alharkan7 / beeblio

# Beeblio
Beeblio is a full-stack AI research workspace. It collaborates with you and works directly in your folders & documents. The Next.js interface and Eve agent run locally on your computer, and each project points to an existing folder.
This is intended for a single trusted user on one computer.
**Download the desktop app:** Get Beeblio for macOS (Apple Silicon) or Windows. Open the release marked **Latest** and choose the `.dmg` or `.exe` under **Assets**. Installers appear there after a release is published.
## What you can do
Beeblio brings writing, an AI research assistant, and your project files into one workspace:
- **Write with your sources beside you.** The document editor supports rich Markdown editing, tables, math, images, inline citations, and a generated bibliography. Search for literature or insert a citation from the editor, and use document review and sentence suggestions while you write. Export finished work as DOCX, citation-aware DOCX, LaTeX, or portable Markdown.
- **Work with an agent in the same project.** The chat can use the open file, `@`-mentioned files, uploaded files, and selected passages as context. The Eve agent can search literature and the web, inspect and edit project files, update bibliographies and literature matrices, and run local Bash or Python workflows. Conversations are saved per project so you can return to earlier work.
- **Keep the research organized.** The workspace rail combines a searchable file explorer with literature search and library. Browse references, data, analysis, reports, and figures in the artifact views; upload files, create folders and research artifacts, and open them alongside the chat.
The project session page connects the workspace to a new or saved conversation. Your project remains an ordinary folder on disk, so files you create or edit in Beeblio are available to your other tools.
## Requirements
- Node.js 24 and pnpm 11
- An OpenRouter API key and model ID for the agent
- Bash and Python 3 for local agent commands and analysis
- Optional: LibreOffice (`soffice`) for agent-assisted Office-to-PDF conversion; other command-line tools and Python packages for the workflows you want to run
In the browser, the folder picker uses macOS's native chooser; on other systems, enter an existing absolute folder path in the project form. The desktop app has a native folder picker on every platform.
## Run locally
 [code] 
Open http://127.0.0.1:3000. The root URL redirects to `/workspace`. `pnpm dev` creates `.beeblio/` if needed, applies SQLite migrations, and starts both the Next.js UI and the Eve agent. They listen on `127.0.0.1:3000` and `127.0.0.1:2000` respectively.
Choose **Link Project Folder** to select a folder. You can also paste its absolute path. Beeblio stores the resolved path in SQLite and works with the files in place. On linking, it adds any missing canonical files & folders: `1-References/`, `2-Data/`, `3-Analysis/`, and `4-Reports/` folders, plus `1-References/referenc