WriteHERE is an open-source AI writing framework designed for long-form content generation, targeting researchers and users who need adaptive and dynamic writing assistance. It addresses the limitations of rigid AI writing workflows by employing a novel Heterogeneous Recursive Planning approach, enabling more human-like writing processes and outperforming state-of-the-art methods in both fiction and technical report generation.

How It Works

WriteHERE utilizes a recursive planning mechanism that breaks down complex writing tasks into smaller, manageable subtasks. This approach allows for the heterogeneous integration of various task types, including retrieval, reasoning, and composition, and dynamically adapts the writing process based on real-time context. This dynamic adaptation, akin to human writing behavior, is a key differentiator from static, workflow-bound AI writers.

Quick Start & Requirements

• Installation: pip install -v -e . (after setting up a virtual environment).
• Prerequisites: Python 3.6+, Node.js 14+ (for frontend), API keys for OpenAI, Anthropic, and SerpAPI.
• Setup: A one-step setup script (./setup_env.sh and ./start.sh) is available for launching the backend and frontend. An alternative script (./run_with_anaconda.sh) is provided for Anaconda users.
• Documentation: https://arxiv.org/abs/2503.08275 (cited paper)

Highlighted Details

• Recursive task decomposition for complex writing.
• Dynamic integration of retrieval, reasoning, and composition.
• Adaptive workflow adjusts to context and requirements.
• Real-time visualization of the agent's "thinking process."

Maintenance & Community

The project welcomes community contributions via pull requests and issue reports. Further details on contributing and community support can be found in the README.

Licensing & Compatibility

Licensed under the MIT License, allowing for free use, modification, and distribution for research, educational, and personal purposes. Commercial use is not explicitly mentioned but is generally permitted under MIT, though the project states it's developed for non-commercial interests.

Limitations & Caveats

The project is primarily developed for research and educational purposes, with an explicit "Non-Commercial" interest stated. Users must provide their own API keys for LLM access, incurring associated costs.