Private · local-first knowledge systems · Graph tooling
Investigation Workbench
A local-first workspace that turns fragmented source material into reviewed, queryable knowledge through explicit stages of evidence, suggestion, review and projection.
- Niche
- Evidence-heavy knowledge review and graph exploration
- Core stack
- Memgraph · Python · Obsidian · Docker · MCP
- Model
- Raw source → candidate → reviewed canonical knowledge → compiled graph
- Focus
- Provenance, reviewability, temporal structure, reproducible projection
The study
Graph and AI tooling benefits from an explicit separation between extraction, interpretation, review and canonical knowledge.
Automated systems propose entities, events and relationships; review establishes canonical knowledge and its provenance.
Raw documents enter a controlled ingestion layer. Candidate knowledge is reviewed before promotion, and reviewed canonical notes are then compiled into graph, timeline and viewer artifacts.
System structure
Review state is part of the architecture and determines which material can establish graph authority.
Ingestion proposes entities, events and relationships; reviewed canonical material establishes downstream graph truth.
The workspace runs locally with explicit storage, container and resource boundaries, keeping source material under local control.
Graph nodes, edges, timelines, manifests and viewer artifacts are compiled from the reviewed layer so outputs can be rebuilt and checked.
Obsidian, Memgraph Lab, generated timeline and graph viewers, and MCP-connected agents consume projections of the same reviewed knowledge under one authority model.
Product and interaction design
The central product problem is epistemic: make source evidence, machine suggestion, reviewed knowledge and derived projections visibly distinct.
That distinction shapes navigation, file structure, validation and graph semantics. The interface presents uncertainty and provenance directly, giving reviewers a clear account of how each conclusion entered the system.
Agents and MCP tools act as query and transformation interfaces within explicit schemas, review gates and deterministic compilation steps.
Development approach
The repository is organized as an inspectable pipeline with distinct stages and authority boundaries.
Docker isolates Memgraph and its tooling; Python dependencies are managed reproducibly; the knowledge vault carries schemas and controlled vocabularies; validation covers links, temporal semantics and orphaned material; fixture cases exercise chronology and concurrency before real data is introduced.
MCP entry points let different AI clients query the same local graph under one local data authority.
Private repository · public case study focuses on architecture; source material remains private