Repository navigation

GH-30: debug runmode — doctor subcommand and live diagnostics - #33

Merged

blainemotsinger merged 33 commits into

mainfrom

GH-30

Sep 29, 2026

Commits on Sep 28, 2026
- GH-30: add eval scenario schema and loader
  blainemotsinger
  committed
  62badd8
- GH-30: add eval deterministic assertion core
  blainemotsinger
  committed
  b12fbb4
- GH-30: fix forbidden_new_paths allow matching and fail closed
  blainemotsinger
  committed
  14aa6db
- GH-30: add eval mock tier and eval-mock target
  blainemotsinger
  committed
  0e2c33f
- GH-30: fix duplicate git_rev_count rows in RunMock
  blainemotsinger
  committed
  2943925
- GH-30: add eval sandbox builder
  blainemotsinger
  committed
  2510f6f
- GH-30: add eval real-tier runner core
  blainemotsinger
  committed
  677c529
- GH-30: fix MCP tool name normalization in runner evidence
  blainemotsinger
  committed
  c3e5cf2
- GH-30: seed pilot store fixture and flip query_knowledge assert
  blainemotsinger
  committed
  b5858b7
- GH-30: raise pilot timeout and sync mock test fixture
  blainemotsinger
  committed
  acf28ca
- GH-30: lift provider block by model prefix
  blainemotsinger
  committed
  20fdf3c
- GH-30: make eval model pin configurable
  blainemotsinger
  committed
  dd17e22
- GH-30: utf8-encode sandbox file writes
  blainemotsinger
  committed
  5efbe9d
- GH-30: pin provider-lift failure message and flag fallthrough
  blainemotsinger
  committed
  3924e5b
- GH-30: add eval judge and make eval target
  blainemotsinger
  committed
  665d8bc
- GH-30: harden judge spawn against arg limits and orphaned children
  blainemotsinger
  committed
  b3a28a7
- GH-30: add wave 1 gate eval scenarios
  blainemotsinger
  committed
  18f8e31
- GH-30: tighten gate asserts and wire mock verification
  blainemotsinger
  committed
  fbf30ed
- GH-30: add wave 2 convention eval scenarios
  blainemotsinger
  committed
  51d80eb
- GH-30: add wave 3 soft eval scenarios and CI workflow
  blainemotsinger
  committed
  47d9236
- GH-30: document behavior eval targets
  blainemotsinger
  committed
  bf204c7
- GH-30: fix review findings in eval corpus and judge clip
  blainemotsinger
  committed
  b63a5f9
- GH-30: add diag logger and arg sanitizer
  blainemotsinger
  committed
  edf33ae
- GH-30: add tool-call debug logging and server debug flags
  blainemotsinger
  committed
  81a6d12
- GH-30: fix tool.done ok flag to honor result errors
  blainemotsinger
  committed
  c0821c4
- GH-30: add hook debug logging via KNM_DEBUG
  blainemotsinger
  committed
  1bec123
- GH-30: move index location logic into search package
  blainemotsinger
  committed
  0812fa4
- GH-30: add debug doctor subcommand
  blainemotsinger
  committed
  421f13e
- GH-30: fix doctor stale flag, close race, and lazy index open
  blainemotsinger
  committed
  58b44af
- GH-30: fix doctor warm-index kick ordering and close race
  blainemotsinger
  committed
  fe7c65d
- GH-30: strengthen warm-index doctor test with stderr race gate
  blainemotsinger
  committed
  ea67245
- GH-30: document debugging and warn on knowledge load errors
  blainemotsinger
  committed
  a55ede2
- GH-30: fix debug example and quiet missing knowledge file warnings
  blainemotsinger
  committed
  7ed72ae

Navigation Menu

Repository navigation

GH-30: debug runmode — doctor subcommand and live diagnostics - #33

GH-30: debug runmode — doctor subcommand and live diagnostics#33
blainemotsinger merged 33 commits into
mainfrom
GH-30

Commits on Sep 28, 2026

GH-30: add eval scenario schema and loader

GH-30: add eval deterministic assertion core

GH-30: fix forbidden_new_paths allow matching and fail closed

GH-30: add eval mock tier and eval-mock target

GH-30: fix duplicate git_rev_count rows in RunMock

GH-30: add eval sandbox builder

GH-30: add eval real-tier runner core

GH-30: fix MCP tool name normalization in runner evidence

GH-30: seed pilot store fixture and flip query_knowledge assert

GH-30: raise pilot timeout and sync mock test fixture

GH-30: lift provider block by model prefix

GH-30: make eval model pin configurable

GH-30: utf8-encode sandbox file writes

GH-30: pin provider-lift failure message and flag fallthrough

GH-30: add eval judge and make eval target

GH-30: harden judge spawn against arg limits and orphaned children

GH-30: add wave 1 gate eval scenarios

GH-30: tighten gate asserts and wire mock verification

GH-30: add wave 2 convention eval scenarios

GH-30: add wave 3 soft eval scenarios and CI workflow

GH-30: document behavior eval targets

GH-30: fix review findings in eval corpus and judge clip

GH-30: add diag logger and arg sanitizer

GH-30: add tool-call debug logging and server debug flags

GH-30: fix tool.done ok flag to honor result errors

GH-30: add hook debug logging via KNM_DEBUG

GH-30: move index location logic into search package

GH-30: add debug doctor subcommand

GH-30: fix doctor stale flag, close race, and lazy index open

GH-30: fix doctor warm-index kick ordering and close race

GH-30: strengthen warm-index doctor test with stderr race gate

GH-30: document debugging and warn on knowledge load errors

GH-30: fix debug example and quiet missing knowledge file warnings