A powerful AI-driven research assistant that automates discovering, summarizing, and analyzing datasets or structured information across multiple sources. Built with OpenAI Agents SDK, this project showcases an agentic pipeline that handles planning, searching, writing, and delivering reports.
- Planner Agent – Generates optimized search queries tailored for dataset discovery or any structured research topic.
- Search Agent – Searches multiple public repositories (Kaggle, UCI, data.gov, GitHub, AWS Open Data) and extracts dataset details in a structured Markdown table.
- Writer Agent – Converts search results into detailed Markdown reports with dataset overviews, use-case recommendations, and project ideas.
- Research Manager – Orchestrates the flow between planner, search, and writer agents.
- Email Agent – Sends the final report via email with clean HTML formatting.
- Deep Research – Core orchestrator that integrates all agents to create a seamless pipeline.
- Python (async/await for concurrent searches)
- OpenAI Agents SDK
- WebSearchTool for live search results
- Markdown for structured reports
- SendGrid API for email delivery
├── deep_research.py # Core orchestrator
├── email_agent.py # Agent to send report emails
├── planner_agent.py # Plans search queries
├── research_manager.py # Manages agent workflow
├── search_agent.py # Agent to perform web searches
├── writer_agent.py # Converts search results into reports
└── README.md # Project documentation
🧑💻 Example Use Case
Discover datasets about "AI impacting human jobs"
Automatically summarize datasets in Markdown tables
Generate 3–5 project ideas per dataset
Deliver the report via email
Extendable to any research topic (product recommendations, market research, analytics, etc.)
💡 Learnings
Agent orchestration and handoffs for task division
Async programming for faster concurrent searches
Structured outputs and Markdown reporting
Automating research and analysis workflows
🔗 References
OpenAI Agents SDK Documentation
Inspiration and mentorship: Ed Donner
📢 Contributions
Feel free to fork this project, submit issues, or create pull requests to improve the agentic workflow or extend its use to other research domains.