Agentic Workspace

A modular LLM-based operating system for data analysis tasks.
It automatically interprets natural language requests and routes them to specialized agents such as SQL and visualization modules.

Features

• SQL Agent – Executes natural language queries against structured databases.
• Visualization Agent – Generates charts and plots automatically using LangChain’s pandas DataFrame agent. Automatically saves to plots folder.
• Modular Architecture – Easily extendable with new agent types or custom workflows.
• Context awareness – Maintains up to the 10 most recent user–system interactions to preserve conversational continuity.

Figure 1: LangGraph workflow showing agent routing and execution flow.

Results

The Results Notebook demonstrates successful end-to-end execution of the agentic workflow using the Chinook database. Key outcomes include:

• Correct routing of tasks between SQL, visualization, and summary agents.
• Automatic generation of SQL queries in response to natural language prompts.
• Automatic visualization of plots created directly from query outputs.
• Context-aware refinement of user requests (e.g., filtering results as demonstrated in the Notebook).

Figure 2: Query — "Query the database for all customers and create a bar chart showing the number of customers per country, sorted in descending order."

Figure 3: Query — "Generate the same chart as in the previous request, but exclude the United States while keeping all other countries."

These results confirm that the Agentic Workspace can autonomously process user queries, generate data insights, and adapt based on conversational context.

Setup

1. Clone the repository:

git clone <your-repo-url>
cd agentic-workspace

2. Install dependencies:

pip install -r requirements.txt

3. Configure environment variables: Create a .env file with your API key and other configuration parameters as in example.env file:

Project Structure:

agentic-workspace/
│
├── agents/
│   ├── __init__.py
│   ├── sql_agent.py
│   └── visualization_agent.py
│
├── utils/
│   ├── __init__.py
│   ├── config.py
│   └── orchestrator.py
|
├── plots/
│
├── main.py
├── Results Notebook.ipynb
├── requirements.txt
├── .env                   
└── README.md

Usage Examples (for relevant databases)

• "Plot a histogram of customer ages"
• "Query the database for top products"
• "Show summary statistics of sales data"
• "Create a scatter plot of price vs rating"

---

Author

Artur Garipov
LinkedIn | GitHub