Cursor Rules for DynoAI

Automatic Model Selection

Sonnet 4.5 Max

Use for:

Any changes to math-critical files:

```
core/ai_tuner_toolkit_dyno_v1_2.py
```
```
core/ve_operations.py
```
```
core/io_contracts.py
```
```
experiments/protos/*.py
```

Repository-wide refactoring operations
Structural changes affecting multiple subsystems
Safety-critical operations (VE apply/rollback logic, clamping enforcement)
Complex debugging involving tuning algorithms

Sonnet 4

Use for:

Normal feature development
Test implementation and updates
Documentation improvements
Web service API development
Script and automation work
Bug fixes in non-critical code

Haiku

Use for:

Single-file trivial edits
Typo fixes
Comment updates
Formatting-only changes
Simple variable renames within one file

Safety Guardrails

Confirmation Required

Deleting any directory
Modifying more than 25 files in a single operation
Changes to
```
core/
```
,
```
experiments/protos/
```
, or test harnesses
Dependency updates (requirements.txt)
CI/CD workflow changes

Read-Only by Default

```
archive/
```
- Historical code and artifacts (reference only)
```
experiments/
```
- Research kernels and experiment outputs (unless explicitly instructed)
```
tables/
```
- Calibration base tables (unless performing VE operations)

Protected Operations

Never modify tuning math or algorithms without maintainer approval
Never weaken test assertions to "make tests pass"
Never skip safety tests (selftests, kernel harnesses, pytest)
Never commit secrets, API keys, or credentials

Testing Expectations

After Modifying Runtime Code

Run in this order:

python tests/selftest.py
python tests/acceptance_test.py
python tests/kernels/test_k1.py
python tests/kernels/test_k2.py
python tests/kernels/test_k2_fixed.py
python tests/kernels/test_k3.py
python -m pytest tests/unit tests/integration -v

For Structural Changes

pwsh scripts/dynoai_safety_check.ps1 -RunSelftests -RunKernels -RunPytest

Before Committing

All tests must pass
No new linter errors introduced
Documentation updated if behavior changed

Code Review Standards

Math-Critical Changes

Require design document
Require validation plan with before/after comparisons
Require maintainer approval
Extended review period (3-7 days minimum)
Must include test evidence in PR

Standard Changes

Self-review completed
Tests added for new functionality
Documentation updated
CI passes

Documentation

When to Update

Always: New CLI flags, API endpoints, or workflows
Always: Changes to test requirements or safety policies
Usually: Bug fixes affecting user-facing behavior
Sometimes: Refactoring (if it impacts usage)

Key Documents

```
README.md
```
- Overview, quick start, troubleshooting
```
docs/DYNOAI_CORE_REFERENCE.md
```
- Minimal runnable examples
```
docs/DYNOAI_SAFETY_RULES.md
```
- Safety policies
```
docs/DYNOAI_ARCHITECTURE_OVERVIEW.md
```
- System architecture

Common Patterns

Adding a New Test

# tests/unit/test_new_feature.py
from pathlib import Path
import pytest

ROOT = Path(__file__).resolve().parents[2]  # Repo root

def test_new_feature():
    """Test description."""
    # Test implementation
    assert True

Adding a New CLI Flag

Update
```
core/ai_tuner_toolkit_dyno_v1_2.py
```
argparse section
Add test coverage in
```
tests/selftest.py
```
or new unit test
Update
```
docs/DYNOAI_CORE_REFERENCE.md
```
with example
Update
```
README.md
```
if user-facing

Adding a New Experimental Kernel

Create
```
experiments/protos/kN_description_v1.py
```
Implement
```
kernel_smooth()
```
function matching signature
Add to
```
experiments/run_experiment.py
```
IDEAS dict
Create
```
tests/kernels/test_kN.py
```
harness
Document in experiment notes

Emergency Procedures

Rollback a Bad Commit

git revert <commit-hash>
git push origin main

Restore VE Table After Bad Apply

from pathlib import Path
from core.ve_operations import VERollback

VERollback().rollback(
    current_ve_path=Path("tables/VE_Front_Bad.csv"),
    metadata_path=Path("tables/VE_Front_Bad_meta.json"),
    output_path=Path("tables/VE_Front_Restored.csv"),
)

Recover from Failed Reorganization

# If backup exists
Remove-Item -Recurse -Force DynoAI_2
Copy-Item -Recurse DynoAI_2_backup_<timestamp> DynoAI_2

By following these rules, we maintain DynoAI's tuning accuracy, customer safety, and code quality standards.

Cursor Rules for DynoAI

Automatic Model Selection

Sonnet 4.5 Max

Use for:

Any changes to math-critical files:

```
core/ai_tuner_toolkit_dyno_v1_2.py
```
```
core/ve_operations.py
```
```
core/io_contracts.py
```
```
experiments/protos/*.py
```

Repository-wide refactoring operations
Structural changes affecting multiple subsystems
Safety-critical operations (VE apply/rollback logic, clamping enforcement)
Complex debugging involving tuning algorithms

Sonnet 4

Use for:

Normal feature development
Test implementation and updates
Documentation improvements
Web service API development
Script and automation work
Bug fixes in non-critical code

Haiku

Use for:

Single-file trivial edits
Typo fixes
Comment updates
Formatting-only changes
Simple variable renames within one file

Safety Guardrails

Confirmation Required

Deleting any directory
Modifying more than 25 files in a single operation
Changes to
```
core/
```
,
```
experiments/protos/
```
, or test harnesses
Dependency updates (requirements.txt)
CI/CD workflow changes

Read-Only by Default

```
archive/
```
- Historical code and artifacts (reference only)
```
experiments/
```
- Research kernels and experiment outputs (unless explicitly instructed)
```
tables/
```
- Calibration base tables (unless performing VE operations)

Protected Operations

Never modify tuning math or algorithms without maintainer approval
Never weaken test assertions to "make tests pass"
Never skip safety tests (selftests, kernel harnesses, pytest)
Never commit secrets, API keys, or credentials

Testing Expectations

After Modifying Runtime Code

Run in this order:

python tests/selftest.py
python tests/acceptance_test.py
python tests/kernels/test_k1.py
python tests/kernels/test_k2.py
python tests/kernels/test_k2_fixed.py
python tests/kernels/test_k3.py
python -m pytest tests/unit tests/integration -v

For Structural Changes

pwsh scripts/dynoai_safety_check.ps1 -RunSelftests -RunKernels -RunPytest

Before Committing

All tests must pass
No new linter errors introduced
Documentation updated if behavior changed

Code Review Standards

Math-Critical Changes

Require design document
Require validation plan with before/after comparisons
Require maintainer approval
Extended review period (3-7 days minimum)
Must include test evidence in PR

Standard Changes

Self-review completed
Tests added for new functionality
Documentation updated
CI passes

Documentation

When to Update

Always: New CLI flags, API endpoints, or workflows
Always: Changes to test requirements or safety policies
Usually: Bug fixes affecting user-facing behavior
Sometimes: Refactoring (if it impacts usage)

Key Documents

```
README.md
```
- Overview, quick start, troubleshooting
```
docs/DYNOAI_CORE_REFERENCE.md
```
- Minimal runnable examples
```
docs/DYNOAI_SAFETY_RULES.md
```
- Safety policies
```
docs/DYNOAI_ARCHITECTURE_OVERVIEW.md
```
- System architecture

Common Patterns

Adding a New Test

# tests/unit/test_new_feature.py
from pathlib import Path
import pytest

ROOT = Path(__file__).resolve().parents[2]  # Repo root

def test_new_feature():
    """Test description."""
    # Test implementation
    assert True

Adding a New CLI Flag

Update
```
core/ai_tuner_toolkit_dyno_v1_2.py
```
argparse section
Add test coverage in
```
tests/selftest.py
```
or new unit test
Update
```
docs/DYNOAI_CORE_REFERENCE.md
```
with example
Update
```
README.md
```
if user-facing

Adding a New Experimental Kernel

Create
```
experiments/protos/kN_description_v1.py
```
Implement
```
kernel_smooth()
```
function matching signature
Add to
```
experiments/run_experiment.py
```
IDEAS dict
Create
```
tests/kernels/test_kN.py
```
harness
Document in experiment notes

Emergency Procedures

Rollback a Bad Commit

git revert <commit-hash>
git push origin main

Restore VE Table After Bad Apply

from pathlib import Path
from core.ve_operations import VERollback

VERollback().rollback(
    current_ve_path=Path("tables/VE_Front_Bad.csv"),
    metadata_path=Path("tables/VE_Front_Bad_meta.json"),
    output_path=Path("tables/VE_Front_Restored.csv"),
)

Recover from Failed Reorganization

# If backup exists
Remove-Item -Recurse -Force DynoAI_2
Copy-Item -Recurse DynoAI_2_backup_<timestamp> DynoAI_2

By following these rules, we maintain DynoAI's tuning accuracy, customer safety, and code quality standards.

Cursor Rules for DynoAI

Related Skills

Markdown Converter

Nano Banana Pro

1password