MVP Rocket 🚀

Two-part MVP acceleration toolkit: from messy LLM output to strict types, and from codebase to defensible launch recommendation.

Zero dependencies — Python 3.10+ stdlib only.

---

Part 1: JSON Repair & Schema Inference

repair_json.py

Extracts and repairs malformed JSON from messy text (LLM output, scraped pages, logs).

# Repair broken JSON from an LLM
echo "{name: 'test', active: True, count: 0xFF, tags: ('a' 'b'), score: .95,}" \
  | python scripts/repair_json.py -

# Output: {"name": "test", "active": true, "count": 255, "tags": ["a", "b"], "score": 0.95}

What It Fixes

Category	Examples
Markdown code fences	```json ... ```
Smart quotes	"curly" → "straight"
Comments	// line and /* block */
Single-quoted strings	'hello' → "hello"
Unquoted object keys	{name: "val"} → {"name": "val"}
Python/JS literals	True/False/None/undefined/NaN/Infinity
Trailing commas	{"a": 1,} and multiple commas
Truncated output	Unclosed strings, objects, arrays auto-closed
Hex/octal numbers	0xFF → 255, 0o77 → 63
Non-standard floats	.5 → 0.5, 5. → 5.0
Missing commas	"a" "b" → "a", "b"
Python tuples	(1, 2, 3) → [1, 2, 3]
UTF-8 BOM	Stripped automatically

CLI

python scripts/repair_json.py raw.txt                  # repair → stdout
python scripts/repair_json.py raw.txt --compact         # minified
python scripts/repair_json.py raw.txt --indent 4        # 4-space indent
python scripts/repair_json.py raw.txt -o clean.json     # write to file
python scripts/repair_json.py raw.txt -i                # in-place
python scripts/repair_json.py raw.txt --check           # exit 0=valid, 1=needs repair
cat raw.txt | python scripts/repair_json.py -           # stdin

Library API

import repair_json
data = repair_json.loads(messy_text)
data, stages = repair_json.repair(messy_text)

---

json_to_schema.py

Analyzes sample JSON and generates strict type definitions.

python scripts/json_to_schema.py sample.json --name User                          # Zod (default)
python scripts/json_to_schema.py sample.json --name User --lang pydantic          # Pydantic v2
python scripts/json_to_schema.py sample.json --name User --lang jsonschema        # JSON Schema 2020-12
python scripts/json_to_schema.py sample.json --name User --lang typescript        # TypeScript interfaces

Features

• Format detection: UUID, email, URL, ISO date, ISO datetime
• Enum inference: low-cardinality strings → z.enum() / Literal[...]
• Nullable handling: null + Type → .nullable() / Type | None
• Multi-sample merging: feed multiple records to discover optional fields
• Loose mode: .passthrough() / extra="allow" / additionalProperties

The Full Pipeline

cat llm_output.txt | python scripts/repair_json.py - --compact \
  | python scripts/json_to_schema.py - --name Response

---

Part 2: Release Auditor

Scans an MVP codebase for release blockers and produces a GO / CONDITIONAL GO / NO-GO recommendation.

Operating Modes

Mode	Behavior
audit-only (default)	Inspect and report. No code changes.
audit-and-fix	Diagnose, fix, measure before/after, retest.
verify-fixes	Retest previous findings + regression check.

What It Covers

Category	Examples
Functional	Auth, CRUD, forms, navigation, edge cases, integrations
Performance	Core Web Vitals, API latency, bundle size, memory, queries
Security	Secrets, auth, XSS, injection, CORS, headers, dependencies
Accessibility	Keyboard, focus, contrast, semantics, screen readers
Cross-browser	Chromium, Firefox, WebKit/Safari
Responsive	320px → 1920px, overflow, layout shifts
Reliability	Failure handling, retries, idempotency, data consistency
Usability	First-time UX, error recovery, empty states
Code Review	Architecture, error handling, type safety, deployment

Audit Scripts

# Scan for hardcoded secrets (API keys, passwords, tokens)
python scripts/audit_secrets.py <directory> --json --gitignore

# Scan for TODO/FIXME/HACK comments with severity
python scripts/audit_todos.py <directory> --json --severity

Reference Documents

• references/test-checklists.md — 200+ test items across 7 categories
• references/performance-methodology.md — metrics, targets, root cause analysis
• references/severity-and-release-gates.md — severity definitions + GO/NO-GO criteria
• references/report-template.md — structured 18-section report template

---

Testing

pip install pytest
python -m pytest tests/ -v          # 67 unit tests
python test_journeys.py             # 27 E2E journey tests

License

MIT