# Inspect the research source

This is the source of a project-operated evidence-handoff candidate, not a Bulla
package release, supported installer, or new evidence standard. The prebuilt
offline installation bundle and private walkthrough materials are not published.

Locally tested on macOS arm64/Python 3.11. An outside-developer walkthrough has not yet been completed.
The standalone rehearsal used a fresh directory and environment on the same host,
not a second physical machine. No adoption or independent implementation is claimed.

## Read a report

The HTML reports work offline without JavaScript or external assets. Their JSON
counterparts contain the same assessment and supporting artifact references. These
are published outputs of local checks, not checks performed by your browser.

The published HTML adds REPORT-WRAPPING.css immediately before the original
closing </style> tag. This publication-only rule lets long status words wrap at
200% enlargement across system fonts. The original tool, JSON reports, case
archives and evidence bytes are unchanged. PUBLICATION.json records both the
original HTML hash and the styled HTML hash; LOCAL-DERIVATIONS.json retains the
original derivation identities. To reproduce the styled HTML, insert the exact
stylesheet bytes before that single closing tag in the tool's output. No script,
external stylesheet request, or evidence transformation is involved.

## Reproduce a check from source

On macOS arm64 with Python 3.11, extract this source archive into a fresh directory.
Dependency installation requires network access; inspection itself denies network
access natively. Python and macOS are not bundled. Use the exact locked wheels:

```sh
python3.11 -m venv .venv
.venv/bin/python -m pip install --only-binary=:all: --require-hashes -r requirements.txt
```

Obtain a case archive and the example CONTEXT.json separately from the source.
The context identifies the original buyer's requirements, accepted observer key
and checker. Its availability on this website does not independently establish
trust in its key or make the requirements your policy. Inspect those assumptions
before using it. A case archive cannot supply a verifier, install code or authorize
a decision.

```sh
.venv/bin/python handoff.py inspect /absolute/path/P05.zip --context /absolute/path/CONTEXT.json --out /absolute/new/P05-review
```

Outputs must not already exist. Open the resulting report.html. A missing response
produces an unresolved check while leaving the historical assessment intact. The
recipient has not thereby accepted the purchase or acquired authority to issue a
refund. Signatures authenticate the local observer's statement, not an inference
provider. No simulator, payment, or supplier request is executed.

The exact 143-member corpus remains available from the original experiment's
source-and-evidence archive at /evidence/routed-buyer-continuity/source-and-evidence.tar.
Export supports P02–P24; P01 is refused because its response remains private.

```sh
.venv/bin/python handoff.py export --source /absolute/path/safe-corpus --purchase P05 --out /absolute/new/P05.zip
```

The original checker, record adapters and schedule rules are unchanged. Equal
admitted evidence produces equal substantive outcomes for signed JSON and
ActionReceipt; reports retain the measured format overhead. This is source
inspection and local reproduction, not independent interoperability or evidence
that a dispute or remedy has been completed. The evidence can inform witnesses,
authorized reviewers or backers without supplying those institutions or powers.

## License and provenance

The research source is covered by the included repository MIT license. Bulla is
an unchanged external Apache-2.0 dependency, installed separately under the pinned
wheel hash; no third-party wheels are redistributed in this archive.
PUBLICATION.json binds the exact source members, examples and generated reports.