Description
ArchiBot is an AI-based digital assistant for archival research. It enables users to query a documentary archive in natural language using large language models (LLMs), retrieving information from thousands of documents, reconstructing it accurately, and always citing the relevant sources.
It is designed to provide accurate answers based exclusively on the contents of the archive.
Features
- Interactive search among documents, letters, photographs, and iconographies
- Extraction of relevant information and composition of responses with references to the sources
- Summaries, analyses, comparisons and generation of texts based on the archive’s documents
- Data verification (metadata consistency and identification of gaps)
- Integrated navigation: immediate visualization of the cited documents
- Handling of simple and complex requests through natural conversation
Architecture and Technologies
ArchiBot uses a multi-agent architecture: multiple artificial intelligence models collaborate to interpret the question, consult the knowledge base, and compose the response. The system includes:
- Document ingestion and automatic organization into a structured knowledge base
- Orchestration software that assigns tasks to the different AI agents
- Research agents that retrieve information from the archive
- Composition agents that combine data and sources into a coherent response
The interface combines chat and document consultation, facilitating the verification of information.
Advantages
- Answers based on verified sources, without invented content (no AI hallucination state)
- Ability to quickly analyze complex archives with thousands of documents
- Greater accuracy compared to general-purpose chatbots
- Flexibility and adaptability to different archival structures
- Faster and more reliable consultation thanks to integrated document navigation
Use Cases
- Free search within the archive
- Comparison between documents or between authors, works, and periods
- Verification and control of metadata
- Production of thematic texts based on archive documents
- Support for researchers, archivists, and scholars in daily consultation