foundation AIs ✅ ❌ ✅

Huggingface AIs ✅ ❌ ✅

Zapier APIs ❌ ❌ ✅

All other APIs, RPA ✅ ❌ ✅

Classifiers ✅ ❌ ✅

Logic ✅ ❌ ✅

Data Lakes ❌ ❌ ✅

DEPLOYMENT OPTIONS (EMBODIMENTS)

Deploy as API ✅ ❌ ✅

Deploy as Webhook ❌ ❌ ✅

Staging Domains ❌ ❌ ✅

Production Domains ❌ ❌ ✅

API Authentication (OAuth + Key) ✅ ❌ ✅

Deploy as Site Chat ✅ ❌ ✅

Deploy as Scheduled Agent ❌ ❌ ✅

Deploy as GPT ✅ ❌ ✅

DATA LAKE SUPPORT

Hosted Vector Database ❌ ❌ ✅

Sitemap Crawler ❌ ❌ ✅

YouTube Transcript Crawler ❌ ❌ ✅

URL Crawler ❌ ❌ ✅

PDF Support ✅ ❌ ✅

Word File Support ❌ ❌ ✅

TXT File Support ✅ ❌ ✅

	LangChain	CrewAI	SmythOS
CORE FEATURES
Hosted Agents (Dev, Production)	✅	❌	✅
Environments (Dev, Production)	✅	❌	✅
Visual Builder	❌	❌	✅
No-Code Options	❌	❌	✅
Explainability & Transparency	✅	❌	✅
Debug Tools	✅	❌	✅
Multimodal	✅	❌	✅
Audit Logs for Analytics	✅	❌	✅
Agent Work Scheduler	❌	✅	✅
SECURITY
Constrained Alignment	❌	❌	✅
Data Encryption	✅	❌	✅
OAuth	✅	❌	✅
IP Control	❌	❌	✅
COMPONENTS
Foundation AIs	✅	❌	✅
Huggingface AIs	✅	❌	✅
Zapier APIs	❌	❌	✅
All other APIs, RPA	✅	❌	✅
Classifiers	✅	❌	✅
Logic	✅	❌	✅
Data Lakes	❌	❌	✅
DEPLOYMENT OPTIONS (EMBODIMENTS)
Deploy as API	✅	❌	✅
Deploy as Webhook	❌	❌	✅
Staging Domains	❌	❌	✅
Production Domains	❌	❌	✅
API Authentication (OAuth + Key)	✅	❌	✅
Deploy as Site Chat	✅	❌	✅
Deploy as Scheduled Agent	❌	❌	✅
Deploy as GPT	✅	❌	✅
DATA LAKE SUPPORT
Hosted Vector Database	❌	❌	✅
Sitemap Crawler	❌	❌	✅
YouTube Transcript Crawler	❌	❌	✅
URL Crawler	❌	❌	✅
PDF Support	✅	❌	✅
Word File Support	❌	❌	✅
TXT File Support	✅	❌	✅

https://python.langchain.com.cn/docs/use_cases/agents/multi_modal_output_agent

audio理解

https://js.langchain.com/v0.2/docs/how_to/tool_calls_multimodal/

https://js.langchain.ac.cn/v0.2/docs/how_to/tool_calls_multimodal/

import { SystemMessage } from "@langchain/core/messages";
import { tool } from "@langchain/core/tools";

const summaryTool = tool(
  (input) => {
    return input.summary;
  },
  {
    name: "summary_tool",
    description: "Log the summary of the content",
    schema: z.object({
      summary: z.string().describe("The summary of the content to log"),
    }),
  }
);

const audioUrl =
  "https://www.pacdv.com/sounds/people_sound_effects/applause-1.wav";

const axiosRes = await axios.get(audioUrl, { responseType: "arraybuffer" });
const base64 = btoa(
  new Uint8Array(axiosRes.data).reduce(
    (data, byte) => data + String.fromCharCode(byte),
    ""
  )
);

const model = new ChatGoogleGenerativeAI({
  model: "gemini-1.5-pro-latest",
}).bindTools([summaryTool]);

const response = await model.invoke([
  new SystemMessage(
    "Summarize this content. always use the summary_tool in your response"
  ),
  new HumanMessage({
    content: [
      {
        type: "media",
        mimeType: "audio/wav",
        data: base64,
      },
    ],
  }),
]);

console.log(response.tool_calls);

How to use multimodal prompts

https://python.langchain.com/docs/how_to/multimodal_prompts/

https://blog.csdn.net/cheese_burger_/article/details/140351968

prompt = ChatPromptTemplate.from_messages(
    [
        ("system", "compare the two pictures provided"),
        (
            "user",
            [
                {
                    "type": "image_url",
                    "image_url": {"url": "data:image/jpeg;base64,{image_data1}"},
                },
                {
                    "type": "image_url",
                    "image_url": {"url": "data:image/jpeg;base64,{image_data2}"},
                },
            ],
        ),
    ]
)

快速封装替换自定义LLM（基于自定义API或本地模型）

https://blog.csdn.net/HYY_2000/article/details/138339333

LangGraph: Multi-Agent Workflows

https://blog.langchain.dev/langgraph-multi-agent-workflows/

Other Frameworks

LangGraph is not the first framework to support multi-agent workflows. Most of the difference between these frameworks largely lies in the mental models and concepts they introduce.

Autogen

Autogen emerged as perhaps the first multi-agent framework. The biggest difference in mental model between LangGraph and Autogen is in construction of the agents. LangGraph prefers an approach where you explicitly define different agents and transition probabilities, preferring to represent it as a graph. Autogen frames it more as a "conversation". We believe that this "graph" framing makes it more intuitive and provides better developer experience for constructing more complex and opinionated workflows where you really want to control the transition probabilities between nodes. It also supports workflows that aren't explicitly captured by "conversations."

Another key difference between Autogen and LangGraph is that LangGraph is fully integrated into the LangChain ecosystem, meaning you take fully advantage of all the LangChain integrations and LangSmith observability.

CrewAI

Another key framework we want to highlight is CrewAI. CrewAI has emerged recently as a popular way to create multi-agent "teams". Compared to LangGraph, CrewAI is a higher-level framework. In fact, we are actively working with the CrewAI team to integration LangGraph into CrewAI! We think CrewAI has arrived at an awesome higher level DevEx and we want to support that.

CrewAI Unleashed: Future of AI Agent Teams

https://blog.langchain.dev/crewai-unleashed-future-of-ai-agent-teams/

CrewAI represents a shift in AI agents by offering a thin framework that leverages collaboration and roleplaying, based on versatility and efficiency. It stands as a tool for engineers and creatives alike, enabling the seamless assembly of AI agents into cohesive, high-performing teams.

Whether it's transforming a single thought into a fully-fledged landing page or conducting complex idea analysis, CrewAI is adept at handling a diverse range of tasks through its processes.

The real-world applications of CrewAI, from boosting social media presence to building interactive landing pages, underscore its practicality and adaptability. Looking forward, CrewAI is set to evolve further, introducing more intricate processes and continuing to redefine the landscape of AI teamwork. With its user-friendly integration and customization options, CrewAI is not just a concept but a tangible, powerful tool to harness the power of AI Agents.

标签：multi,url,image,agent,langchain,modal,import,tools,support
From： https://www.cnblogs.com/lightsong/p/18471157

langchain multi modal support

How to pass multimodal data directly to models

How to use ImagePromptTemplate

Image-caption-with-LLM-and-Langchain

how_to_use_crew_ai_for_image_understanding

LangChain vs. CrewAI: Comparing AI Development Platforms

Feature Comparison Table

audio理解

How to use multimodal prompts

快速封装替换自定义LLM（基于自定义API或本地模型）

LangGraph: Multi-Agent Workflows

Other Frameworks

Autogen

CrewAI

CrewAI Unleashed: Future of AI Agent Teams

相关文章

赞助商

阅读排行

langchain multi modal support

How to pass multimodal data directly to models

How to use ImagePromptTemplate

Image-caption-with-LLM-and-Langchain

how_to_use_crew_ai_for_image_understanding

LangChain vs. CrewAI: Comparing AI Development Platforms

Feature Comparison Table

Multi-modal outputs: Image & Text

audio理解

How to use multimodal prompts

快速封装替换自定义LLM（基于自定义API或本地模型）

LangGraph: Multi-Agent Workflows

Other Frameworks

Autogen

CrewAI

CrewAI Unleashed: Future of AI Agent Teams

相关文章

赞助商

阅读排行