PaperForge

简体中文 · English

铸知识为器，启洞见之明。 — Forge Knowledge, Empower Insight.

PaperForge brings your Zotero library into Obsidian. Sync papers, run OCR, extract figures, and do AI-assisted deep reading — all inside a single vault.

---

0. What PaperForge Is

PaperForge is not just an Obsidian plugin. It has two parts:

Part	What	Does	Where
Obsidian Plugin	main.js + manifest.json + styles.css	Dashboard, buttons, settings UI	.obsidian/plugins/paperforge/ in your vault
Python Package	paperforge	Sync, OCR, Doctor, repair	Your system Python (pip install)

The plugin is the interface. The Python package is the engine. Every button you click in the plugin actually runs a Python command behind the scenes.

After installing the plugin, you MUST verify that the Python package is also installed and version-matched.

---

1. Install the Obsidian Plugin

Option A: Community Plugin Browser (Recommended)

1. Open Obsidian → Settings → Community plugins → Browse
2. Search for PaperForge
3. Click Install, then Enable

Community plugins auto-update through Obsidian. No extra steps needed.

Option B: BRAT

If you need beta versions or the plugin hasn't appeared in search yet:

1. Install BRAT from the Obsidian community plugin browser
2. Open BRAT settings → Add Beta Plugin
3. Enter: https://github.com/LLLin000/PaperForge
4. Enable PaperForge in Settings → Community Plugins

Option C: Manual Download

1. Go to Releases
2. Download the three files: main.js, manifest.json, styles.css
3. Create .obsidian/plugins/paperforge/ in your vault
4. Put the three files there
5. Restart Obsidian → Settings → Community Plugins → enable PaperForge

Manual install does not auto-update. You'll need to re-download for each new version.

---

2. Install the Python Package

After enabling the plugin, open the PaperForge settings tab. You'll see a Runtime Status section:

Plugin v1.5.0 → Python Package v1.5.0 ✓ Matched

• If it says "Not installed" → click Open Wizard to re-run the setup process
• If it says "Mismatch" → the Python package auto-updates when the plugin updates. If it didn't succeed, click Update Runtime to manually trigger

---

3. How Python Interpreter Resolution Works

PaperForge needs to find a working Python on your system. It searches in this order:

Priority	Source	Description
1	Manual override	Settings → Custom Python Path, enter the full path (e.g., C:\Users\you\...\python.exe). This is the most reliable method.
2	venv auto-detect	Scans .paperforge-test-venv, .venv, venv under your vault root
3	System auto-detect	Tries py -3, python, python3 in order, verifies with --version
4	Fallback	Defaults to python if nothing else works

If you have multiple Python installations (e.g., system 3.9 + self-installed 3.11), strongly recommend setting a manual path in settings to avoid hitting the wrong one.

The Validate button in settings immediately tests the resolved interpreter and shows its version.

---

4. Setup Wizard — What Each Step Means

Open the plugin settings panel (Settings → Community plugins → PaperForge) and click the Open Wizard button. The wizard walks you through configuration. Here's what every step does.

4.1 Vault Path

Your Obsidian vault root. Auto-detected, usually no need to change.

4.2 AI Agent Platform

PaperForge's deep reading features run through an AI Agent. The core mechanism is trigger phrases, not registered plugins: you type /pf-deep <key> directly into the Agent chat, and the Agent recognizes the trigger and loads the literature-qa Skill automatically.

The setup wizard deploys Skill files to the correct location:

Agent	Skill location	Trigger example
OpenCode	.opencode/skills/ + .opencode/command/	/pf-deep <key>
Claude Code	.claude/skills/	/pf-deep <key>
Cursor	.cursor/skills/	/pf-deep <key>
GitHub Copilot	.github/skills/	/pf-deep <key>
Windsurf	.windsurf/skills/	/pf-deep <key>
Codex	.codex/skills/	$pf-deep <key>
Cline	.clinerules/	/pf-deep <key>

Key concept: /pf-deep is NOT a plugin you install on the Agent platform — it's a Skill file deployed inside your Vault. Once the setup wizard copies the files into place, the Agent auto-discovers the triggers on startup. You type the trigger phrase just like any other chat input.

4.3 Directory Names

The wizard asks what to name several directories. These are for organizing files inside your vault. Defaults work for most users.

Parameter	Default	Purpose
system_dir	System	Root for PaperForge internal data. Contains exports/ (Zotero JSON exports), ocr/ (OCR results), config/. You rarely need to open this manually.
resources_dir	Resources	Resources root. Your formal literature notes live under this directory, inside literature_dir.
literature_dir	Literature	Formal literature notes directory. paperforge sync generates frontmatter .md notes here.
base_dir	Bases	Obsidian Base view definitions. Dashboard filters ("Pending OCR", "Ready to Read", etc.) are stored here.

4.4 PaddleOCR API Token

OCR requires a PaddleOCR API key. Configured in .env:

PADDLEOCR_API_TOKEN=your-api-key

The wizard guides you through setting this. You can also edit .env later. The OCR URL usually stays at the default.

4.5 Zotero Data Directory

PaperForge creates a junction (Windows) or symlink (macOS/Linux) linking your Zotero data directory into the vault. This is how Obsidian wikilinks resolve to PDF files.

The wizard auto-detects your Zotero installation. If detection fails, manually enter the path to your Zotero data directory — the folder that contains the storage/ subdirectory (not the Zotero executable).

4.6 What Happens During Setup

After confirming your choices, the wizard automatically:

• Creates all needed directory structures
• Deploys Agent command files to the correct locations
• Installs Obsidian plugin files
• Creates the Zotero junction/symlink
• Writes paperforge.json and .env

The process is incremental — if files already exist in the chosen directories, the wizard only adds what's missing and never deletes existing content.

---

5. First-Time Setup Checklist

1. Version match: Settings → Runtime Status → confirm plugin and Python package match
2. Python path: Settings → Validate button → confirm it's the Python you want
3. Setup wizard: Settings → PaperForge → Open Wizard
4. PaddleOCR key: Enter your API token in .env (wizard guides this)
5. Export from Zotero: Right-click your library → Export... → format Better BibTeX JSON → check Keep updated → save to <system_dir>/PaperForge/exports/
6. Run Doctor: Dashboard → Run Doctor → all checks should pass

---

6. Daily Use

Dashboard (Three-Mode Views)

Ctrl+P → PaperForge: Open Dashboard opens the control panel with three views:

View	Purpose
Global	System homepage: run Sync, OCR, Doctor, and other mechanical operations
Collection	Batch workspace: browse paper queues by domain, batch tagging
Per-paper	Reading companion: do_ocr / analyze toggle checkboxes, discussion record cards

PDF files in the Dashboard automatically switch to Per-paper mode — no manual switching needed.

AI Deep Reading & Q&A (Requires Agent)

Launch your Agent app and type commands into its chat input. The more specific you are about the paper (Zotero Key, title, DOI), the faster the Agent locates it.

Route	Command	Does	Trigger examples	Prerequisites
Deep Read	/pf-deep <key>	Keshav three-pass deep reading, writes to formal note	deep read XX, walk me through, journal club	OCR done, analyze: true
Q&A	/pf-paper <key>	Interactive paper Q&A, OCR not required	take a look at XX, what does this paper say	Formal note exists
Archive	/pf-end	Save current /pf-paper Q&A session	save, end discussion	During /pf-paper session

/pf-end Details

• /pf-end only applies to /pf-paper Q&A sessions. Deep reading (/pf-deep) writes directly to the formal note and does not need /pf-end.
• When executed, two files are created in the paper's workspace:
  • discussion.md — human-readable Q&A discussion record
  • discussion.json — structured Q&A data (with timestamps, source tags)
• Dashboard Per-paper view automatically displays these as discussion record cards

Command prefixes vary by platform (mostly /, Codex uses $).

---

7. Full Workflow

Add paper to Zotero
  ↓ Better BibTeX auto-exports JSON to exports/
Dashboard → Sync Library
  ↓ Generates formal note (in Literature/, with frontmatter metadata)
Set do_ocr: true in the note's frontmatter
  ↓
Dashboard → Run OCR
  ↓ PaddleOCR extracts full text + figures → ocr/ directory
Set analyze: true in the note's frontmatter
  ↓
Open Agent → type /pf-deep <zotero_key>
  ↓ Agent performs three-pass deep reading
## 🔍 Deep Reading section appears in the note
  ↓ (for additional Q&A)
Open Agent → type /pf-paper <zotero_key>
  ↓ Interactive Q&A
Type /pf-end to save the discussion record
  ↓
Dashboard Per-paper view shows discussion cards

---

8. Troubleshooting

Plugin fails to load

• Confirm .obsidian/plugins/paperforge/ has main.js, manifest.json, styles.css
• If upgrading from an old version: delete the entire paperforge plugin folder and reinstall via the community plugin browser
• Open Developer Console (Ctrl+Shift+I) and check the red errors

"Sync Runtime" doesn't update the version

• The plugin may be calling a different Python than your terminal. Check Settings → Python path
• Try with --no-cache-dir to bypass pip cache
• Confirm https://github.com/LLLin000/PaperForge is reachable

OCR stays pending

• Confirm .env has PADDLEOCR_API_TOKEN
• Run paperforge ocr --diagnose to check API connectivity
• PDF paths may be broken: run paperforge repair --fix-paths

No notes generated after sync

• Is Better BibTeX auto-export configured in Zotero? Are JSON files in exports/?
• Run paperforge doctor to find which step failed

/pf-deep command does nothing

• Make sure you're running it in your Agent app, not a terminal
• Confirm OCR is done (ocr_status: done)
• Confirm analyze is set to true

---

9. Updating

The Obsidian plugin auto-updates through the community plugin browser. For the Python package:

paperforge update
# or
pip install --upgrade paperforge

If you installed via BRAT, it also auto-detects GitHub Release updates.

---

10. Architecture

paperforge/
├── core/          Contract layer — PFResult/ErrorCode/state machine
├── adapters/      Adapter layer — BBT parsing, paths, frontmatter I/O
├── services/      Service layer — SyncService orchestration
├── worker/        Worker layer — OCR, status, repair
├── commands/      CLI dispatch
├── setup/         Setup wizard (directories, agent deployment, Zotero linking)
├── plugin/        Obsidian plugin (Dashboard, settings panel)
└── schema/        Field registry

---

License

CC BY-NC-SA 4.0. Non-commercial use only.

Acknowledgments

Built on PaddleOCR, Obsidian, Better BibTeX for Zotero, and other great open-source projects.