triz-solver

name	triz-solver
description	Generic TRIZ solver for resolving architectural and technical contradictions using inventive principles.

When to use

Use after triz-analysis has identified technical or physical contradictions. The solver provides a structured way to find innovative solutions beyond simple compromises.

Inventive Principles for Software/Rust

1. Segmentation: Divide into independent parts (crates, async tasks, plugins).
2. Extraction: Remove the "disturbing" part (isolate side effects, move hot loops to SIMD).
3. Local Quality: Make each part of the system function under optimal conditions (custom thresholds per audio profile).
4. Asymmetry: Use asymmetric structures for asymmetric problems (different processing for speech vs. silence).
5. Prior Action: Perform required changes before they are needed (pre-allocate buffers, pre-compute spectral features).
6. Feedback: Introduce feedback to improve the process (self-learning calibration).
7. Taking Out: Only keep the essential part (shared types crate).

Solver Examples (Project-Specific)

1. VAD Throughput vs. False Positives

• Contradiction: High speed vs. high accuracy.
• TRIZ Solution (Segmentation + Extraction):
  • Use a multi-pass VAD.
  • Pass 1: Extremely fast, coarse energy-based VAD to extract potential speech regions.
  • Pass 2: More expensive, high-accuracy analysis (spectral flux, entropy) only on the regions identified in Pass 1.
• TRIZ Solution (Local Quality):
  • Apply noise reduction/pre-filtering only to regions with low signal-to-noise ratio.

2. Review Quality vs. Pipeline Latency

• Contradiction: High data detail vs. fast feedback.
• TRIZ Solution (Prior Action):
  • Generate a "Draft" interactive report as soon as VAD is done.
  • Stream high-resolution spectral data in the background to update the report incrementally.
• TRIZ Solution (Local Quality):
  • Only compute high-resolution spectral features for segments where VAD confidence is low or a "Review Required" tag is applied.

3. Deterministic Output vs. Self-Learning

• Contradiction: Reproducibility vs. Adaptation.
• TRIZ Solution (Asymmetry):
  • Keep the core audio pipeline deterministic by taking all parameters as input.
  • Store the "Learned" parameters in a sidecar SQLite/libsql database (learning.db).
  • Provide a "Synchronize" step that updates the static config from the learned database periodically.
• TRIZ Solution (Feedback):
  • Use the verified_segments table to continuously refine the spectral thresholds, but apply them in discrete, versioned "Profile" updates.

4. Modularity vs. CI/Runtime Complexity

• Contradiction: Clean organization vs. Build overhead.
• TRIZ Solution (Taking Out):
  • Extract minimal, common data structures into a zero-dependency types crate to prevent dependency bloat in high-level modules.
• TRIZ Solution (Homogeneity):
  • Use standardized error handling (anyhow, thiserror) and logging across all modules to reduce cognitive load and boilerplate.

Output

All TRIZ solver outcomes must be written to analysis/triz/ with a descriptive filename (e.g., analysis/triz/feature-x-solutions.md).

Related Skills

• triz-analysis: For identifying the contradictions to solve.
• triz-audio-timeline: Domain-specific TRIZ for audio segmentation.