QWED = Query with Evidence and Determinism
Thank you for your interest in contributing to QWED Documentation.
This repository documents deterministic verification systems, security boundaries, and fail-closed verification workflows. Because documentation directly affects how users understand and implement QWED systems, contributions are reviewed carefully for technical accuracy, clarity, and consistency.
| File | Why It Matters |
|---|---|
| README.md | Understand the docs repository scope and ecosystem |
| QWED Verification README | Core protocol philosophy and architecture |
| QWED_RULES.md | Canonical enforcement rules — the source of truth |
When documentation content and enforcement rules appear to conflict, QWED_RULES.md is the authoritative source for boundary behavior.
At this stage, contributions are intentionally limited while the documentation architecture stabilizes.
- Typo and grammar fixes
- Broken link fixes
- Formatting and readability improvements
- Documentation clarification requests
- Code example corrections (must be reproducible)
- API reference corrections
- Small deterministic examples
- Large architectural rewrites
- Speculative or roadmap content
- Marketing-heavy or promotional edits
- Undocumented feature additions
- Content that weakens precision or introduces ambiguity
- Probabilistic framing without explicit
HEURISTIClabeling
| Prefer | Avoid |
|---|---|
| Precise claims with explicit boundaries | Hype language or exaggerated guarantees |
| Operational details and verifiable behavior | Vague "AI magic" explanations |
| Concrete code examples with expected output | Hand-wavy descriptions |
Never describe uncertain behavior as guaranteed behavior.
If a workflow is partial, experimental, advisory, heuristic-based, or incomplete — it must be labeled clearly. Undeclared uncertainty is a documentation bug.
Use terms consistently across all documentation:
| ✅ Preferred | ❌ Avoid |
|---|---|
verified |
smart |
deterministic |
intelligent |
fail-closed |
safe enough |
unverifiable |
probably valid |
advisory |
guaranteed |
HEURISTIC (for LLM-dependent results) |
accurate |
Do not blur the boundaries between:
- validation (input format checking)
- parsing (structural extraction)
- simplification (reduction)
- heuristics (probabilistic estimation)
- verification (deterministic proof)
Do not claim:
- ❌ Perfect security
- ❌ Guaranteed correctness beyond stated boundaries
- ❌ Impossible-to-bypass protection
Documentation must accurately represent the actual implemented behavior — nothing more, nothing less.
Open an issue for:
- Incorrect or misleading documentation
- Broken code examples
- Unclear verification boundary descriptions
- Missing fail-closed behavior documentation
# 1. Fork and clone
git clone https://github.com/YOUR_USERNAME/docs.git
# 2. Create a branch
git checkout -b fix/your-fix-description
# 3. Make changes
# 4. Preview locally
mintlify dev
# 5. Commit with conventional commits
git commit -m "docs: fix deadline guard example output"
# 6. Push and create PR
git push origin fix/your-fix-descriptiontype(scope): description
docs(legal): fix statute guard response fields table
docs(finance): correct NPV formula in example
fix(changelog): update broken PR link
docs: clarify fail-closed behavior for unknown jurisdictions
- Stay scoped to a single fix or improvement
- Explain why the change is needed
- Preserve deterministic terminology
- Keep examples minimal, reproducible, and explicit about assumptions
- Follow the existing writing style and MDX conventions
- Introduce hype-oriented or marketing language
- Weaken precision of verification claims
- Add unverifiable or speculative claims
- Conflict with fail-closed philosophy
- Introduce probabilistic framing without explicit labeling
- Bypass the review process
- All non-trivial changes require maintainer review
- Automated checks (Mintlify build, link validation) must pass
- Self-merging is restricted to typo fixes by maintainers
| ✅ Preferred | ❌ Avoid |
|---|---|
| Technical, calm, direct | Aggressive marketing tone |
| Audit-friendly and precise | Exaggerated confidence |
| Operationally useful | Buzzword-heavy language |
Examples must be:
- Minimal — shortest code that demonstrates the behavior
- Deterministic — same input always produces same output
- Reproducible — reader can copy-paste and run
- Explicit — state assumptions, expected output, and edge cases
If an example is incomplete or illustrative only, label it clearly:
# ⚠️ Simplified example — production code should handle error casesThese belong in their respective repositories, not in documentation PRs:
| Don't Add Here | Where It Belongs |
|---|---|
| New guard implementations | qwed-verification, qwed-legal, qwed-finance |
| SDK code changes | Respective SDK repos |
| Enterprise features (audit, SSO, RBAC) | Separate enterprise repo |
| Telemetry / observability | Separate enterprise repo |
If you discover:
- Unsafe or dangerous code examples
- Misleading verification claims that could cause operational risk
- Security-sensitive documentation issues
- Secret exposure in examples
- Incorrect operational guidance
Do NOT open a public issue. Use the repository security policy instead.
To contribute, you must sign off on your commits:
Signed-off-by: Your Name <your@email.com>
You can sign off automatically with git commit -s.
This project is governed under a BDFL model. See the QWED Governance document for details.
By contributing, you agree that your contributions will be licensed under the Apache License 2.0.
- Open an issue with the
questionlabel - Join discussions in GitHub Discussions
- Email: rahul@qwedai.com
QWED documentation is part of the trust boundary itself. Ambiguous documentation creates operational risk, incorrect deployments, and false assumptions about verification guarantees. Clarity and correctness matter more than speed or volume.