← Velune Compare
SAMPLE COMPARISON — generated by running the real,
unmodified Velune Adapter and Velune Compare against a small synthetic
Reference Run and Target Run, not a live customer run. There is no
self-service upload workflow live yet; this is the same Compare Report
produced once you have a Reference Run and Target Run prepared.
Compare Result
1 change stands out
Stream 004
Target activity disappears here
Persistent across 4 consecutive seconds
Compare narrowed 5 Streams across 9 seconds to one 4-second review window.
Full Run
00:0000:09
Review First
00:05 → 00:09
Explore evidence ↓
Compare found 7 review intervals
across 4 affected Streams.
The sustained change ranked above
3 shorter variations.
Level 2 · Evidence
Stream 004
Message count, whole Run
Reference
100
→
Target
50
-50 messages
Affected window
00:05 → 00:09
Level 2 · Also detected, ranked lower
- Stream 003 — Activity briefly disappears for one second, then returns to normal.second 1
- Stream 002 — Activity briefly disappears for one second, then returns to normal.second 3
- Stream 001 — A small message-count change, isolated to 1 second.second 2
Level 3 · Bucket-level detail
- Comparable range: 9 complete one-second Buckets (complete_common_one_second_buckets_only rule).
- Top finding canonical Stream identifier: stream.unmapped.004, Buckets 5–8.
- Top finding difference type: Reference active, Target empty.
- Original topic/source names are not included in the canonical bundle by design ("original_topic_name_included": false) -- Stream labels above are the neutral identifier Compare itself receives.
- Candidate-D (alternating empty/non-empty pattern) findings: 0 — a sustained phase-offset-shaped pattern is a structural classification, not a determination of cause.
Level 3 · How the ranking works
Comparison basis: Reference Run and Target Run were both processed by
the same Adapter bundle (one bucket contract), so this comparison is
exact by construction.
Findings are ordered by Compare's own deterministic ranking
rule: a complete Reference/Target activity mismatch always ranks above
a partial message-count difference; within the same kind of difference,
a Stream that differs more persistently across the comparable Buckets
ranks above one that differs only briefly. This is a fixed, reproducible
ordering rule — never a severity, fault, or root-cause judgment.
There is no AI decision stage in this ranking.
Level 3 · Technical provenance
- Bundle ID: vcib_sha256_f7611da446035a984525386a6ed9d32030d99212520e9e1d7593ca7cce586ef9
- Report schema: 0.2.0
- Adapter version: 0.2.0
- Mapping profile: generic.mcap.structural v0.1.0