The task
How can an analyst discover relationships in a changing corpus and distinguish new knowledge, real-world changes, corrections, and withdrawals?
My solution
I built a local evidence workspace that combines search, graph tracing, versioned sources, saved investigations, and comparisons between knowledge snapshots.
Search selects evidence within a pinned snapshot. Graph relationships retain their source passages and versions. Baseline comparisons distinguish when a claim was known from when an event was said to occur, while local model adapters support extraction and embeddings.
How it works
Python · SQLite · lexical and vector retrieval · llama.cpp- 01
Pin
Choose a knowledge snapshot and an evidence scope for the investigation.
- 02
Trace
Search passages and follow cited relationships back to source revisions.
- 03
Compare
Compare with a saved baseline and inspect corrections and withdrawals separately.
Try it
Interactive browser demonstrationInteractive browser demo
Change the date. Keep the evidence.
Set when information was known and when the claim applies. Watch a plan, a late correction, conflicting reports, and a withdrawal change the answer.
Loading the bundled evidence…
Methods / evidence / reproduction
Read the research.
An authored five-batch Pump P fixture exercises versioned claims, corrections, and withdrawal behavior. Separate local-model records document actual extraction and embedding calls.
The paper connects the research question to the implementation, worked examples, and evidence. The project directory includes the runnable software, documentation, and tests.