feat: execute isolated Bambu Studio slices

README.md

@@ -5,13 +5,15 @@ regression plans and physical-print evidence. It is designed to work with
 desktop slicers already installed by the user, beginning with Bambu Studio and
 later OrcaSlicer.
 
-The first development milestone can:
+The current pre-alpha milestone can:
 
 - discover a local Bambu Studio or OrcaSlicer executable;
 - report the detected slicer version without using a cloud account;
 - fingerprint an STL or 3MF input;
 - produce an inspectable, shell-free Bambu Studio slicing command;
-- emit a JSON plan that can become part of a reproducible regression report.
+- emit a JSON plan that can become part of a reproducible regression report;
+- execute one explicitly approved slice in a new isolated directory;
+- record a portable run manifest, output hashes, and a redacted slicer log.
 
 OpenPrintBench does not upload models, control printers, use Bambu Cloud,
 request access tokens, or load proprietary network plugins.
@@ -23,8 +25,9 @@ request access tokens, or load proprietary network plugins.
 
 ## Status
 
-Pre-alpha. The current code creates and validates slicing plans; it does not
-yet execute a slice or claim physical-print validation. See
+Pre-alpha. A run requires an explicit `--approve` flag and a fixture with pinned
+source and license provenance. The project does not claim physical-print
+validation. See
 [the roadmap](docs/ROADMAP.md), [daily progress log](docs/PROGRESS.md), and
 [evidence-gated promotion plan](docs/PROMOTION.md).
 
@@ -68,6 +71,39 @@ uv run openprintbench plan \
 The generated command is an argument array. OpenPrintBench does not construct
 or execute a shell string.
 
+Execute an isolated run only after reviewing the plan and fixture provenance:
+
+```bash
+uv run openprintbench run \
+  --input fixtures/cube-20mm.stl \
+  --run-dir /tmp/openprintbench-run-001 \
+  --machine-settings path/to/machine.json \
+  --process-settings path/to/process.json \
+  --filament-settings path/to/filament.json \
+  --profile-root path/to/installed/profiles/BBL \
+  --profile-source-url \
+    https://github.com/bambulab/BambuStudio/tree/FULL_COMMIT/resources/profiles/BBL \
+  --profile-source-commit FULL_COMMIT \
+  --profile-license AGPL-3.0-only \
+  --fixture-source-url \
+    https://github.com/kforris/OpenPrintBench/blob/FULL_COMMIT/fixtures/cube-20mm.stl \
+  --fixture-source-commit FULL_COMMIT \
+  --fixture-license CC0-1.0 \
+  --approve
+```
+
+The run directory must not already exist. OpenPrintBench gives the slicer a
+private `HOME` and temporary directory, removes secret-like environment
+variables, captures stdout/stderr into a redacted log, and fingerprints regular
+files under the isolated output directory. For STL inputs, it resolves the
+installed Bambu profile inheritance into complete private run-local JSON files
+and records both source and materialized hashes. Generated 3MF/G-code remains
+local and must not be committed.
+
+See the first
+[repeatability evidence](docs/evidence/2026-07-28-bambu-stl-repeatability.md)
+for two runs using the project-authored CC0 cube fixture.
+
 ## Development
 
 ```bash

docs/ARCHITECTURE.md

@@ -16,7 +16,31 @@ The Python package currently contains:
 - `slicers/` — slicer-specific command construction;
 - `fingerprint.py` — streaming SHA-256 file fingerprints;
 - `plan.py` — portable JSON plan assembly;
-- `cli.py` — user-facing `doctor` and `plan` commands.
+- `profiles.py` — pinned, local Bambu profile inheritance materialization;
+- `run.py` — explicit approval, isolated process execution, output
+  fingerprinting, and privacy-reviewed run manifests;
+- `cli.py` — user-facing `doctor`, `plan`, and `run` commands.
 
 Slicers are executed as separate local processes using argument arrays.
 OpenPrintBench does not link to, embed, or redistribute slicer code.
+
+## Run isolation
+
+Execution is opt-in and creates a new run directory. The slicer receives that
+directory as its working directory, a private `HOME`, a private temporary
+directory, and an environment with secret-like variables removed. Existing run
+directories are rejected to avoid mixing evidence.
+
+The run manifest uses placeholders instead of machine-local paths. It records
+the full fixture source commit and license, input/settings hashes, slicer
+version, start time, duration, exit status, timeout state, output hashes, and
+the hash of a redacted log. It sets `state` to `executed` and keeps
+`physical_validation` null; a successful process does not imply a physical
+print.
+
+For a bare STL, Bambu Studio's CLI needs complete machine, process, and
+filament configuration. OpenPrintBench recursively resolves `inherits` and
+`include` references from a user-installed profile tree, then writes complete
+profiles inside the private run directory. The source tree commit, license,
+leaf hashes, materializer version, and resulting hashes are recorded. Profile
+source files and generated slicer outputs are never added to the repository.

docs/PROGRESS.md

@@ -64,6 +64,24 @@ instead of manufacturing activity.
 - Kept X/Twitter cadence at no more than two original milestone posts or
   threads per week and excluded automated likes, follows, reposts, quote-posts,
   private messages, and generic engagement.
+- Added the first redistributable fixture: a project-authored 20 mm ASCII STL
+  cube released under CC0-1.0, with a sidecar license and pinned SHA-256.
+- Implemented the explicitly approved Bambu Studio executor linked to Issue
+  #1. It uses an argument array with `shell=False`, a new private run
+  directory, isolated `HOME`/temporary paths, secret-like environment removal,
+  a bounded timeout, redacted logging, and portable evidence manifests.
+- Added a local profile materializer because Bambu Studio CLI requires complete
+  machine/process/filament settings rather than installed leaf profiles.
+  Source commit/license, leaf hashes, materializer version, and resulting
+  profile hashes are recorded without redistributing the profiles.
+- Completed two successful Bambu Studio `02.06.00.51` digital slices with the
+  same fixture and settings. Both exited `0`; the G-code SHA-256 was identical.
+  Start time, duration, log timestamps, timing fields in `result.json`, and the
+  3MF container hash varied. No generated 3MF/G-code was committed and no
+  physical-print claim was made.
+- Audited GitHub and the authorized `@kforris_w` X account before development.
+  Issue #1 was the only open repository item; there were no open PRs, relevant
+  X mentions, or OpenPrintBench search results. No X reply or post was needed.
 
 ### Evidence
 
@@ -72,25 +90,45 @@ instead of manufacturing activity.
 - P0 draft: `docs/promotion/P0_BUILDING_IN_PUBLIC.md`
 - X destination: <https://x.com/kforris_w>
 - Public baseline CI remains green:
-  <https://github.com/kforris/OpenPrintBench/actions/runs/30328095151>
+  <https://github.com/kforris/OpenPrintBench/actions/runs/30328367492>
+- Fixture SHA-256:
+  `369c23daac96f4cde40ec6a0e13afb9be5ec4cbbf974c071c60f1009824477c4`
+- Bambu profile source:
+  `b506005bc4ee62124e24bf00e0f58656db3646a6` (`AGPL-3.0-only`)
+- Repeatability record:
+  `docs/evidence/2026-07-28-bambu-stl-repeatability.md`
+- Run 1: duration `0.650262` seconds, exit `0`, G-code SHA-256
+  `f6c6365d65ecf1110f4aefd1097558378006ad4f1738648fba74c0bfa95205c5`.
+- Run 2: duration `0.472667` seconds, exit `0`, G-code SHA-256
+  `f6c6365d65ecf1110f4aefd1097558378006ad4f1738648fba74c0bfa95205c5`.
+- Local validation after implementation: Ruff lint/format, mypy, 45 tests,
+  0 skipped/xfail, 87.38% branch coverage, package build, and
+  `git diff --check` passed.
+- Functional PR: <https://github.com/kforris/OpenPrintBench/pull/2>
+- PR CI (Python 3.11, 3.12, and 3.13):
+  <https://github.com/kforris/OpenPrintBench/actions/runs/30368257211>
+- No public release exists yet.
 
 ### Blockers
 
-- No functional PR has been opened or merged yet, so the P0 publication gate is
-  not met.
-- No redistributable STL/3MF fixture has been selected yet.
+- The first functional PR is not merged and its `main` CI gate is not yet
+  satisfied, so P0 publication remains blocked.
+- Only one of the ten required redistributable fixtures is complete.
 - The target Bambu printer model is not yet recorded.
+- OrcaSlicer is not installed locally and no verified CI fixture is available.
 
 ### Next
 
-1. Select the first open fixture and implement the isolated Bambu Studio
-   execution path in a functional PR linked to Issue #1.
+1. Review the functional PR linked to Issue #1 and merge only after its CI is
+   green; then verify the resulting `main` CI.
 2. After that PR is merged and `main` CI is green, create and verify the P0
    visual, refresh the exact post copy, and publish through the authorized X
    account.
-3. Check GitHub issues/PRs and relevant X replies before selecting daily work;
+3. Add the next license-clear fixture or generator without weakening the
+   provenance gate.
+4. Check GitHub issues/PRs and relevant X replies before selecting daily work;
    reply with evidence after changes and close only resolved GitHub issues.
-4. Measure public repository and X signals 24 and 72 hours after the first
+5. Measure public repository and X signals 24 and 72 hours after the first
    post.
 
 ## 2026-07-29
@@ -147,3 +185,59 @@ instead of manufacturing activity.
 4. When the P0 gate is satisfied, refresh the exact copy and metrics, create
    and inspect the evidence-backed 1600x900 visual, then publish and verify the
    authorized two-post thread.
+
+## 2026-07-30
+
+### Completed
+
+- Refreshed the public repository before selecting work. `origin/main` was
+  `2e61c5c29d7522d4aa4be8dcf6afdb563bde8121`; Issue #1 and PR #2 remained
+  open, with no new review, review comment, issue comment, or external
+  contribution.
+- Found PR #2 had changed from mergeable/clean to conflicting/dirty after the
+  2026-07-29 progress-only commit modified `docs/PROGRESS.md` on `main`.
+- Selected one bounded unblocker instead of opening a second implementation
+  PR: merged the latest `origin/main` into the existing PR branch and resolved
+  the progress-log conflict without dropping either day's verified record.
+- Verified the existing Comet X session was logged in as `@kforris_w`. The
+  mentions view was empty and an exact latest search for `OpenPrintBench`
+  returned no results after one retry. No private message was inspected and no
+  X reply or post was made.
+
+### Evidence
+
+- Open Issue #1: <https://github.com/kforris/OpenPrintBench/issues/1>
+- Open PR #2: <https://github.com/kforris/OpenPrintBench/pull/2>
+- Pre-resolution PR state: `CONFLICTING` / `DIRTY` at
+  `817e8a7c8fb7c21264fead9229d277d3da55b6e2`, with no reviews or comments.
+- Previous PR CI:
+  <https://github.com/kforris/OpenPrintBench/actions/runs/30368347793>
+- Latest baseline `main` CI:
+  <https://github.com/kforris/OpenPrintBench/actions/runs/30459104729>
+- Local validation after conflict resolution: `uv lock --check`, Ruff
+  lint/format, mypy, 45 tests with 0 skipped/xfail and 87.38% branch coverage,
+  package build, and `git diff --check` passed.
+- Public repository snapshot: 1 star, 0 forks, no tags, and no releases.
+
+### Blockers
+
+- PR #2 is still unmerged and requires review/merge; its resulting green
+  `main` CI does not yet exist, so Issue #1 cannot be closed.
+- The P0 visual and two-post thread remain blocked on the merged functional PR
+  and resulting green `main` CI.
+- The target Bambu printer model, material, and human observation remain
+  unavailable; no physical validation is claimed.
+- OrcaSlicer is not installed locally, and nine additional redistributable
+  fixtures or generators remain to reach the v0.1 fixture target.
+
+### Next
+
+1. Verify the updated PR branch CI and mergeability, then wait for real review
+   and merge without self-merging.
+2. After PR #2 is merged, verify the resulting `main` CI, update and close
+   Issue #1 with durable evidence, and mark only the proven roadmap items.
+3. When the P0 gate is satisfied, refresh the exact copy and metrics, create
+   and inspect the evidence-backed 1600x900 visual, then publish and verify the
+   authorized two-post thread.
+4. Add the next license-clear fixture or generator only after the current
+   implementation PR is no longer blocking the one-topic workflow.

docs/ROADMAP.md

@@ -9,9 +9,10 @@ requirements and do not guarantee acceptance into any support program.
 - [x] Probe Bambu Studio version and supported command-line options.
 - [x] Produce a deterministic, shell-free slice plan.
 - [x] Fingerprint source models with SHA-256.
-- [ ] Execute an explicitly approved Bambu Studio slice in an isolated run
-      directory.
-- [ ] Capture duration, exit status, output hashes, and a redacted log.
+- [x] Execute an explicitly approved Bambu Studio slice in an isolated run
+      directory. See the
+      [2026-07-28 evidence](evidence/2026-07-28-bambu-stl-repeatability.md).
+- [x] Capture duration, exit status, output hashes, and a redacted log.
 - [ ] Add an OrcaSlicer execution adapter after local installation or a
       verified CI fixture is available.
 - [ ] Publish at least ten redistributable test fixtures or generators.

docs/evidence/2026-07-28-bambu-stl-repeatability.md

@@ -0,0 +1,76 @@
+# Bambu STL repeatability evidence — 2026-07-28
+
+This record covers two local, explicitly approved digital slicing runs. It
+does not claim a physical print, printer control, cloud access, or hardware
+validation.
+
+## Provenance
+
+- OpenPrintBench fixture commit:
+  `2d2c54e9d9ddc53a365606b9d9d070b94caeb64c`
+- Fixture: `fixtures/cube-20mm.stl`, project-authored, `CC0-1.0`
+- Fixture SHA-256:
+  `369c23daac96f4cde40ec6a0e13afb9be5ec4cbbf974c071c60f1009824477c4`
+- Fixture size: `1517` bytes
+- Slicer: Bambu Studio `02.06.00.51`
+- Installed profile source commit:
+  `b506005bc4ee62124e24bf00e0f58656db3646a6`
+- Installed profile source license: `AGPL-3.0-only`
+- Materializer: `openprintbench-bambu-profile-merge-v1`
+- Machine: `Bambu Lab X1 Carbon 0.4 nozzle.json`
+- Process: `0.20mm Standard @BBL X1C.json`
+- Filament: `Generic PLA.json`
+
+The selected profile files were used only as local inputs. Their inheritance
+was materialized into the isolated run directory; neither those profiles nor
+generated 3MF/G-code files are stored in this repository.
+
+## Captured development attempts
+
+Before the two successful runs, the executor captured two non-success states
+without producing output files:
+
+- Installed leaf profiles were passed directly: duration `0.278480` seconds,
+  exit status `-11` (signal 11), redacted log SHA-256
+  `f1de4e2a8b778b70ef41d56836046f5e73b8e07aa756e46471a11e3850dcc3fc`.
+  This led to the complete-profile materializer.
+- Materialized profiles were used with an incomplete orientation argument:
+  duration `0.070092` seconds, exit status `254`, redacted log SHA-256
+  `5a9942777f3bb22d04188093b67f318b551a4b09ad90b9430f289b1aeb410a8d`.
+  The argument was corrected to the documented `--orient 1` form.
+
+Both manifests retained `physical_validation: null`.
+
+## Run results
+
+| Field | Run 1 | Run 2 | Classification |
+| --- | --- | --- | --- |
+| Exit status | `0` | `0` | stable |
+| Timed out | `false` | `false` | stable |
+| Duration (seconds) | `0.650262` | `0.472667` | expected variable |
+| G-code size (bytes) | `278335` | `278335` | stable |
+| G-code SHA-256 | `f6c6365d65ecf1110f4aefd1097558378006ad4f1738648fba74c0bfa95205c5` | same | stable |
+| 3MF size (bytes) | `55621` | `55621` | stable |
+| 3MF SHA-256 | `57ce07f1176c466cd20a97908038a5c3a8d167937eec796544bb73750d0d3c9d` | `146dcd84ea9ed468ca52abc8e8149d1d44b4310445dd341d71aa9b126f6bab31` | variable container bytes |
+| `result.json` SHA-256 | `d97fc5b7900c937e64bfb114a26089fa0196ec79dbfbb7210147baed2dbcb547` | `17cee9b309239fe0dc6a74681a76f73ca936f58e1aec10d53047a1ccdf591e23` | variable timing fields |
+| Redacted log SHA-256 | `5d701094d4e0cd5535a456dd9c28b1923070f83a007eeee9b055e0e59c41429e` | `25fa3d3424d53cddee09f9dd3a0330000908a9fd6d01b9be5956b1c3586f3550` | variable timestamps |
+| Physical validation | `null` | `null` | not performed |
+
+## Stable inputs and settings
+
+The portable command array, fixture hash, slicer version, source profile
+hashes, materialized profile hashes, output names, approval state, and exit
+status matched across both runs. In particular:
+
+- materialized machine SHA-256:
+  `b5a60ef9b156acce52fa5b083595af1a00d7c478ccdc05f1b7f4e59b5e6c6a7b`
+- materialized process SHA-256:
+  `5d6ec0717d02b0293d45798d3c776d141e6a3f03876ee519863028de6b992039`
+- materialized filament SHA-256:
+  `2ee424b81096dbdf4f6c8e09a7f7f46a840a5b44fa857db19c290871ff2c67b8`
+
+The differing `result.json` fields were execution-time measurements such as
+`export_time`, `prepare_time`, perimeter/infill time, and total sliced time.
+The redacted logs differed at timestamp positions. The matching G-code hash is
+the strongest byte-level repeatability result from these two runs; the varying
+3MF container hash must not be treated as a slicing regression by itself.

fixtures/README.md

@@ -0,0 +1,18 @@
+# Redistributable fixtures
+
+`cube-20mm.stl` is an original, deterministic ASCII STL created for
+OpenPrintBench. It contains a 20 mm cube represented by 12 triangles and is
+dedicated to the public domain under CC0-1.0. Its adjacent `.license` file is
+the authoritative redistribution notice.
+
+For an evidence run, pin the fixture to the full Git commit that contains the
+exact bytes and use:
+
+- source URL:
+  `https://github.com/kforris/OpenPrintBench/blob/<commit>/fixtures/cube-20mm.stl`
+- source commit: the same full 40-character Git SHA;
+- license: `CC0-1.0`;
+- input SHA-256: calculated by OpenPrintBench at execution time.
+
+The fixture is digital test geometry only. Its presence does not imply that a
+slice has run or that a physical print has been validated.