Wordbot: A Microsoft Word desktop plugin for Markdown formatting + local LLM + ZotSeek (using your local Zotero library). #46
Addy-ad
started this conversation in
Show and tell
Replies: 0 comments
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Uh oh!
There was an error while loading. Please reload this page.
Uh oh!
There was an error while loading. Please reload this page.
As a researcher, a huge chunk of my actual work isn't writing, it's fighting with Microsoft Word formatting. I do use AI tools to correct grammar, rephrase, write equations, draft, brainstorm, and etc. But the workflow always felt fundamentally broken: give context to an LLM in a browser, copy the response, paste it into Word, and waste time manually rebuilding tables, converting formulas into Word equations, fixing broken headers, and styling code blocks. Doing this dozens of times during a single paper drains focus and wastes hours on repetitive chores that add zero intellectual value.
I built Wordbot to solve my own frustration and eliminate that friction entirely.
What It Does
Automatic Markdown-to-Word Formatting
Instantly converts AI outputs into native Word tables, styled headers (H1–H6), code blocks, bulleted lists, and LaTeX equations. Format existing Markdown text on demand with a single click.
Seamless In-Document LLM Communication
Prompt any local model directly inside Word. Highlight text and click Run (or use keyboard shortcuts) to generate, summarize, expand, or translate text.
Grounded Academic Research via Zotero
Ask questions directly against your reference library using ZotSeek. Wordbot performs semantic searches, extracts key findings, and generates grounded text with matching inline citations that adhere strictly to your active Zotero citation style (APA, IEEE, Harvard, etc.). The official Zotero plugin handles bibliography generation seamlessly.
Which Edition to Pick
Demo
https://github.com/user-attachments/assets/9b2e8f6c-2df8-4a3e-9b7f-a32f5d0b8e3d
3.5x speed due to file size restriction. Every time I press Run, I ask a local AI model with no Zotero library context, so it gives generic responses. For the last query, I do semantic search on Zotero library and it brings up context with inline citations.
How It Works
The system bridges Word (VBA macros) with a Python server that communicates with your local LLM and ZotSeek:
The VBA macros handle the UI, text extraction, and final formatting. The Python server manages API calls and response processing. ZotSeek performs semantic searches against your local Zotero library to ground responses with your own papers.
Getting Started
Why It Matters for Researchers
Smaller local models running on a laptop often lack domain knowledge and produce generic responses or hallucinations. By hooking the backend into my local Zotero library via ZotSeek, Wordbot performs semantic searches across my personal paper collection, extracts grounding facts, and writes context-aware text complete with Zotero-aligned inline citations. Because these match Zotero's exact citation structure, the native Zotero Word plugin treats them as its own, allowing me to manually insert additional citations and generate a unified bibliography effortlessly.
Documentation & Support
License
CC BY-NC 4.0 — Free for non-commercial use with attribution.
A huge thank you to the ZotSeek author @introfini for building ZotSeek with its local API, which made the Research edition possible. Also, thanks to him for testing my plugin on macOS. This is the first independent project built on ZotSeek, and I'm excited to see what others will create.
I'd love to hear your feedback, questions, or suggestions. Try it out and let me know what you think!
All reactions