About KnowledgeMapper
KnowledgeMapper is a visual research workspace for building, testing, and documenting topic networks. It combines map editing, evidence collection, and assistant-guided expansion in one interface.
The core goal is traceability: what each topic means, why links exist, and which sources support node-level and edge-level claims.
How It Works
- Create a named map, or open the Combined map to inspect cross-map overlap.
- Add topics manually with +, or use AI Assistant (General mode) to build a new map from a prompt.
- Use Extend current map with a prompt to steer what gets added next.
- Open nodes and edges to save notes, attach evidence, and inspect relationship reasoning.
- Use grounded assistant scope (map/node/edge) for evidence-aware answers and cited papers.
- Tune Similarity, Cluster, Link pull, layout mode, and TDA recommendation to control structure.
- Share a map read-only, or export Download .txt for node/edge notes + BibTeX-style entries.
Research Workspace
- Multiple named maps
- Automatic Combined map across all user maps (deduplicated topics)
- Node expansion and bridge-topic generation
- Node evidence + edge evidence trails
- Persistent notes on both nodes and edges
AI Assistant
- Grounded mode for map-aware answers with citations
- General mode for broader ideation with optional map/node/edge focus
- Build map from prompt
- Prompt-driven Extend current map
- Save assistant output into notes and evidence records
Graph Controls
- Similarity threshold to gate weak links
- Cluster threshold for color-group structure
- Link pull to tune connected-node compactness
- UMAP or classic layout mode
- TDA recommendation baseline for layout tuning
- Fast settle mode and manual cluster positioning
Sharing
- Private maps by default
- Public read-only share links
- Shared links carry current layout settings
- Recipients can view notes and saved papers
- Recipients can open learning links and load evidence
- Recipients cannot save, edit, or modify map data
Map Management
- Per-map saved layout settings
- Delete map with Bin action
- Combined map remains read-only by design
- Fullscreen graph mode with centered detail overlays
Export
- Per-map Download .txt export in the dashboard header
- Includes every node with BibTeX-style entries + node notes
- Includes every edge with BibTeX-style entries + edge notes
- Structured for easy copy into LaTeX/BibTeX writing pipelines
Node vs Edge Evidence
Node evidence answers: What supports this topic?
Edge evidence answers: What supports this relationship between two topics?
Using both layers turns a visual map into a defensible research artifact.
Best Practices
- Keep topic names specific and testable.
- Use separate maps for unrelated domains rather than forcing one giant graph.
- Treat node evidence as support for a concept, and edge evidence as support for a claim.
- Use extension prompts to intentionally steer what the map grows toward.
- Use edge notes to record assumptions, caveats, and confidence level.
- Start with TDA recommendations, then tune manually for your use case.
- Use grounded assistant mode when you need traceable, source-linked responses.
- Validate all suggested papers before relying on them in real outputs.
Scope and Limits
KnowledgeMapper is strongest for early- to mid-stage research planning, literature mapping, and hypothesis exploration across related concepts.
It does not replace formal literature review, domain expertise, citation-management tooling, or source-quality verification. Assistant outputs should be treated as drafts to validate, not final authority.
Data and Validation
Your maps, topics, notes, and saved evidence are tied to your account. Shared links are read-only, map-specific, and can be disabled or regenerated. Exported `.txt` files are generated client-side from your current map data.
Evidence suggestions are a starting point. Verify relevance and quality before citing or acting on them.