EvidionAI: An Automated AI Research System Based on Multi-Agent Workflows

EvidionAI is a Windows-based autonomous multi-agent AI research system developed by Hairspuriouscorrelation574. It automates end-to-end research tasks (literature retrieval, code execution, result analysis) using FastAPI and LangGraph, aiming to reduce researchers' repetitive workload and let them focus on creative thinking. Key features include multi-agent collaboration, task decomposition, structured output, and support for various research input modes.

In scientific research, literature review, data collection, and result analysis are time-consuming and repetitive. EvidionAI addresses this pain point by decomposing complex research tasks into sub-tasks handled by specialized agents. It is a Windows desktop application designed to automate the full research lifecycle, easing researchers' burden of transactional work.

EvidionAI uses a multi-agent system architecture where specialized agents handle specific sub-tasks (e.g., literature retrieval, code execution). Core tech stack:

• FastAPI: Backend framework with async support, auto API docs, and type safety for efficient agent communication.
• LangGraph: Orchestrates agent workflows (defines interaction graphs, manages state, supports persistence for long-running tasks).

Core capabilities:

1. Task decomposition & planning: Breaks down research topics into executable sub-tasks using LLM prompt engineering.
2. Literature retrieval: Automatically searches academic resources (data sources not detailed but implies integration with academic databases/search APIs).
3. Multi-angle analysis: Identifies consensus and disagreement in research findings.
4. Structured output: Organizes results for easy use in reports.

Input modes: Theme-driven (e.g., "find main views on X"), goal-oriented (e.g., compare tools), iterative refinement. Typical workflow: Start → Input topic → Agent planning → Auto execution → Result review → Export output.

Requirements:

• OS: Windows 10/11 (no macOS/Linux support).
• Hardware: ≥8GB RAM, ≥5GB storage, stable internet connection.
• Installation: Admin rights needed.

Optimization tips: Close memory-heavy apps, use simple file paths (e.g., C:\EvidionAI), organize folders (root/data/projects).

Scenarios:

1. Academic research: Literature review, scientific topic exploration, research foundation building.
2. Market/trend research: Market overview, trend tracking, competitive analysis.
3. Knowledge management: Structured note-taking, continuous learning.

Target users: Researchers, students, analysts needing frequent literature reviews or data organization.

Current limitations:

• Platform restriction (only Windows).
• Hardware requirements may exclude low-config devices.
• Tech docs are brief (focus on usage over implementation).
• Dependent on external AI/services (availability/cost impact experience).

Future directions: Cross-platform support (macOS/Linux), cloud SaaS version, more data source integration, collaboration features.

EvidionAI is a valuable auxiliary tool for automating repetitive research tasks, reducing entry barriers, improving efficiency, and aiding structured thinking. However:

• Quality depends on source data (potential bias/omission).
• Deep understanding and critical analysis still require human input.
• Users must comply with copyright and ethical rules for literature use.

Recommendation: Use as an initial research and data collection tool, not a replacement for human expertise.