SCIOS

Your Own Personal AI Research Assistant

An intelligent, locally-run academic agent focused on dynamic research landscape analysis and interactive assistant workflows.

🌟 Features

• Dynamic Research Landscape Build a structured, visual research workspace for a topic through three core modules:

  1. Tech Tree (technology evolution paths),
  2. Collaboration Network (scholar groups and co-author relationships),
  3. Research Gaps (evidence-grounded open problems and opportunities). Tasks are managed in a unified left-sidebar workspace.

• Interactive Academic Assistant A powerful AI agent equipped with a local sandbox workspace. It can help you search for literature, analyze experimental data, and even write and compile LaTeX documents autonomously.

📸 Screenshots

Landscape Workspace

Create and manage landscape tasks in a sidebar, then inspect each task's visual analysis in the main panel.

Tech Tree

Visualize methodology evolution, key milestones, and branch structures over time.

Collaboration Network

Inspect active scholars, collaboration strength, and major research groups in the topic.

Research Gaps

Surface high-value open questions with evidence papers and suggested research directions.

Assistant Mode

Use the interactive assistant to search papers, edit LaTeX files, and compile PDFs directly in the local workspace.

🧰 Built-in Agent Tools

The Interactive Assistant is equipped with a variety of powerful tools to perform complex tasks:

• Academic Search (search_academic_papers): Search and retrieve papers from academic sources.
• Web Search (web_search): Access the internet via Tavily to find the latest news, tutorials, and general knowledge.
• Workspace Operations (read_file, write_file, edit_file, glob_search): Read, create, and precisely edit files within a secure local sandbox.
• Persistent Shell (run_bash_command): Execute shell commands, manage files, and install dependencies in a persistent bash session.
• Python REPL (run_python_code): Execute Python code for data analysis, plotting, or testing algorithms.
• LaTeX Compiler (compile_latex): Automatically compile .tex files into PDF documents and fix compilation errors.
• Data Parser (parse_csv_log): Quickly extract and analyze metrics from experimental CSV logs.

📚 Supported Academic Sources

SCIOS currently uses a focused source strategy:

• Landscape pipeline: Semantic Scholar only (papers, citations, authors, and scholar profiles)
• Assistant tools: Semantic Scholar + Tavily Web Search (tool-augmented exploration and web context lookup)

---

🚀 Getting Started

SCIOS is split into a Python backend (FastAPI) and a Next.js frontend. Below are the instructions to set up and run the project.

Prerequisites

• Python >= 3.10
• Node.js >= 18.x
• uv (Recommended Python package manager)
• pdflatex (Optional, required only if you want the Assistant to compile LaTeX)

1. Backend Setup

# Navigate to the backend directory
cd backend

# Copy the environment template
cp .env.example .env

Configuration (.env)

You must configure the API keys in your .env file for SCIOS to function properly:

• LLM_API_KEY: API key for your selected provider.
• LLM_BASE_URL: API base URL. Keep the default for OpenAI; set it for OpenAI-compatible providers.
• LLM_MODEL: Model name in LiteLLM format.
• TAVILY_API_KEY: API key for Tavily Web Search.
• (Optional) Keep additional source-related variables for future experimentation. Current core flow uses:
  • Landscape: Semantic Scholar
  • Assistant: Semantic Scholar + Tavily

Multi-model provider examples (via LiteLLM)

# OpenAI
LLM_BASE_URL=https://api.openai.com/v1
LLM_MODEL=gpt-4o
LLM_API_KEY=<OPENAI_API_KEY>

# Anthropic Claude
LLM_BASE_URL=
LLM_MODEL=anthropic/claude-3-5-sonnet-20241022
LLM_API_KEY=<ANTHROPIC_API_KEY>

# Google Gemini
LLM_BASE_URL=
LLM_MODEL=gemini/gemini-1.5-pro
LLM_API_KEY=<GEMINI_API_KEY>

# DeepSeek
LLM_BASE_URL=https://api.deepseek.com/v1
LLM_MODEL=deepseek-chat
LLM_API_KEY=<DEEPSEEK_API_KEY>

Run the Backend

Instead of using development hot-reloading, run the backend using standard FastAPI commands:

# Install dependencies
uv sync

# Run the backend server
uv run fastapi run src/main.py --host 0.0.0.0 --port 8000

The backend API will be available at http://localhost:8000. API documentation is at http://localhost:8000/docs.

2. Frontend Setup

Open a new terminal window to start the Next.js frontend.

# Navigate to the frontend directory
cd frontend

# Install dependencies
npm install

# Build the project for production
npm run build

# Start the production server
npm run start

The frontend interface will be accessible at http://localhost:3000.

---

💡 How to Use

1. Landscape: Open http://localhost:3000, create a topic task in the left sidebar, and follow real-time task progress. Once completed, inspect tech tree, collaboration graph, and research gaps in the main workspace.
2. Assistant: Switch to the Assistant tab for tool-augmented academic workflows. For example: "Search recent papers on graph neural networks, summarize top 3, draft a LaTeX intro, and compile PDF."