CNRS·Université Côte d'Azur·Institut de Chimie de Nice·Centre Inria de l'Université Côte d'Azur·Interdisciplinary Institute for Artificial Intelligence (3iA) Côte d'Azur
HolobiomicsLab

Research workflows

Put the literature to work

From exploring a new field to preparing a proposal, use Perspicacité to assemble a collection, question it, and return to the underlying papers.

Explore a field

Build a reading map before a long reading list.

For a researcher entering a topic or a student beginning a literature review.

Start with a topic search or a small set of seed papers. Build a knowledge base, expand it through citation links, and use Literature Survey to organize the retrieved work into themes. Follow promising groups with focused Basic or Advanced questions.

The useful result is a collection of papers and an initial map of themes to inspect. Review source coverage and inclusion choices yourself before treating the collection as a systematic review.

Literature Survey mode ↗

Compare findings

Locate the disagreement, then read it in context.

For questions where several papers seem to support different conclusions.

Bring the relevant papers into the same collection and frame the comparison as a focused question. Contradiction mode extracts claims and groups them into agreement, disagreement and open questions.

Use those groups to decide which papers and passages to read next. Check whether differences in methods, conditions or scope explain an apparent conflict; the grouping is an aid to analysis, not an adjudication of the science.

Compare the research modes →

A documented proposal workflow

Keep project context and scientific evidence distinct.

The grant-proposal guide describes a two-knowledge-base strategy. One collection holds context such as earlier proposals, reviewer reports and call documents; another holds the scientific literature, for example from Zotero. Retrieve from the context collection when addressing project history, and from the literature collection when looking for evidence to support a scientific claim.

Context for the argument

Give local documents descriptions that explain their relevance to the proposal. Keep an explicit list of what was ingested, and retrieve the relevant context when working on a section or responding to a previous review.

Literature for the claims

Search the scientific collection with Advanced mode, use deeper retrieval when needed, and inspect the cited passages. The documented workflow can archive annotations back to Zotero for later use.

This is a workflow described by the project, not a reported evaluation of proposal outcomes. Choose your inference provider with the content of your project documents in mind: retrieved context may be included in model requests.

Read the proposal workflow ↗

Maintain a useful collection

Let the evidence travel with your notes.

For recurring literature mining, bibliography analysis and research planning.

Import an existing bibliography or Zotero collection, add relevant papers by DOI, and follow citations as the topic develops. Keep the indexed collection available for new questions, and export papers and notes to the tools you already use.

BibTeX with cached PDFs, and Obsidian vault exports, provide routes out of the research workspace. An MCP-connected assistant can use the same collection as part of a broader workflow; Mimosa-AI is one of the clients named by the project.

Use the collection from an agent →