What is archAIc

archAIc is a probabilistic search and intelligence layer for handwritten archival documents.

The product takes scanned manuscripts, produces a multi-hypothesis transcription rather than a single best guess (this is the Probabilistic Indexing, or PrIx, foundation), and exposes the result as a probabilistically searchable index plus a graph of structured entities, events, and relationships anchored to specific lines on specific pages.

The pieces:

• PrIx generation. HTR over degraded handwriting plus a probabilistic indexing pipeline that preserves alternative readings with their confidence scores, instead of collapsing to a single best guess.
• The search engine. Probabilistic keyword search, semantic search, and a structured-document layer (Document Logical Units) all built on top of the same PrIx data.
• The interface surfaces. A web application, an MCP server for AI agents, and IIIF endpoints for integration with existing archival infrastructure.

archAIc is sold and run self-hosted today, packaged as a Docker bundle for the client’s own servers.

This is a placeholder concept page; more detailed concept documentation lands as the site grows.