Commits

Breadcrumbs

History for

onmain

User selector

All users

Datepicker

All time

Commit history

Commits on Jul 12, 2026

• fix(evals): remediate eval quality gaps to 192/201, advance baseline

  Show description for 331441c
  

  ChelseaKR

  and

  claude

  committed
  331441cView commit details

  Copy full SHA for 331441c

  View code at this pointBrowse repository at this point

Commits on Jul 11, 2026

• FIX-01: wire the provenance gate so published artifacts must match HEAD (#23)

  Show description for 484d1d0
  

  ChelseaKR

  and

  claude

  authored
  484d1d0View commit details

  Copy full SHA for 484d1d0

  View code at this pointBrowse repository at this point

Commits on Jun 17, 2026

• Consistency pass: system prompt v5 (five agencies) + variance-robust gate

  Show description for 746df4c
  

  ChelseaKR

  and

  claude

  committed
  746df4cView commit details

  Copy full SHA for 746df4c

  View code at this pointBrowse repository at this point

• Fifth agency (Humboldt Transit) + privacy-safe feedback (115/118)

  Show description for 27320af
  

  ChelseaKR

  and

  claude

  committed
  27320afView commit details

  Copy full SHA for 27320af

  View code at this pointBrowse repository at this point

• Formal two-turn eval sub-suite: conversation (103/109)

  Show description for f6ceea1
  

  ChelseaKR

  and

  claude

  committed
  f6ceea1View commit details

  Copy full SHA for f6ceea1

  View code at this pointBrowse repository at this point

• P0: make the model-card claims real, fix fresh-001 (97 -> 98/103)

  Show description for 62b8a2f
  

  ChelseaKR

  and

  claude

  committed
  62b8a2fView commit details

  Copy full SHA for 62b8a2f

  View code at this pointBrowse repository at this point

• Normalize transposed tables at ingest: 96/103 -> 97/103 (ADR 0005)

  Show description for b54182c
  

  ChelseaKR

  and

  claude

  committed
  b54182cView commit details

  Copy full SHA for b54182c

  View code at this pointBrowse repository at this point

Commits on Jun 13, 2026

• Expand eval suites to 103 cases, hitting the 100+ target: 96/103 live

  Show description for fcde7dc
  

  ChelseaKR

  and

  claude

  committed
  fcde7dcView commit details

  Copy full SHA for fcde7dc

  View code at this pointBrowse repository at this point

• Remediate remaining eval failures: 65/70 → 67/70, refusal suite 14/14

  Show description for 88e74c2
  

  ChelseaKR

  and

  claude

  committed
  88e74c2View commit details

  Copy full SHA for 88e74c2

  View code at this pointBrowse repository at this point

Commits on Jun 12, 2026

• Live Bedrock evals: 51/70 baseline tuned to 65/70 across three iterations

  Show description for 35bfe90
  

  ChelseaKR

  and

  claude

  committed
  35bfe90View commit details

  Copy full SHA for 35bfe90

  View code at this pointBrowse repository at this point

• Phase 1+2: corpus, assistant core, eval harness with offline baseline

  Show description for 4ab6931
  

  ChelseaKR

  and

  claude

  committed
  4ab6931View commit details

  Copy full SHA for 4ab6931

  View code at this pointBrowse repository at this point

Loading