ARCAContact
← All guides

Libraries

Build document knowledge bases that power smarter AI research and responses.

Overview

Libraries are curated collections of documents that serve as knowledge bases for Arca's AI. When you attach a library to a research chat, the AI can search through every document in it—finding relevant clauses, referencing precedent, and citing specific passages. The more relevant your libraries, the better the AI's answers.

Creating a library

  1. Click New library.
  2. Give your library a descriptive name.
  3. Choose visibility:
    • Private — Only you can see and use it.
    • Organization — Everyone on your team can access it.
  4. Upload documents (see below).

Adding documents

You can add documents to a library from several sources:

  • Local upload — Drag and drop files or use the file picker. Supports PDFs, Word documents, text files, and images.
  • Google Drive — If connected, browse and select files from your Drive.
  • SharePoint — If connected, browse and import documents from your organization's SharePoint.
  • Ironclad — If connected, import contracts directly from Ironclad.

You can upload multiple files at once. Each file shows its upload progress individually.

How documents are processed

When you add a document, Arca processes it through several steps:

  1. Text extraction — For PDFs and scanned documents, Arca uses OCR to extract text while preserving document structure (headers, tables, paragraphs).
  2. Metadata generation — The AI generates a summary, extracts the title, identifies key entities (companies, people, laws), assigns topic tags, and classifies the document type.
  3. Chunking — The document is split into smaller, overlapping sections that respect the original structure (e.g., not cutting in the middle of a paragraph).
  4. Embedding — Each chunk is converted into a vector embedding and stored in a search index, making it findable by semantic meaning rather than just keywords.

This process happens in the background. You can check the status of each document—a green indicator means it's fully processed and searchable.

Managing documents

  • Search — Find documents within a library by name or content.
  • Preview — View a document's contents without downloading it.
  • Download — Get the original file.
  • Delete — Remove individual documents from the library.
  • Embedding status — Each document shows whether it has been successfully embedded. If embedding fails, an error message and retry option are available.

Using libraries in research

Libraries become powerful when attached to research chats:

  • Select one or more libraries when starting a chat. The AI will search these libraries to find relevant information for your questions.
  • Libraries can be linked to playbooks, so they're automatically attached whenever that playbook is used.
  • The same library can be used across multiple chats and playbooks.
  • Organization-visible libraries are available to all team members.

Library settings

  • Rename — Change the library name at any time.
  • Change visibility — Switch between private and organization-wide access.
  • Delete library — Permanently remove the library and all its documents.

Tips for organizing libraries

  • Group documents by topic or use case—e.g., “Vendor Agreements,” “Employment Templates,” “Regulatory Guidance.”
  • Keep libraries focused. A library with 20 relevant documents will produce better results than one with 200 loosely related files.
  • Use organization visibility for shared reference materials and private visibility for work-in-progress collections.
  • Link libraries to playbooks to create ready-made research configurations that your whole team can use.