Add Matplotlib Agent Toolkit

libs/experimental/langchain_experimental/agents/agent_toolkits/matplotlib/README.md

@@ -0,0 +1,124 @@
+cat <<'EOF' > lib/langchain_experimental/agents/agent_toolkits/matplotlib/README.md
+# 🧠 Matplotlib Agent Toolkit
+
+The **Matplotlib Agent Toolkit** extends LangChain with intelligent data visualization capabilities using **Matplotlib** and **Python REPL** tools.  
+It enables agents to understand your data and **generate, execute, and explain plots** — all from natural language instructions.
+
+---
+
+## 🚀 Overview
+
+This toolkit lets a language model:
+- Analyze one or multiple Pandas DataFrames  
+- Generate valid Python + Matplotlib plotting code  
+- Execute it safely in a sandboxed REPL  
+- Save or display visualizations automatically  
+
+It supports all major LangChain agent types, including:
+- `AgentType.ZERO_SHOT_REACT_DESCRIPTION`
+- `AgentType.OPENAI_FUNCTIONS`
+- `"openai-tools"`
+- `"tool-calling"`
+
+---
+
+## ⚙️ Installation
+
+Clone the experimental repo and install dependencies:
+
+```bash
+git clone https://github.com/langchain-ai/langchain-experimental.git
+cd langchain-experimental
+pip install -e .
+```
+
+
+Ensure the following are installed
+```bash
+pip install pandas matplotlib langchain python-dotenv
+```
+
+## Example Usage:
+``` bash
+from langchain_openai import ChatOpenAI
+from langchain_experimental.agents.agent_toolkits.matplotlib.base import create_matplotlib_agent
+import pandas as pd
+from dotenv import load_dotenv
+
+# Load your OpenAI API key
+load_dotenv()
+
+# Load a sample dataset
+df = pd.read_csv("https://raw.githubusercontent.com/pandas-dev/pandas/main/doc/data/titanic.csv")
+
+# Initialize an LLM
+llm = ChatOpenAI(model="gpt-4o", temperature=0)
+
+# Create a Matplotlib-powered agent
+agent = create_matplotlib_agent(
+    llm=llm,
+    df=df,
+    verbose=True,
+    allow_dangerous_code=True,  # ⚠️ Enables Python execution for plotting
+)
+
+# Natural language plotting query
+query = """
+Create a scatter plot of Age vs Fare, color-coded by passenger class.
+Save the figure as 'titanic_scatter.png'.
+"""
+
+# Run the agent
+result = agent.invoke({"input": query})
+
+print("Agent output:", result.get("output", ""))
+print("✅ Plot generated and saved as titanic_scatter.png")
+
+```
+After Execution you will have a chart named titanic_scatter.png
+
+
+## Agent Creator:
+
+``` bash
+
+agent = create_matplotlib_agent(
+    llm=llm,
+    df=df,
+    verbose=True,
+    allow_dangerous_code=True,  
+)
+```
+
+
+## 🧩 Parameters
+
+| Parameter | Type | Description |
+|------------|------|-------------|
+| `llm` | `LanguageModelLike` | Any LLM supporting tool-calling (e.g. `ChatOpenAI`) |
+| `df` | `pd.DataFrame` or `List[pd.DataFrame]` | Dataset(s) for visualization |
+| `agent_type` | `AgentType` or `str` | Type of agent (`react`, `openai-functions`, etc.) |
+| `allow_dangerous_code` | `bool` | Must be `True` to enable Python REPL execution |
+| `include_df_in_prompt` | `bool` | Whether to embed sample DataFrame rows in the prompt |
+| `extra_tools` | `Sequence[BaseTool]` | Add additional LangChain tools if needed |
+
+
+## 🧠 How It Works
+
+Under the hood, `create_matplotlib_agent()`:
+
+- Builds a customized prompt based on your DataFrame(s)
+- Uses a Python REPL tool (`PythonAstREPLTool`) for safe code execution
+- Integrates Matplotlib and Pandas contexts for inline plotting
+- Returns an `AgentExecutor` ready to interpret natural language queries into plots
+
+---
+
+## 🧪 Running Tests
+
+To verify functionality:
+
+```bash
+pytest tests/test_matplotlib_toolkit.py -v
+
+

libs/experimental/langchain_experimental/agents/agent_toolkits/matplotlib/__init__.py

@@ -0,0 +1 @@
+"matplotlib"