Skip to content

FLSkill v0.1.0-alpha

Pre-release
Pre-release

Choose a tag to compare

@ryurikoneko ryurikoneko released this 26 Sep 06:04
· 163 commits to main since this release

FLSkill v0.1.0-alpha

The first clean public alpha of FLSkill.

FLSkill is an experimental framework for building verifiable, resumable AI-assisted music-production workflows for FL Studio.

Instead of treating a DAW operation as successful merely because an agent attempted it, FLSkill is designed around a stricter principle:

Plan → Write → Read Back → Verify → PASS / STOP

This first public release intentionally contains only the clean, DAW-independent core.

What's included

Musical Grid

Represent musical positions using deterministic bar / beat / tick coordinates.

Absolute Tick Resolution

Convert musical positions into absolute timeline locations that can later be mapped to a DAW.

Note Plan

Structured note events containing pitch, start position, duration and velocity.

Writer / Reader abstraction

A generic boundary for future DAW adapters. The core does not require FL Studio, Computer Use, MCP or SysEx.

Exact-Set Verification

After writing events, FLSkill reads the resulting event set back and compares it with the plan. Duplicate-event counts are preserved, and mismatches are diagnosed instead of being treated as success.

PASS / STOP semantics

Write failures, read failures, missing events, extra events or mismatched event fields result in STOP.

Verification status

Offline Algorithm Verified

The release was published after:

  • 14 passing offline tests
  • Python compile checks
  • JSON Schema checks
  • clean public provenance tracking

Important:

Offline Algorithm Verified ≠ FL Studio Verified

This release does not yet claim successful live FL Studio operation.

Why FLSkill exists

Most AI + DAW experiments focus on whether an agent can perform an action. FLSkill focuses on a second question:

Can the system prove that the musical operation it intended to perform is what actually happened in the target?

The long-term architecture separates composition planning, musical timing, note-event planning, execution adapters, state readback, verification and recovery/resume workflows.

Planned execution paths

Future optional adapters may include:

  • Native Computer Use
  • Community FL Studio MCP integration
  • FLSkill SysEx RPC
  • compatibility Computer Use bridges

These are intended to remain adapters around FLSkill Core rather than mandatory dependencies.

Roadmap

Planned work includes:

  • live FL Studio note write/readback verification
  • multi-channel / multi-pattern workflows
  • SysEx RPC
  • Mixer control with readback verification
  • plugin parameter read/write/readback
  • Computer Use and community MCP adapters
  • audio analysis and section-level composition workflows

See docs/ROADMAP.md for the current development plan.

Provenance

The public repository is a clean implementation of the FLSkill core. Experimental third-party integrations and unknown-provenance implementation files from the private development environment are not included.

See PROVENANCE.md for file-level provenance.

Current limitations

v0.1.0-alpha does not yet provide:

  • live FL Studio control
  • SysEx RPC
  • Computer Use integration
  • community FL Studio MCP adapter
  • Mixer control
  • plugin parameter control
  • autonomous song production

Those capabilities remain on the roadmap.

Download

The GitHub release source archives contain the exact v0.1.0-alpha tagged source tree.

Licensed under the MIT License.