Skip to content
Open
Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
53 commits
Select commit Hold shift + click to select a range
ea5ae80
feat(research): isolate durable automation and preserve capture lineage
jdharcourt Nov 17, 2026
4bad1bf
fix(research): align automation verification with current ci
jdharcourt Nov 17, 2026
a920bf7
docs(agents): require research automation for decoding challenges
jdharcourt Nov 17, 2026
e282bf1
Docs - A01 Restore the research plan and phase register
Stephen316 Nov 17, 2026
fb3063a
Feat - A02 Restore pinned capture and evidence intake
Stephen316 Nov 17, 2026
78754ae
Docs - A03 Restore controller recovery records
Stephen316 Nov 17, 2026
7a5348b
Feat - A04 Restore SOLIDWORKS decoding and census
Stephen316 Nov 17, 2026
ed54366
Feat - A05 Restore Fusion decoding and exact census
Stephen316 Nov 17, 2026
430316c
Feat - A06 Restore paired semantic evidence
Stephen316 Nov 17, 2026
ce9c8d1
Feat - A07 Restore native inference and authenticated grids
Stephen316 Nov 17, 2026
9a91ed6
Feat - A08 Restore application worker research
Stephen316 Nov 17, 2026
108dbe7
Feat - A09 Restore two-pad field and codec research
Stephen316 Nov 17, 2026
4fae1ef
Feat - A11 Restore Fusion writer research
Stephen316 Nov 17, 2026
7facadf
Feat - A12 Restore SOLIDWORKS writer research
Stephen316 Nov 17, 2026
a3b9677
Feat - A13 Restore native qualification guards
Stephen316 Nov 17, 2026
888e7bc
Fix - A14 Restore native verification in the merge queue
Stephen316 Nov 17, 2026
7bc3d12
Docs - A15 Restore pilot commissioning evidence
Stephen316 Nov 17, 2026
00ea6bb
Feat - A16 Restore FreeCAD shape and edit history work
Stephen316 Nov 17, 2026
1e87ddf
fix(research): prepare automated decoding for native testing
jdharcourt Nov 17, 2026
a6cc279
fix(research): make reviewed call contracts explicit
jdharcourt Nov 17, 2026
7b33e12
feat(research): queue circular solidworks readback (A08f)
Stephen316 Nov 17, 2026
c572f1a
fix(research): fence worker execution and retain native progress (A07…
Stephen316 Nov 17, 2026
ee53cf7
fix(research): retain owned SOLIDWORKS sessions across document closes
Stephen316 Nov 17, 2026
56b2ad9
feat(solidworks): project standalone circles and construct visual state
Stephen316 Nov 17, 2026
9fc3c9a
fix(solidworks): keep the final readback control hidden
Stephen316 Nov 17, 2026
8509332
fix(solidworks): complete owned Boss witness cleanup
Stephen316 Nov 17, 2026
c7cbdd2
feat(research): freeze and replay offline pilot evidence
Stephen316 Nov 17, 2026
5b0608d
feat(research): prepare and supervise fixed offline pilot window
Stephen316 Nov 17, 2026
d3f4855
docs(research): record live offline pilot and reproducible evidence
Stephen316 Nov 17, 2026
810ba00
fix(freecad): separate native replay from editable reconstruction
Stephen316 Nov 17, 2026
bce0c0b
test(freecad): seal A16 rectangle and native saved-edit evidence
Stephen316 Nov 17, 2026
d4d3b64
Merge remote-tracking branch 'origin/feat/pr94-automation-only-202611…
Stephen316 Nov 17, 2026
5018f10
feat(solidworks): construct explicit document appearance (A12)
Stephen316 Nov 17, 2026
b475f94
Fix - Bind Fusion inference phase evidence
Stephen316 Nov 18, 2026
0788b0d
Fix - Prepare owned Fusion scope capture
Stephen316 Nov 18, 2026
62884aa
Feat - Replay recovered Fusion admission evidence
Stephen316 Nov 18, 2026
e976c92
Fix - Preserve native dimension roles and graph matches
Stephen316 Nov 18, 2026
dfe2bd4
fix(research): bind inference dependency inputs to accepted results
Stephen316 Nov 18, 2026
51f299d
fix: isolate fusion runner git source reads
Stephen316 Nov 18, 2026
2467ae8
fix(research): validate semantic comparison dependency snapshots
Stephen316 Nov 18, 2026
5503cbe
fix(research): finish deterministic verifier and Windows fixture repairs
Stephen316 Nov 18, 2026
0b66cc0
Fix - Retain local msgpack dependency update
Stephen316 Nov 18, 2026
cc2fb28
Merge branch 'main' into feat/pr94-automation-only-20261117
jdharcourt Nov 18, 2026
5df61c9
fix(convert): validate scoped ends and align freecad fixtures
Stephen316 Nov 19, 2026
e57a5ff
Feat - Add Codex advice for parked decoder jobs
Stephen316 Nov 19, 2026
16edc4f
perf(research): batch SOLIDWORKS discovery readbacks
Stephen316 Nov 19, 2026
640e5ed
Perf - Reuse the SDK for grouped rescans
Stephen316 Nov 19, 2026
40980db
perf(research): streamline SOLIDWORKS native authoring
Stephen316 Nov 19, 2026
553956d
Perf - Speed up deterministic decoding
Stephen316 Nov 19, 2026
2a143e8
test(research): place batch binding test with controller tests
Stephen316 Nov 19, 2026
a6062ea
feat(research): connect cut-extrude planning to decoder evidence
Stephen316 Nov 19, 2026
e282c31
Fix - Meet the rationale comment policy in three research tests
Stephen316 Nov 19, 2026
12a94e5
Fix - Remove side effects from rescan group assertions
Stephen316 Nov 19, 2026
File filter

Filter by extension

Filter by extension


Conversations
Failed to load comments.
Loading
Jump to
The table of contents is too big for display.
Diff view
Diff view
  •  
  •  
  •  
2 changes: 2 additions & 0 deletions .gitattributes
Original file line number Diff line number Diff line change
Expand Up @@ -13,3 +13,5 @@
*.x_t binary

/re export-ignore

re/automation/Corpus.json linguist-generated=true
3 changes: 3 additions & 0 deletions .github/scripts/CiAllowlist.cjs
Original file line number Diff line number Diff line change
Expand Up @@ -7,6 +7,9 @@ const QueueJobs = [
"Run the complete test suite",
"Test suite / Plan",
"Test suite / Report",
"Test suite / Platform (windows-2025)",
"Test suite / Platform (macos-26)",
"Linux kernel OOM witness",
"Analyze actions",
"Analyze python",
"Analyze javascript-typescript",
Expand Down
228 changes: 225 additions & 3 deletions .github/workflows/MergeQueue.yml
Original file line number Diff line number Diff line change
Expand Up @@ -33,6 +33,8 @@ jobs:
uses: actions/checkout@3d3c42e5aac5ba805825da76410c181273ba90b1 # v7.0.1
with:
persist-credentials: false
fetch-depth: 0
filter: blob:none
ref: ${{ github.event_name == 'push' && env.AuditSha || github.sha }}

- name: Confirm the pinned audit revision
Expand Down Expand Up @@ -69,6 +71,9 @@ jobs:
- name: Verify packaged native schema provenance
run: uv run --no-sync --python 3.14 python re/Fusion/Reconstruction/EntityGraph/PromoteCatalog.py --check

- name: Hydrate sealed original research inputs before collection
run: uv run --no-sync --python 3.14 python re/automation/CaptureHistory.py fetch

- name: Collect the complete test suite
run: |
set -euo pipefail
Expand Down Expand Up @@ -154,6 +159,7 @@ jobs:
--max-tests 1200 \
--output "${RUNNER_TEMP}/ShardPlan/ShardPlan.json" \
--summary "$GITHUB_STEP_SUMMARY"
cp "${RUNNER_TEMP}/Collected.txt" "${RUNNER_TEMP}/ShardPlan/Collected.txt"
echo "shards=$(jq -c '[.Shards[].Index]' "${RUNNER_TEMP}/ShardPlan/ShardPlan.json")" >> "$GITHUB_OUTPUT"

- name: Upload the shard plan
Expand All @@ -179,6 +185,8 @@ jobs:
uses: actions/checkout@3d3c42e5aac5ba805825da76410c181273ba90b1 # v7.0.1
with:
persist-credentials: false
fetch-depth: 0
filter: blob:none
ref: ${{ github.event_name == 'push' && env.AuditSha || github.sha }}

- name: Set up uv
Expand Down Expand Up @@ -229,17 +237,35 @@ jobs:
if grep --quiet '^tests/CliResearch/' "${RUNNER_TEMP}/ShardTests.txt"; then
echo "research=true" >> "$GITHUB_OUTPUT"
fi
if grep --quiet '^tests/Research/Inference/' "${RUNNER_TEMP}/ShardTests.txt"; then
echo "inference=true" >> "$GITHUB_OUTPUT"
fi
if grep --extended-regexp --quiet '^tests/(CliResearch/|Research/(Archives/|Inference/|Outputs/Ci|CorpusFlowTests|CaptureLineageFlowTests|Reader|FusionReader))' "${RUNNER_TEMP}/ShardTests.txt"; then
echo "lineage=true" >> "$GITHUB_OUTPUT"
fi

- name: Checkout pinned PR94 tools for research CLI cases
- name: Checkout candidate tools separately for research CLI cases
if: steps.selection.outputs.research == 'true'
uses: actions/checkout@3d3c42e5aac5ba805825da76410c181273ba90b1 # v7.0.1
with:
repository: Stephen316/Kit
ref: b86f30c7e41abdc3812ada795a193377f6ddd220
ref: ${{ github.event_name == 'push' && env.AuditSha || github.sha }}
path: ResearchCheckout
filter: blob:none
fetch-depth: 0
persist-credentials: false

- name: Synchronize locked dependencies for inference shards
if: steps.selection.outputs.inference == 'true'
run: uv sync --frozen --no-default-groups --extra dev --group dev --python 3.14 --compile-bytecode

- name: Hydrate sealed original inputs for research shards
if: steps.selection.outputs.lineage == 'true'
run: uv run --no-sync --python 3.14 python re/automation/CaptureHistory.py fetch

- name: Hydrate candidate CLI fixture inputs
if: steps.selection.outputs.research == 'true'
run: uv run --no-sync --python 3.14 python re/automation/CaptureHistory.py fetch --root ResearchCheckout

- name: Run this shard's tests
id: tests
env:
Expand All @@ -261,6 +287,199 @@ jobs:
if-no-files-found: error
retention-days: 7

research-windows:
if: github.event_name == 'merge_group' || github.event_name == 'push'
needs: test-plan
name: Test suite / Platform (windows-2025)
runs-on: windows-2025
timeout-minutes: 30
env:
PlatformRunner: windows-2025
NativePattern: "^tests/Research/Windows/"
SelectionPatterns: |-
^tests/Research/Windows/
^tests/Research/Outputs/
^tests/Research/Applications/ApplicationParkTests\.py::
^tests/Research/Applications/Acceptance/AcceptanceSupervisorTests\.py::
^tests/Research/Applications/Acceptance/AcceptanceCrashTests\.py::
^tests/Research/Supervisor/Grouping/SupervisorBatchTests\.py::CheckBatchCap
^tests/Research/Supervisor/Grouping/SupervisorSoloTests\.py::CheckBatchCrash
^tests/Research/Owners/
^tests/Research/WindowsRecords/
defaults:
run:
shell: bash
steps: &ResearchPlatformSteps
- name: Checkout exact candidate
uses: actions/checkout@3d3c42e5aac5ba805825da76410c181273ba90b1 # v7.0.1
with:
persist-credentials: false
fetch-depth: 0
filter: blob:none
ref: ${{ github.event_name == 'push' && env.AuditSha || github.sha }}

- name: Set up uv
uses: astral-sh/setup-uv@68559e99cc807c1eb7e9238fd43a15bba460eec2 # v10.3.3
with:
version: "0.12.5"
enable-cache: true
cache-dependency-glob: uv.lock

- name: Synchronize locked platform process dependencies
run: uv sync --frozen --no-default-groups --extra dev --python 3.14

- name: Download the complete shared plan
uses: actions/download-artifact@9000827ccba6bdab643e8b6fd33ac0654aef8333 # v8.0.2
with:
name: ShardPlan
path: ${{ runner.temp }}/ShardPlan

- name: Select required platform cases from the shared collected roster
run: |
set -euo pipefail
mkdir -p "${RUNNER_TEMP}/PlatformReport"
Roster="${RUNNER_TEMP}/ShardPlan/Collected.txt"
Selected="${RUNNER_TEMP}/PlatformReport/Selected.txt"
test -s "$Roster"
: > "$Selected"
while IFS= read -r Pattern; do
test -n "$Pattern"
grep -E "$Pattern" "$Roster" >> "$Selected"
done <<< "$SelectionPatterns"
sort -u "$Selected" > "${Selected}.sorted"
mv "${Selected}.sorted" "$Selected"
test -s "$Selected"
grep -E "$NativePattern" "$Selected" > "${RUNNER_TEMP}/PlatformReport/Native.txt"
test -s "${RUNNER_TEMP}/PlatformReport/Native.txt"

- name: Execute the required platform process cases
id: tests
run: >-
uv run --no-sync --python 3.14 python -m pytest -p no:cacheprovider
--report-log="${RUNNER_TEMP}/PlatformReport/Report.jsonl"
"@${RUNNER_TEMP}/PlatformReport/Selected.txt"

- name: Require complete platform outcomes and actual native-platform passes
run: |
set -euo pipefail
jq --slurp --exit-status \
--rawfile Selected "${RUNNER_TEMP}/PlatformReport/Selected.txt" \
--rawfile Native "${RUNNER_TEMP}/PlatformReport/Native.txt" '
($Selected | split("\n") | map(select(length > 0))) as $Expected
| ($Native | split("\n") | map(select(length > 0))) as $NativeIds
| any(.[]; .["$report_type"] == "SessionFinish" and .exitstatus == 0)
and all(.[]; .outcome != "failed")
and ([.[] | select(.["$report_type"] == "TestReport")
| select(.when == "call" or (.when == "setup" and .outcome == "skipped"))
| .nodeid] | sort) == ($Expected | sort)
and any(.[]; .["$report_type"] == "TestReport" and .when == "call"
and .outcome == "passed" and (.nodeid as $Id | $NativeIds | index($Id) != null))
' "${RUNNER_TEMP}/PlatformReport/Report.jsonl"

- name: Publish exact platform selection and process outcomes
if: always() && steps.tests.outcome != 'skipped'
uses: actions/upload-artifact@cf430e030ddbb5b0abf93d22962f4752f3646cd9 # v7.0.2
with:
name: PlatformReport-${{ env.PlatformRunner }}-${{ github.run_attempt }}
path: ${{ runner.temp }}/PlatformReport
if-no-files-found: error
retention-days: 14

research-macos:
if: github.event_name == 'merge_group' || github.event_name == 'push'
needs: test-plan
name: Test suite / Platform (macos-26)
runs-on: macos-26
timeout-minutes: 30
env:
PlatformRunner: macos-26
NativePattern: '^tests/Research/Applications/FusionColdStartTests\.py::'
SelectionPatterns: |-
^tests/Research/Supervisor[^/]*Tests\.py::
^tests/Research/Supervisor/Supervisor[^/]*Tests\.py::
^tests/Research/Outputs/
^tests/Research/Applications/
^tests/Research/Supervisor/Grouping/
^tests/Research/Supervisor/Launch/
^tests/Research/Owners/
^tests/Research/Calibration/
defaults:
run:
shell: bash
steps: *ResearchPlatformSteps

linux-oom-witness:
if: github.event_name == 'merge_group' || github.event_name == 'push'
name: Linux kernel OOM witness
runs-on: ubuntu-24.04
timeout-minutes: 5
permissions:
contents: read
steps:
- name: Checkout exact candidate
uses: actions/checkout@3d3c42e5aac5ba805825da76410c181273ba90b1 # v7.0.1
with:
ref: ${{ github.event_name == 'push' && env.AuditSha || github.sha }}
persist-credentials: false

- name: Observe the bounded Linux kernel memory failure
shell: bash
env:
WITNESS_REVISION: ${{ github.event_name == 'push' && env.AuditSha || github.sha }}
WITNESS_REPORT: ${{ runner.temp }}/LinuxOomWitness.json
run: |
test "$(git rev-parse HEAD)" = "$WITNESS_REVISION"
source_blob="$(git rev-parse "$WITNESS_REVISION:re/automation/LinuxOomWitness.py")"
witness="$(mktemp -d "$RUNNER_TEMP/kit-oom-witness.XXXXXXXX")/LinuxOomWitness.py"
git cat-file blob "$source_blob" > "$witness"
source_sha="$(sha256sum "$witness")"
chmod 0400 "$witness"
sudo -n -- /usr/bin/env -i SUDO_UID="$(id -u)" SUDO_GID="$(id -g)" \
/usr/bin/python3 -I -S "$witness" --source-revision "$WITNESS_REVISION" \
--source-blob "$source_blob" --source-sha256 "${source_sha%% *}" \
--output "$WITNESS_REPORT"

- name: Publish sanitized kernel evidence
uses: actions/upload-artifact@cf430e030ddbb5b0abf93d22962f4752f3646cd9 # v7.0.2
with:
name: LinuxOomWitness-${{ github.run_attempt }}
path: ${{ runner.temp }}/LinuxOomWitness.json
if-no-files-found: error
retention-days: 14

- name: Observe kernel OOM through the research supervisor
shell: bash
env:
WITNESS_REVISION: ${{ github.event_name == 'push' && env.AuditSha || github.sha }}
WITNESS_REPORT: ${{ runner.temp }}/LinuxSupervisorOomWitness.json
run: |
test "$(git rev-parse HEAD)" = "$WITNESS_REVISION"
test -z "$(git status --porcelain --untracked-files=all -- re/automation)"
stage="$(sudo -n -- /usr/bin/mktemp -d /tmp/kit-supervisor-oom.XXXXXXXX)"
trap 'sudo -n -- /usr/bin/rm -f -- "$stage/LinuxOomGuest.py" "$stage/LinuxOomSupervisor.py" "$stage/LinuxOomWitness.py"; sudo -n -- /usr/bin/rmdir -- "$stage"' EXIT
sudo -n -- /usr/bin/chmod 0755 "$stage"
sources=()
for file in LinuxOomGuest.py LinuxOomSupervisor.py LinuxOomWitness.py; do
blob="$(git rev-parse "$WITNESS_REVISION:re/automation/$file")"
digest="$(git cat-file blob "$blob" | sha256sum)"
git cat-file blob "$blob" | sudo -n -- /usr/bin/tee "$stage/$file" > /dev/null
sudo -n -- /usr/bin/chmod 0444 "$stage/$file"
actual="$(sha256sum "$stage/$file")"
test "${digest%% *}" = "${actual%% *}"
sources+=(--source "$file:$blob:${digest%% *}")
done
sudo -n -- /usr/bin/env -i SUDO_UID="$(id -u)" SUDO_GID="$(id -g)" \
/usr/bin/python3 -B -I -S "$stage/LinuxOomSupervisor.py" --repository "$(pwd -P)" \
--revision "$WITNESS_REVISION" --output "$WITNESS_REPORT" "${sources[@]}"

- name: Publish sanitized supervisor OOM evidence
uses: actions/upload-artifact@cf430e030ddbb5b0abf93d22962f4752f3646cd9 # v7.0.2
with:
name: LinuxSupervisorOomWitness-${{ github.run_attempt }}
path: ${{ runner.temp }}/LinuxSupervisorOomWitness.json
if-no-files-found: error
retention-days: 14

test-report:
if: ${{ !cancelled() && (github.event_name == 'merge_group' || github.event_name == 'push') && needs.test-plan.result == 'success' }}
needs: [test-plan, test-shard]
Expand Down Expand Up @@ -973,6 +1192,9 @@ jobs:
test-plan,
test-shard,
test-report,
research-windows,
research-macos,
linux-oom-witness,
verify-end-to-end,
analyze,
bandit,
Expand Down
22 changes: 14 additions & 8 deletions .github/workflows/WorkflowQuality.yml
Original file line number Diff line number Diff line change
Expand Up @@ -81,18 +81,18 @@ jobs:
env:
KIT_RESEARCH_CHECKOUT: ${{ github.workspace }}/ResearchCheckout
steps:
- name: Checkout CLI repository
- name: Checkout candidate CLI source
uses: actions/checkout@3d3c42e5aac5ba805825da76410c181273ba90b1 # v7.0.1
with:
persist-credentials: false

- name: Checkout pinned PR94 research tools
- name: Checkout candidate research tools separately
uses: actions/checkout@3d3c42e5aac5ba805825da76410c181273ba90b1 # v7.0.1
with:
repository: Stephen316/Kit
ref: b86f30c7e41abdc3812ada795a193377f6ddd220
ref: ${{ github.sha }}
path: ResearchCheckout
fetch-depth: 0
filter: blob:none
persist-credentials: false

- name: Set up uv
Expand All @@ -105,6 +105,9 @@ jobs:
- name: Install CLI and compatibility dependencies
run: uv sync --frozen --no-default-groups --extra dev --python 3.14

- name: Hydrate sealed research inputs for candidate CLI cases
run: uv run --no-sync --python 3.14 python re/automation/CaptureHistory.py fetch --root ResearchCheckout

- name: Verify deterministic tools through the installed CLI
run: >-
uv run --no-sync --python 3.14 python -m pytest tests/CliResearch -q
Expand All @@ -119,18 +122,18 @@ jobs:
env:
KIT_RESEARCH_CHECKOUT: ${{ github.workspace }}/ResearchCheckout
steps:
- name: Checkout CLI repository
- name: Checkout candidate CLI source
uses: actions/checkout@3d3c42e5aac5ba805825da76410c181273ba90b1 # v7.0.1
with:
persist-credentials: false

- name: Checkout pinned PR94 research tools
- name: Checkout candidate research tools separately
uses: actions/checkout@3d3c42e5aac5ba805825da76410c181273ba90b1 # v7.0.1
with:
repository: Stephen316/Kit
ref: b86f30c7e41abdc3812ada795a193377f6ddd220
ref: ${{ github.sha }}
path: ResearchCheckout
fetch-depth: 0
filter: blob:none
persist-credentials: false

- name: Set up uv
Expand All @@ -143,6 +146,9 @@ jobs:
- name: Install CLI and compatibility dependencies
run: uv sync --frozen --no-default-groups --extra dev --python 3.14

- name: Hydrate sealed research inputs for candidate CLI cases
run: uv run --no-sync --python 3.14 python re/automation/CaptureHistory.py fetch --root ResearchCheckout

- name: Verify reader and controller exit statuses through the installed CLI
run: >-
uv run --no-sync --python 3.14 python -m pytest
Expand Down
2 changes: 2 additions & 0 deletions AGENTS.md
Original file line number Diff line number Diff line change
Expand Up @@ -19,6 +19,8 @@ key boundaries with the REA MCP server or the `npx rea-agents` CLI, generating a
of extreme models through the vendor API, correlating that corpus to locate every field, and only
then building the writer. Put both documents in every sub-agent brief for CAD work.

For decoding challenges, agents MUST use the research automation pipeline for supported capture, deterministic comparison, inference and evidence tracking. Investigate unresolved findings and implement decoder rules from the recorded evidence. Validate changes through deterministic checks; a detected pattern alone is not a proven rule. Report missing automation capabilities explicitly

## Documentation and gotchas

Read the [completion roadmap](Roadmap.md) and [conversion backlog](Documentation/Maintenance/ConversionBacklog.md)
Expand Down
Original file line number Diff line number Diff line change
@@ -0,0 +1,7 @@
# Verify the installed research CLI against the workflow candidate

The Linux and Windows research CLI compatibility jobs use the exact GitHub candidate for the installed CLI source and a separate `ResearchCheckout` for its research tools. The research checkout retains full Git history so the readers and corpus can resolve historical object pins. A merge-queue shard checks out the same candidate tools when its test plan includes research CLI cases.

The compatibility fixtures check both local Git revisions against the workflow SHA before selecting the research root. The supervisor scenario launches its configured runner from that root. The launcher signal scenarios use a separate disposable script to hold the actual child startup boundary. Reader cases compare direct and CLI output and check the selected SDK's receipt provenance.

Follow [the research CLI procedure](../../Tasks/ResearchCommandLine.md) for a working installation. Exact-candidate hosted results are pending.
Loading
Loading