Introduction
We present a drafting workflow that keeps bibliography editing, compilation, and revision history in the same environment.
The agent proposes wording changes, diagnoses compile failures, and tracks the exact file scope before it writes.
AI is polishing references and the abstract
AI is polishing references and the abstract
Pin down the target
Rewrite only the abstract section in `main.tex` and leave every other file alone.
Compile check
Detected a conflict between `biblatex` and the journal template, and suggested a fix.
Version Flow
12
Draft versions and restore points made for repetitive paper revisions.
Team Mode
Live
Shared links, collaborative comments, and compile results all stay on the same page.
One workflow, three layers
Not stuffing AI into the editor,
but re-laying out the paper workflow.
The homepage does not list features. It shows how this workspace connects writing, verification, and submission for the user.
Draft
Write only when the goal is clear
Separate questions from write requests so AI does not accidentally enter draft mode for prompts like "how should I change this".
Review
Compile errors and fixes appear side by side
Show authors what failed, why it failed, and which files will change after the fix.
Deploy
Fits solo work and campus deployments alike
Ready for public cloud use, while private deployments keep data boundaries clear for labs and enterprise knowledge teams.
AI Automated Review
Before you submit,
let AI review it first.
Simulates a reviewer's perspective to automatically check formatting compliance, citation completeness, logical coherence, and language quality — surfacing issues before submission.
Format compliance
Margins, spacing, and fonts meet the journal template requirements
Citation check
Section 3 has 2 references not found in the bibliography
Logical coherence
Arguments and evidence are well-matched; transitions are natural
Language quality
Abstract paragraph 2 contains a long sentence (>60 chars); consider splitting
Review Summary
AI-based automated review does not replace human peer review
For labs and institutions
A deployment version for labs, research groups, and enterprise knowledge teams.
Keep data inside your own boundary
Private deployments support local object storage, your own compile environment, and internal model access for sensitive research teams.
Built to make Chinese paper workflows easier
Templates, terminology, and multi-file structures are designed around mixed Chinese/English text, bibliography cleanup, and pre-submission review.
Keep extending it to fit your flow
Project management, collaboration permissions, model vendors, and compile queues can all be tailored to the way you work.
AI Model Architecture
Hybrid architecture powered by domestic LLMs
We combine leading Chinese large language models into a hybrid reasoning architecture, balancing Chinese understanding, code generation, and long-context processing for the best academic writing experience.
Kimi K2.6
By Moonshot AI, excels at long-context understanding and multi-turn dialogue
DeepSeek V4
By DeepSeek, powerful reasoning and code generation capabilities
GLM-5.1
By Zhipu AI, outstanding Chinese understanding and academic text processing
Models are automatically routed by task type — no manual selection needed
Start from the first paragraph
Next time you write a paper, let the editor feel like a desk that is already set up.
Start with a 7-day trial. When it ends, you will automatically fall back to the free plan without getting interrupted by a forgotten cancellation.