ShivSupane/fliprate

No problems traced to this repository code in these checks.

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

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 →

• Clean messy text directly with the normalize command: Real input produced the expected result: mala product avadla nahi

  Source: fliprate normalize "mala product avadlaaa nhi"

  Input: mala product avadlaaa nhi

  Expected: mala product avadla nahi

  Command: <guest-path> normalize 'mala product avadlaaa nhi'

  Observed: mala product avadla nahi avadlaaa -> avadla nhi -> nahi

• Gate decision TRANSFORM means unstable but fixed by the built-in normalizer: Real input produced the expected result: TRANSFORM: Stable after approved normalization

  Source: | **TRANSFORM** | Unstable, but the built-in normalizer fixes it.

  Input: mala product avadlaaa nahi

  Expected: TRANSFORM: Stable after approved normalization

  Command: <guest-path> gate 'mala product avadlaaa nahi'

  Observed: TRANSFORM: Stable after approved normalization prediction='product_feedback' raw_stability=0.625 normalized_stability=1.0 flipped on: mala produkt avdla nahi flipped on: mala prodct avdla nahee flipped on: mala produuuct avadlaaa nahi

Examples

• Basic usage example runs the library quick-check, normalization and audit flow: command completed successfully

  Source: examples/01_basic_usage.py (docs/QUICKSTART.md Python API)

  Input: examples/01_basic_usage.py with no arguments

  Expected: prints a clean text, safe prediction and audit report

  Command: <guest-path> <guest-path>

  Observed: on: Prediction is stable across safe variants Messy: 'Mala product avadlaaa nhiii!!!' Clean: 'Mala product avadla nahi!' Safe prediction on 'ye delivery nhi aayi': negative Auditing model on dataset... Messages: 4 | safe variants tested: 12 Accuracy on originals: 75.0% (macro-F1 0.78) Accuracy on safe variants: 75.0% (macro-F1 0.78) Accuracy drop: +0.0% 95% CI [+0.0%, +0.0%] Flip rate: 0.0% 95% CI [0.0%, 0.0%] Most fragile change type: spelling (flip rate 0.0%) Note: flip rate shows instability, not correctness. A model can flip to the right answer.

Local or hosted: local

Install FlipRate from source and run the CLI quality tools (normalize, gate) plus the documented Python quick-start example

cd <guest-path> && pwd && ls -la && git status 2>&1 | head -3; id -u; python3 --version; cat README.md 2>/dev/null | head -150
cd <guest-path> && tail -c +7000 README.md | sed 's/\\x0d//g' | head -220
cd <guest-path> && ls -la .coldpass examples scripts tests src/fliprate 2>&1; sed -n 1,80p pyproject.toml
mkdir -p <guest-path> && python3 -m venv <guest-path> && <guest-path> install -q --upgrade pip 2>&1 | tail -3; echo rc=$?
cd <guest-path> && <guest-path> install -e ".[dev,api,sklearn]" 2>&1 | tail -5; echo rc=${PIPESTATUS[0]}
cd /tmp && F=<guest-path>; $F normalize "mala product avadlaaa nhi"; echo rc=$?; $F gate "mala product avadlaaa nahi"; echo rc=$?
cd <guest-path> && ls .github/workflows && sed 's/\r//' .github/workflows/*.yml | grep -n "run:" ; grep -n "@app\.\(get\|post\)" src/fliprate/api.py; grep -n "def \|add_parser" src/fliprate/cli.py | head -60
cd <guest-path> && sed 's/\r//' .github/workflows/ci.yml | sed -n 40,80p; sed 's/\r//' src/fliprate/api.py | sed -n 100,130p
cd <guest-path> && timeout 900 <guest-path> -m pytest -q 2>&1 | tail -15; echo rc=${PIPESTATUS[0]}
cd <guest-path> && grep -c 'x0d' README.md; grep -n 'fliprate normalize "mala product avadlaaa nhi"' README.md | cat -A | cut -c1-200; grep -n 'Unstable, but the built-in normalizer fixes it' README.md | cut -c1-200; grep -n 'fliprate gate "mala' README.md | cut -c1-120; <guest-path> perturb "delivery nahi aaya" | head -8; echo rc=$?
cd <guest-path> && ls data; grep -rniE "cron|schedul|webhook|--once|--now|while True" src/fliprate | head; cd /tmp && <guest-path> <guest-path> 2>&1 | tail -15; echo rc=${PIPESTATUS[0]}
cd <guest-path> && sed -n 1,30p src/fliprate/events.py; grep -n "def \|argv" src/fliprate/cli.py | sed -n 1,12p
cd <guest-path> && cat > probe.json <<'EOF'
{"kind":"cli","argv":["<guest-path>","normalize","mala product avadlaaa nhi"],"cwd":"<guest-path>","expected_text":"mala product avadla nahi"}
EOF
cat > claims.json <<'EOF'
[{"kind":"claim","claim":"Clean messy text directly with the normalize command","source":"fliprate normalize \"mala product avadlaaa nhi\"","input":"mala product avadlaaa nhi","expected":"mala product avadla nahi","probe_kind":"cli","argv":["<guest-path>","normalize","mala product avadlaaa nhi"],"expected_text":"mala product avadla nahi","cwd":"<guest-path>"},
{"kind":"claim","claim":"Gate decision TRANSFORM means unstable but fixed by the built-in normalizer","source":"| **TRANSFORM** | Unstable, but the built-in normalizer fixes it.","input":"mala product avadlaaa nahi","expected":"TRANSFORM: Stable after approved normalization","probe_kind":"cli","argv":["<guest-path>","gate","mala product avadlaaa nahi"],"expected_text":"TRANSFORM: Stable after approved normalization","cwd":"<guest-path>"}]
EOF
cat > examples.json <<'EOF'
[{"kind":"example","claim":"Basic usage example runs the library quick-check, normalization and audit flow","source":"examples/01_basic_usage.py (docs/QUICKSTART.md Python API)","input":"examples/01_basic_usage.py with no arguments","expected":"prints a clean text, safe prediction and audit report","argv":["<guest-path>","<guest-path>"],"cwd":"/tmp","expected_text":"Clean: 'Mala product avadla nahi!'"}]
EOF
cat > cycles.json <<'EOF'
{"detected":false,"reason":"Inspected README (CLI, API, production deployment sections), CI workflow, pyproject entry points and src/fliprate. No cron, scheduler, --once/--now, or webhook trigger exists. The only while True loop (src/fliprate/events.py drain) is a generator consumer, not a product cycle. FastAPI jobs are request-driven.","checks":[]}
EOF
for f in probe.json claims.json examples.json cycles.json; do <guest-path> -I -c "import json,sys;json.load(open(sys.argv[1]));print('ok',sys.argv[1])" $f; done
cd <guest-path> && <env-var>=<guest-path> python3 -I <guest-path> --validate --replay 2>&1 | tail -60; echo rc=${PIPESTATUS[0]}
cd <guest-path> && cat > report.json <<'EOF'
{"outcome":"working","fault_owner":null,"operation":"Install FlipRate from source and run the CLI quality tools (normalize, gate) plus the documented Python quick-start example","operation_scope":"Ubuntu guest, Python 3.12.3 (README requires >=3.9), venv at <guest-path>, extras dev,api,sklearn; no Hugging Face/torch model, no browser, no credentials needed","steps":["python3 -m venv <guest-path> && pip install --upgrade pip","pip install -e \".[dev,api,sklearn]\" (From source section, in workspace)","fliprate normalize \"mala product avadlaaa nhi\" (CLI section)","fliprate gate \"mala product avadlaaa nahi\" (CLI section)","python -m pytest -q (CI workflow, 407 passed)","python examples/01_basic_usage.py (Quick start / examples)"],"deviations":["Installed extras .[dev,api,sklearn] instead of .[dev,api,hf,sklearn] to skip the large torch/transformers download; the hf-backed model claims were not exercised.","Used a dedicated venv at <guest-path> with the absolute fliprate path instead of a global pip install, as the task requires absolute venv paths.","Workspace was not a git checkout; no git init was needed because the pinned source was already present and pip install used the local tree."],"stuck_step":null,"main_blocker":null,"plain_failure":null,"suggested_fix":"None required for this run. Optional: the README text contains CRLF line endings and the Docker/Production sections were not executed."}
EOF
<guest-path> -I -c "import json;d=json.load(open('report.json'));print(d['outcome'],d['fault_owner'])"; ls -la <guest-path>