Every fact is a claim with provenance and a validity window. Overwrites and contradictions are first-class graph structure. A graph-probe router answers cheaply, escalates conflicts, and abstains when the graph can't back an answer.
When deadlines move, owners hand off, and status changes across Slack, Linear, and meetings — flat chunk stores retrieve stale text. They can't tell you what's current, what was believed last week, or when to say "I don't know."
New facts create SUPERSEDES edges — history is never overwritten. Reconstruct any fact's full timeline with bounded path traversal.
When Slack says one thing and Linear says another, CONTRADICTS edges mark the disagreement. Trust scoring arbitrates.
If no claim exists for the asked (subject, predicate), the system declines and reports what it searched. No guessing.
"What was believed as of last Tuesday?" is a property filter on valid_from/valid_to — a Cypher query, not inference.
2–3 bounded Cypher queries measure coverage, conflicts, and chain depth. Clean → FAST. Messy → DEEP. Empty → ABSTAIN.
Every answer traces back to verbatim quotes, authors, and sources. Provenance is structure, not prompt engineering.
Slack, Linear, meeting transcripts → LLM extractor pulls atomic claims with a closed predicate vocabulary (~12 types).
Deterministic rules detect overwrites (SUPERSEDES) and conflicts (CONTRADICTS). Deduplication. No ambiguity.
Claims, evidence, entities, and typed edges land in HydraDB via batched UNWIND writes. Idempotent.
Classify → graph probe → route FAST / DEEP / ABSTAIN → cited answer with timeline and provenance.
| Arm | Accuracy | Knowledge update | Abstention P / R | Queries/q | p95 latency |
|---|---|---|---|---|---|
| Naïve RAG | 24.0% | 100% | 0% / 0% | 0.9 | 6 ms |
| Question Router | 76.0% | 100% | 100% / 50% | 4.8 | 111 ms |
| Always Deep | 84.0% | 100% | 100% / 50% | 5.0 | 84 ms |
| Router + Graph Probe | 100% | 100% | 100% / 100% | 4.7 | 71 ms |
The naïve RAG baseline picks the single active claim with the most word overlap. It cannot see supersession chains, cannot surface conflicts, and guesses on every abstention question. The graph probe gives HydraClaim precise, typed coverage.
Connected to a live HydraDB instance with 8 pre-ingested scenarios. Ask about deadlines, owners, conflicts, or things that were never recorded.