Btike/bulk-bg-remove

Review signals: 1 connection review.

Reviewed as expected for a test run: attempted writes that failed, so nothing was written.

Repository processes attempted these non-loopback connections; proxies can hide the final hostname. 12 distinct locations. 15 attributed processes.

OSV scanned supported lockfiles; unpinned dependencies and unsupported formats are outside this check.

A short run cannot establish that code is safe.

Connections show IP addresses; proxies can hide the final hostname.

Tracing covers dispatched commands and their children, not the whole OS; guest root can interfere with tracing.

Full safety evidence →

• Each photo is cut out into a new output folder (tested with -o, since the default <name>-no-bg location is not an argv destination): Real input produced the expected result: 1 done

  Source: Output goes to a new `<name>-no-bg/` folder next to the input.

  Input: <guest-path>

  Expected: one image processed into the new folder <guest-path>

  Command: <guest-path> <guest-path> -o <guest-path> -y

  Observed: …1 done · 0 failed · 0 already done · 13s ! showcase.png: edge <email> MP to fit free RAM output → <guest-path>

• --dry-run --json plans the run as JSON with the model chosen: Real input produced the expected result: "model": "silueta"

  Source: scan and plan only; one JSON event per line for agents

  Input: <guest-path>

  Expected: JSON plan event naming the auto-chosen model silueta

  Command: <guest-path> <guest-path> --dry-run --json

  Observed: …"model": "silueta", "model_choice": {"profile": "uniform", "model": "silueta", "score": 8.1, "best_model": "birefnet-general-lite", "best_score": 8.6, "best_needs_gb": 11.7, "needs_gb": 3.0, "fits": true, "gap": 0.5}, "edges": "matte", "output_format": "png", "finish": null, "already_done": 0}

• --shop plans a white square 2000 px JPEG output: Real input produced the expected result: "output_format": "jpg"

  Source: `--shop` puts every product on white, square, 2000 px and ready to

  Input: <guest-path>

  Expected: plan with output_format jpg

  Command: <guest-path> <guest-path> --shop --dry-run --json

  Observed: …"output_format": "jpg", "finish": {"background": "#FFFFFF", "square": true, "pad": 7.5, "size": 2000}, "already_done": 0}

Examples

• Documented folder run with --format webp for smaller lossless files: command completed successfully

  Source: README.md Options table: --format webp|jpg

  Input: <guest-path> --format webp

  Expected: one image processed into <guest-path> as webp

  Command: <guest-path> <guest-path> -o <guest-path> --format webp -y

  Observed: by RAM) · ETA ~2s model silueta · edges matte · needs 3.0 GB free RAM auto: silueta fits your free RAM (8.1/10 on uniform images; best birefnet-general-lite 8.6) output → <guest-path> [3/4 RUN] loading model in each worker… worker 0 ready ✓ demo.webp 1280×720 1.6s [4/4 VERIFY] checking 1 outputs… GAME CLEAR · 1 done · 0 failed · 0 already done · 5s output → <guest-path>

Local or hosted: local

Remove backgrounds from every photo in a folder with the bulk-bg-remove CLI, writing a new output folder with cut-out PNGs

cd <guest-path> && pwd && ls -la && id -u && cat README.md | head -c 6000
cd <guest-path> && git log --oneline -1 2>&1; python3 --version; echo $CDP_URL; curl -s -m 10 -i $CDP_URL/json/version | head -20
cd <guest-path> && tail -c 1000 README.md; echo ----; cat .python-version; grep -n requires-python pyproject.toml; ls scripts tests .coldpass; which uv; uv --version 2>&1
cd /tmp && timeout 900 uv tool install bulk-bg-remove 2>&1 | tail -20; echo "rc=$?"; which bulk-bg-remove; bulk-bg-remove --help 2>&1 | head -60
export PATH=<guest-path>; cd /tmp && bulk-bg-remove --help 2>&1 | head -80; bulk-bg-remove --version 2>&1 | head -3
cd <guest-path> && grep -n '^version' pyproject.toml; ls assets benchmark | head -50; export PATH=<guest-path>; cd /tmp && timeout 600 bulk-bg-remove doctor 2>&1 | head -60
mkdir -p <guest-path> && cp <guest-path> <guest-path> && export PATH=<guest-path> && cd /tmp && timeout 600 bulk-bg-remove <guest-path> --dry-run --json 2>&1 | head -30; ls <guest-path>
export PATH=<guest-path> && cd /tmp && timeout 1500 bulk-bg-remove <guest-path> -y 2>&1 | tail -30; echo rc=${PIPESTATUS[0]}; find <guest-path> -type f 2>/dev/null | head
export PATH=<guest-path>; cd /tmp && bulk-bg-remove <guest-path> --shop --dry-run --json 2>&1 | tail -1 | head -c 1500; echo; cd <guest-path> && grep -rnE "while True|schedule|cron|--once|--now|sleep\(|webhook" src | head -20; ls src/*
mkdir -p <guest-path> <guest-path> && cp <guest-path> <guest-path> && ls -la <guest-path> <guest-path> && ls <guest-path> 2>&1 | head -1
cd <guest-path> && cat > probe.json <<'EOF'
{"kind":"cli","argv":["<guest-path>","<guest-path>","-o","<guest-path>","-y"],"cwd":"<guest-path>","env":{"HOME":"<guest-path>"},"output_directory":"<guest-path>","expected_text":"1 done"}
EOF
python3 -I -c "import json;print(json.load(open('probe.json'))['argv'])"
cd <guest-path> && cat > claims.json <<'EOF'
[
{"kind":"claim","claim":"Default output folder is <name>-no-bg next to the input and a real photo is cut out","source":"Output goes to a new `<name>-no-bg/` folder next to the input.","input":"<guest-path>","expected":"one image processed into <guest-path>","probe_kind":"cli","argv":["<guest-path>","<guest-path>","-y"],"env":{"HOME":"<guest-path>"},"output_directory":"<guest-path>","expected_text":"1 done","cwd":"<guest-path>"},
{"kind":"claim","claim":"--dry-run --json plans the run as JSON with the model chosen","source":"scan and plan only; one JSON event per line for agents","input":"<guest-path>","expected":"JSON plan event naming the auto-chosen model silueta","probe_kind":"cli","argv":["<guest-path>","<guest-path>","--dry-run","--json"],"env":{"HOME":"<guest-path>"},"expected_text":"\"model\": \"silueta\"","cwd":"<guest-path>"},
{"kind":"claim","claim":"--shop plans a white square 2000 px JPEG output","source":"`--shop` puts every product on white, square, 2000 px and ready to","input":"<guest-path>","expected":"plan with output_format jpg","probe_kind":"cli","argv":["<guest-path>","<guest-path>","--shop","--dry-run","--json"],"env":{"HOME":"<guest-path>"},"expected_text":"\"output_format\": \"jpg\"","cwd":"<guest-path>"}
]
EOF
python3 -I -c "import json;print(len(json.load(open('claims.json'))))"
mkdir -p <guest-path> && cp <guest-path> <guest-path> && cd <guest-path> && cat > examples.json <<'EOF'
[{"kind":"example","claim":"Documented folder run: bulk-bg-remove PATH writes PATH-no-bg, with --format webp for smaller lossless files","source":"README.md Options table: --format webp|jpg","input":"<guest-path> --format webp","expected":"one image processed into <guest-path> as webp","argv":["<guest-path>","<guest-path>","--format","webp","-y"],"cwd":"<guest-path>","env":{"HOME":"<guest-path>"},"output_directory":"<guest-path>","expected_text":"1 done"}]
EOF
cat > cycles.json <<'EOF'
{"detected":false,"reason":"Entry points inspected (README, src/bulk_bg_remove cli.py and the ui server). The product is a one-shot CLI run per folder; no cron, scheduler, --once/--now, or webhook loop was found. The only while-True loop is the local UI server's event stream (src/bulk_bg_remove/ui/server.py:343), which is not a product cycle.","checks":[]}
EOF
python3 -I -c "import json;[json.load(open(f)) for f in ['examples.json','cycles.json','claims.json','probe.json']];print('ok')"
cd /tmp && <env-var>=<guest-path> timeout 1500 python3 -I <guest-path> --validate --replay 2>&1 | tail -c 5000; echo rc=$?
cd <guest-path> && cat > claims.json <<'EOF'
[
{"kind":"claim","claim":"Each photo is cut out into a new output folder (tested with -o, since the default <name>-no-bg location is not an argv destination)","source":"Output goes to a new `<name>-no-bg/` folder next to the input.","input":"<guest-path>","expected":"one image processed into the new folder <guest-path>","probe_kind":"cli","argv":["<guest-path>","<guest-path>","-o","<guest-path>","-y"],"env":{"HOME":"<guest-path>"},"output_directory":"<guest-path>","expected_text":"1 done","cwd":"<guest-path>"},
{"kind":"claim","claim":"--dry-run --json plans the run as JSON with the model chosen","source":"scan and plan only; one JSON event per line for agents","input":"<guest-path>","expected":"JSON plan event naming the auto-chosen model silueta","probe_kind":"cli","argv":["<guest-path>","<guest-path>","--dry-run","--json"],"env":{"HOME":"<guest-path>"},"expected_text":"\"model\": \"silueta\"","cwd":"<guest-path>"},
{"kind":"claim","claim":"--shop plans a white square 2000 px JPEG output","source":"`--shop` puts every product on white, square, 2000 px and ready to","input":"<guest-path>","expected":"plan with output_format jpg","probe_kind":"cli","argv":["<guest-path>","<guest-path>","--shop","--dry-run","--json"],"env":{"HOME":"<guest-path>"},"expected_text":"\"output_format\": \"jpg\"","cwd":"<guest-path>"}
]
EOF
cat > examples.json <<'EOF'
[{"kind":"example","claim":"Documented folder run with --format webp for smaller lossless files","source":"README.md Options table: --format webp|jpg","input":"<guest-path> --format webp","expected":"one image processed into <guest-path> as webp","argv":["<guest-path>","<guest-path>","-o","<guest-path>","--format","webp","-y"],"cwd":"<guest-path>","env":{"HOME":"<guest-path>"},"output_directory":"<guest-path>","expected_text":"1 done"}]
EOF
python3 -I -c "import json;[json.load(open(f)) for f in ['examples.json','claims.json']];print('ok')"
cd /tmp && <env-var>=<guest-path> timeout 1500 python3 -I <guest-path> --validate --replay > <guest-path> 2>&1; echo rc=$?; python3 -I -c "
import json
d=json.load(open('<guest-path>'))
print('valid',d.get('valid'),d.get('problems'))
r=d.get('replay',{})
print('status',r.get('status'))
det=r.get('details',{})
print('probe rc',det.get('returncode'),det.get('argv'))
for c in det.get('claims',[]): print('CLAIM',c.get('verdict'),c.get('outcome'),c.get('evidence'))
for e in det.get('example_checks',[]): print('EX',e.get('verdict'),e.get('outcome'),e.get('evidence'))
for c in det.get('cycle_checks',[]): print('CYC',c.get('verdict'))
"
cd /tmp && grep -o '"verdict": "[a-z_]*"' <guest-path> | sort | uniq -c; grep -o '"outcome": "[^"]\{0,160\}' <guest-path> | head -8; rm -rf <guest-path>; ls <guest-path>
cd <guest-path> && cat > report.json <<'EOF'
{"outcome":"working","fault_owner":null,"operation":"Remove backgrounds from every photo in a folder with the bulk-bg-remove CLI, writing a new output folder with cut-out PNGs","operation_scope":"Linux x86_64 guest (Ubuntu, Python 3.12.3, uv 0.13.5), CPU only, 3.5 GB free RAM; downloads the silueta model (44 MB) on first run; no browser or credentials needed for the CLI path","steps":["curl -s $CDP_URL/json/version (HTTP 200, Chrome/155 shared endpoint reachable; not needed for CLI)","uv tool install bulk-bg-remove (installed 0.1.0 into <guest-path>)","bulk-bg-remove doctor (READY)","bulk-bg-remove <guest-path> --dry-run --json (plan: model silueta)","bulk-bg-remove <guest-path> -y (1 done, 0 failed; output showcase.png)","bulk-bg-remove <guest-path> -o <guest-path> -y (probe, replayed PASS)","bulk-bg-remove <guest-path> --shop --dry-run --json (output_format jpg)"],"deviations":["Installed the PyPI release (uv tool install bulk-bg-remove, version 0.1.0) as the README documents, instead of building from the pinned checkout; the workspace pyproject also says 0.1.0 but I did not diff the sources.","Input photo was assets/showcase.webp from the repo (a composite of four real photos), copied into fresh input folders.","Claim and example runs use an explicit -o output folder, because the checker requires output_directory to match an argv destination; the README default location (<name>-no-bg next to the input) was observed manually in the first run but not replayed.","The --ui browser drag-and-drop page was not exercised; the CLI path was used instead.","No Git metadata setup was needed; the workspace was used directly, not as a git checkout.","Removed my manual photos-no-bg output before replay; checker-generated output folders were left in <guest-path>"],"stuck_step":null,"main_blocker":null,"plain_failure":null,"suggested_fix":null}
EOF
python3 -I -c "import json;d=json.load(open('report.json'));print(d['outcome'], len(d['steps']))"; ls -la <guest-path>