Multi-Doc Chat for Projects

Multi-Doc Chat lets you have a conversation with all the documents in your project simultaneously. Instead of opening each paper individually and asking questions one at a time, you can ask a single question and receive an answer that draws from every document attached to your project.

How It Works

Multi-Doc Chat is powered by AcaTrove's Retrieval-Augmented Generation (RAG) system. When you ask a question, the system:

1. Searches across all documents in your project using semantic similarity to find the most relevant passages.
2. Retrieves the top-matching chunks of text from multiple documents.
3. Generates a response that synthesizes information from those passages.
4. Cites the specific documents and passages used, so you can verify every claim.

This means you are not limited to keyword matching. The AI understands the meaning of your question and finds relevant content even when different papers use different terminology for the same concept.

Starting a Multi-Doc Chat

1. Open your project.
2. Open the Research AI Panel from the sidebar.
3. Select Multi-Doc Chat from the panel's feature menu.
4. The chat automatically scopes to the current project's documents. You will see a message confirming how many documents are included.
5. Type your question and press Send.

Multi-Doc Chat interface within a project

Understanding Citations

Every response from Multi-Doc Chat includes inline citations that reference the source documents and passages. Citations appear as clickable markers within the response text. Clicking a citation:

• Shows the exact passage the AI used.
• Identifies the source document by title and author.
• Provides a link to open the full document in Paper Chat for deeper exploration.

This citation transparency is essential for academic work. You can verify that the AI's synthesis accurately represents what each source says, rather than taking the response at face value.

Response with inline citations linking to source documents

What You Can Ask

Multi-Doc Chat is effective for questions that span multiple sources:

• Synthesis questions -- "What do my sources say about the relationship between sleep duration and cognitive performance?"
• Comparison questions -- "How do the methodologies in my uploaded studies differ?"
• Gap identification -- "Are there any contradictions between the findings in these papers?"
• Fact lookup -- "Which of my papers reports a sample size greater than 500?"
• Summary requests -- "Summarize the key findings across all my project documents."

The quality of responses depends on the documents in your project. The more relevant and well-structured your uploaded documents are, the better the AI's answers will be.

Managing Documents for Chat

Multi-Doc Chat includes all documents currently linked to your project. To improve chat quality:

• Add relevant documents -- Upload or link papers, data reports, and drafts that are central to your research question.
• Remove irrelevant documents -- Unlink documents that are not related to the project's focus. Fewer, more relevant documents produce more precise responses.
• Upload full text -- The AI works best with full-text documents. Abstracts-only uploads produce limited results.

Tips

• Start with broad synthesis questions to understand how your sources relate, then drill down into specifics.
• Use Multi-Doc Chat before writing a literature review to quickly identify themes and gaps across your sources.
• Always verify citations by clicking through to the source passage. AI-generated synthesis should be treated as a research aid, not a final product.