Context Graph vs. Context Database
Does explicit graph structure improve retrieval traceability and multi-hop reasoning over a context-database approach, at the same token cost?
Hypothesis
Semantica's graph-native context wins on retrieval traceability and multi-hop correctness; OpenViking's context database wins on raw retrieval latency and cost — neither dominates on every axis.
Background
The two top-scoring repos out of the Sept 9 GitHub Open Source Radar run (98/100 and 97/100) are architecturally opposite answers to the same agent-memory problem.
Method
Same source corpus, same task set, same underlying model; build ingestion adapters for both Semantica and OpenViking; log every retrieval trace.
Measurement Plan
Retrieval traceability, multi-hop correctness, latency, and token cost.
Expected Artifact
A benchmark notebook plus architecture diagrams comparing both systems on identical tasks.
Limitations
Both repos are queued to study, not yet run locally — this experiment depends on that first pass completing.
Status
This experiment is Proposed — queued, not started. Results, data, and measurements will appear here once it actually runs.
