deep-research

Gemini Deep Research - Autonomous Multi-step AI Research Tool

Overview of Capabilities

Gemini Deep Research is an autonomous research assistant that can automatically plan search strategies, read web content, and synthesize detailed research reports with citations.

Use Cases

1. Market Analysis and Competitive Research

Quickly complete market sizing, competitor product comparisons, and industry trend research. Suitable for product managers and strategy analysts conducting market entry assessments and gathering competitive intelligence.

2. Technical Literature Review and Technology Selection

Automatically aggregate technical documentation, compare the strengths and weaknesses of frameworks, and organize best practices. Suitable for developers to conduct systematic research before technology selection, saving time on manual searching and organizing.

3. Due Diligence and Historical Research

Conduct background checks on investment targets, compile company historical timelines, and analyze risk factors. Suitable for investors and researchers who need in-depth information.

Core Features

Autonomous Research Planning

The tool automatically breaks a research task into multiple search steps, performs web reading and information aggregation, and can complete a 2–10 minute deep research process without human intervention.

Structured Report Output

Supports Markdown, JSON, and Raw output formats. By default it generates a human-readable report with clear source citations, making verification and sharing easy.

Flexible Task Management

Supports real-time streamed progress viewing, background asynchronous execution, follow-up queries based on existing research, listing recent research records, and other full task lifecycle management features.

Frequently Asked Questions

How long does each Deep Research session take?

Depending on the complexity of the research topic, a single task usually takes 2–10 minutes. The system provides an estimated time at the start; you can use the --stream flag to view progress in real time or use --no-wait to run it in the background.

What does it cost to use this tool?

Each research run consumes about $2–5, depending on the complexity and number of search results. Token usage is roughly 250,000–900,000 input tokens and 60,000–80,000 output tokens.

How do I start my first research?

First obtain a Gemini API Key from Google AI Studio, set the GEMINI_API_KEY environment variable, then run:
python3 scripts/research.py --query "your research question" to start.