Mantis Docs

Appearance

Welcome to Mantis! ​

Overview & Orientation ​

What is Mantis? ​

Mantis is an agentic, programmable workbench for exploring, creating, and transforming conceptual knowledge spaces across multiple data modalities (text, images, audio, networks, etc.).
Unlike passive data viewers or black-box chatbots, Mantis provides a glass-box interface for hands-on, visual, and semantic reasoning—with AI agents as copilots in your workflow.

  • Think: Cognitive map + AI agent cockpit + visual programming notebook + marketplace of data tools
  • Built for: Scientists, analysts, developers, teams, and anyone who needs to see and shape complex data

A visual of Mantis' teams, applications (verticals), and core functionality.

Core Concepts ​

1. Cognitive Cartography ​

  • Visual maps of knowledge, where each point represents a piece of data (document, image, protein, etc.)
  • Proximity = conceptual similarity
  • Clusters, contours, and trajectories show semantic relationships and patterns. Interactive navigation lets you zoom, filter, and annotate—just like a living map.

2. Multi-Modal Embeddings ​

  • Unify text, images, molecules, networks, and more in shared spaces
  • Joint and cross-modal representations—see connections across data types
  • Examples: clustering podcasts by theme, mapping molecules by chemical similarity, linking papers and grants on different “floors”

3. Agentic Workflows ​

  • Embedded AI agents (narrators, searchers, summarizers) assist, automate, and learn
  • Voice, chat, or code: command Mantis by speaking, typing, or scripting
  • Save and replay macros; use agents to cluster, summarize, plot, or generate reports

4. Outbeddings ​

  • Generate new artifacts from positions in latent space (e.g., summaries, images, molecules)
  • Move beyond “finding” to creating—powered by AI + user input

1. Map

  • The main canvas: explore points, clusters, and relationships visually
  • Zoom in/out, select with lasso, view overlays (gradients, cluster contours)

2. Tree

  • Hierarchical explorer: browse and filter content by semantic categories or ontology branches
  • Expand/collapse to focus on what matters

3. Reader

  • Read documents, summaries, or item details in context
  • Semantic highlighting, annotation, and multi-item comparison

4. Bags

  • Custom collections: save sets of points (“bags”) for further analysis or export
  • Share, compare, or summarize bags

5. Summarizer

  • Automatic, agentic summarization of selected points, clusters, or bags
  • Control summary length, style, and modality

6. Plots

  • Visualize quantitative features: histograms, scatter plots, density maps, overlays
  • Interactive customization and direct integration with map selections

7. Notebook

  • Embedded, programmable workspace (like Jupyter) linked to map elements
  • Script, annotate, run macros, and connect code outputs to visual UI

Interactive Features ​

  • Voice/Command Palette: Control the platform with voice or quick-commands: “Summarize this cluster”, “Search: climate change”, “Plot bag A vs bag B”
  • Action Log: See all your actions, agent responses, and replay/undo workflows
  • Multi-Floor Mapping: Jump between 2D/3D floors for different modalities or abstraction levels
  • Mini-Map: Always stay oriented—see where you are in the global landscape
  • Agent Panel: Manage, deploy, and create custom agents for automation or assistance