paper-cli（paper CLI 搜索）

paper CLI — Academic Paper Management

paper is a local-first CLI tool for managing academic papers with AI-powered semantic search. It stores data in SQLite databases and uses FAISS vector stores for similarity search.

Core Concepts

Scope: User vs Project

Every resource (config, knowledge base, literature) lives in one of two scopes:

• Project scope (default): stored in ./.paper-manager/ — tied to the current directory, good for project-specific paper collections
• User scope (--user flag): stored in ~/.paper-manager/ — shared across all projects, good for a personal paper library

Project-level config overrides user-level config. When looking up a knowledge base or literature, the tool checks project scope first, then falls back to user scope.

Data Hierarchy

Config (embedding models)
  └── Knowledge Base (has a name, description, and embedding model)
        └── Literature (a paper with metadata, source file, and vector embeddings)
              └── Notes (key-value pairs for personal annotations)

You must configure an embedding model before creating a knowledge base, and you must create a knowledge base before adding literature.

Troubleshooting: Installation & Setup

The user should already have paper installed and configured. Only use this section if something is missing.

Installation (if paper command is not found)

npm install -g paper-manager    # or: pnpm add -g paper-manager

Requires Node.js >= 24. Verify with paper --version.

Config setup (if commands fail due to missing config)

# If data directory is missing or incomplete, initialize it first
paper config init [--user]

# Set up an embedding model (OpenAI-compatible provider)
paper config set embeddingModels '{"my-model": {"provider": "openai", "model": "text-embedding-3-small", "apiKey": "sk-...", "dimensions": 1536}}' --user

# Set a default model so you don't have to specify it every time
paper config set defaultEmbeddingModelId my-model --user

The embedding model config fields:

• provider: currently only "openai" (works with any OpenAI-compatible API)
• model: the model name (e.g., "text-embedding-3-small")
• baseUrl: optional custom API endpoint
• apiKey: API key for the provider
• dimensions: embedding vector dimensions (e.g., 1536)
• batchSize: optional max number of texts per embedding API request (set this if your provider limits batch size)

Command Reference

paper config — Configuration Management

paper config init [--user]              # Initialize data directory (config, db, subdirs)
paper config list [--user]              # Show all config (merged or user-only)
paper config get <key> [--user]         # Get a specific config value
paper config set <key> <value> [--user] # Set a config value (value is parsed as JSON, falls back to string)
paper config remove <key> [--user]      # Remove a config key

Config keys:

• embeddingModels — a JSON object of {[modelId]: {provider, model, baseUrl?, apiKey, dimensions, batchSize?}}
• defaultEmbeddingModelId — which model ID to use when none is specified

paper kb — Knowledge Base Management

# Create a knowledge base (requires an embedding model in config)
paper kb create <name> -d <description> [-e <embedding-model-id>] [--user]

# List knowledge bases
paper kb list [--user | --all]

# Update knowledge base metadata (name and/or description)
paper kb update <id> [-n <name>] [-d <description>]

# Remove a knowledge base and ALL its data (literatures, PDFs, vectors)
paper kb remove <id>

# Semantic search across a knowledge base
paper kb query <id> <query-text> [-k <top-k>]   # default top-k is 5

The <id> for knowledge bases is a UUID assigned at creation time. Use paper kb list to find it.

paper lit — Literature Management

# Add a paper (extracts content, splits text, creates embeddings)
# Supports PDF, TXT, MD, TEX, and other text-based formats
# For PDFs, automatically extracts metadata (title, author, keywords, DOI, etc.)
paper lit add <kb-id> <file-path> [-t <title>]
# Title defaults to PDF metadata title, then filename if not specified

# List papers in a knowledge base
paper lit list <kb-id>

# Show full details of a paper
paper lit show <kb-id> <lit-id>

# Update paper metadata
paper lit update <kb-id> <lit-id> [options]
#   -t, --title <title>
#   --title-translation <translation>
#   -a, --author <author>
#   --abstract <abstract>
#   --summary <summary>
#   --url <url>
#   --doi <doi>
#   --keywords <comma-separated-keywords>

# Remove a paper (deletes DB record, source file; vectors remain in store)
paper lit remove <kb-id> <lit-id>

paper lit note — Literature Notes

Notes are key-value string pairs attached to a literature entry — useful for personal annotations.

paper lit note list <lit-id>                    # List all notes
paper lit note set <lit-id> <key> <value>       # Set a note
paper lit note remove <lit-id> <key>            # Remove a note

Note: the note commands take <lit-id> directly (not <kb-id> <lit-id>).

paper util — Utilities

# Convert a DOI to BibTeX citation (accepts DOI identifier or full URL)
paper util doi2bib <doi>

Common Workflows

Start a new research project

1. paper kb create "my-project" -d "Papers about X" — create a project-scoped KB
2. paper lit add <kb-id>./paper.pdf -t "Paper Title" — add papers (also works with.txt,.md,.tex)
3. paper kb query <kb-id> "your research question" — search

Manage paper metadata

1. paper lit list <kb-id> — find the literature ID
2. paper lit update <kb-id> <lit-id> -a "Author Name" --keywords "ML,NLP"
3. paper lit note set <lit-id> takeaway "Key insight from this paper"

Important Notes

• All IDs (knowledge base, literature) are UUIDs — always use list commands to look them up
• Adding a paper (lit add) extracts the file content (PDF or text), splits it into chunks, and creates vector embeddings — this calls the embedding API and may take some time
• Text files have a 10 MB size limit
• The kb remove command is destructive: it deletes the knowledge base, all its literatures, source files, and vector stores
• The --user flag on config/kb create controls scope; omitting it uses project scope
• Config values passed to config set are parsed as JSON first; if JSON parsing fails, the raw string is stored

Skill Maintenance

Skill version: v0.6.0. To update to the latest version, run npx/pnpx/bunx skills add paper-manager.