Skip to content

ruvnet.

Python

docs(ruvnet): finalize 2026-09-19 evidence

rUva7375c617f70
diff --git a/docs/self-improvement/2026-09-19-evidence.json b/docs/self-improvement/2026-09-19-evidence.json
index a0da934..df73dd3 100644
--- a/docs/self-improvement/2026-09-19-evidence.json
+++ b/docs/self-improvement/2026-09-19-evidence.json
@@ -1,9 +1,9 @@
{
"date": "2026-09-19",
"baselineCommit": "015725be93fadb3484474cae218cdc51dba2372b",
- "candidateCommit": null,
+ "candidateCommit": "29fec9d63e1e713891c14042a12cf0c4a998327e",
"finalCommit": null,
- "decision": "accept-pending-publish",
+ "decision": "accept",
"candidate": "raw-relevance-threshold",
"version": "0.7.0",
"snapshotId": "sha256:5ac0ca2f58e510a3e449b04c39d02a8538899ef3e56637773beccaa68058db07",
@@ -94,6 +94,12 @@
"sources": ["https://github.com/ruvnet/ruflo/releases/tag/v3.42.4","https://arxiv.org/abs/2608.26846","https://arxiv.org/abs/2605.06132"],
"priorForecastsExpired": []
},
- "ci": null,
+ "ci": {
+ "runUrl": "https://github.com/ruvnet/ruvnet/actions/runs/35438772319",
+ "status": "success",
+ "node22": "success",
+ "node24": "success",
+ "durationSeconds": 16
+ },
"dashboard": "https://ruvnet-harness-evidence.ruv.chatgpt.site"
}
diff --git a/docs/self-improvement/2026-09-19.md b/docs/self-improvement/2026-09-19.md
index a48a460..8f5803a 100644
--- a/docs/self-improvement/2026-09-19.md
+++ b/docs/self-improvement/2026-09-19.md
@@ -2,7 +2,7 @@
## Decision
-Accepted an optional minimum raw-relevance gate for reviewed-change search as RuVnet entry point 0.7.0, pending publication and CI. CLI callers use `--min-raw-relevance`; MCP callers use `minRawRelevance`. The value is bounded from 0 to 1 and filters the already exposed exact-token overlap before limiting. Omission defaults to zero and preserves 0.6.0 behavior. It is not semantic similarity, ranking confidence, answer confidence or value of information. The companion remains offline, read only and unable to refresh or learn by itself.
+Accepted and published an optional minimum raw-relevance gate for reviewed-change search as RuVnet entry point 0.7.0. CLI callers use `--min-raw-relevance`; MCP callers use `minRawRelevance`. The value is bounded from 0 to 1 and filters the already exposed exact-token overlap before limiting. Omission defaults to zero and preserves 0.6.0 behavior. It is not semantic similarity, ranking confidence, answer confidence or value of information. The companion remains offline, read only and unable to refresh or learn by itself.
## Upstream review
@@ -16,6 +16,8 @@ Fifteen alternating pairs used 10,000 discovery calls and 20 cold CLI starts per
Pinned Ruflo 3.25.6 deep code and dependency scans, npm audit and a tracked-file credential-pattern check reported zero known findings. Its secret scanner covered five files and STRIDE four, so coverage is limited. The new numeric input is checked by both MCP schema and runtime validation for finite values from 0 to 1. Residual risks include lexical recall loss and hosts overinterpreting the score despite explicit labels. This is not a claim of complete security.
+Candidate commit [`29fec9d`](https://github.com/ruvnet/ruvnet/commit/29fec9d63e1e713891c14042a12cf0c4a998327e) passed both Node 22 and Node 24 jobs in [GitHub Actions run 35438772319](https://github.com/ruvnet/ruvnet/actions/runs/35438772319).
+
## RuVector, WASM and GNN decision
The corpus remains at 10 catalog projects plus 38 reviewed changes: 48 sanitized public records, below the frozen 100-record gate. The candidate adds no persistent storage. RuVector retrieval, WASM acceleration, GNN training and persistent learning remain unimplemented and unclaimed. The future evaluation remains exact versus vector-only, fusion, GNN reranking and shuffled-graph control, requiring at least a ten-point held-out task-success gain plus latency, cold-start, memory, artifact-size and reproducibility evidence.