rew-workflows

name

rew-workflows

description

This skill should be used when the user asks to "measure my room", "calibrate my monitors", "analyze my measurements", "optimize speaker placement", "run a calibration", "check my levels", "do a measurement session", "compare before and after", "what should I measure next", or when working with REW MCP server tools and needing guidance on workflow sequencing, tool chaining, and best practices for room measurement campaigns.

REW MCP Workflows

This skill encodes the canonical tool-chaining sequences, prompt invocations, and resource access patterns for the REW MCP server. Apply these patterns whenever orchestrating room measurement, calibration, analysis, or optimization tasks.

Tool Overview

The REW MCP server exposes tools across six categories. Each tool follows an action-based input pattern and returns structured JSON with status, data, and error fields.

Setup

• rew_api_connect -- Establish connection to a running REW instance. Default port 4735. Must call before any other API tool. Returns REW version, measurement count, and capability flags (pro_features, blocking_mode).
• rew_api_audio -- Configure audio devices. Actions: status, list_devices, set_input, set_output, set_sample_rate. Use to select the measurement microphone and output device before measuring.
• rew_api_check_levels -- Read input levels and classify into zones: CLIPPING, HOT, OPTIMAL (-20 to -10 dBFS), LOW, VERY_LOW. Blocks measurement for CLIPPING or VERY_LOW zones. Provides per-channel RMS/peak and L/R mismatch detection (>3 dB triggers warning).

Calibration

• rew_api_calibrate_spl -- Semi-automated SPL calibration. Three-phase state machine: start (play pink noise at -20 dBFS + start SPL meter with C-weighting Slow), check (read current SPL, calculate adjustment from target, generate guidance), stop (stop generator and meter). Default target: 85 dB. Tolerance: 1.0 dB.
• rew_api_generator -- Signal generator control. Actions: status, start, stop, set_signal, set_level, set_frequency, list_signals. Generates pink noise, sweeps, tones.

Measurement

• rew_api_measure -- Direct measurement control. Actions: status, sweep, spl, cancel, configure. Sweep measurements require REW Pro license.
• rew_api_measurement_session -- Guided L/R/Sub measurement sequence with persistent session state. Actions: start_session (create new session, get UUID), measure (trigger measurement for a channel: left/right/sub), get_status (check progress or list active sessions), stop_session. Enforces sequence order: left -> right -> sub -> complete. Auto-names measurements with session prefix.
• rew_api_measure_workflow -- Orchestrated measurement workflow. Actions: setup (auto-configure devices), check_levels, calibrate_level, measure (single sweep), measure_sequence (L/R or multi-position). Handles device selection, blocking mode, and result retrieval.

Analysis

• rew_analyze_room -- Unified analysis combining peaks/nulls, room modes (requires dimensions), sub integration (requires sub measurement), L/R symmetry (requires L+R measurements), and GLM calibration transparency (full with pre+post, heuristic with post-only). Returns top 5 prioritized recommendations scored 60% fixability + 40% severity. Always use this as the primary analysis entry point.
• rew_analyze_room_modes -- Standalone room mode analysis. Correlates detected peaks with theoretical axial/tangential/oblique modes from room dimensions.
• rew_analyze_decay -- ISO 3382 decay analysis (T20/T30/EDT) from impulse response data. Identifies frequencies with excessive ringing.
• rew_analyze_impulse -- Early reflection detection and path estimation from impulse data.
• rew_compare_measurements -- Compare 2-10 measurements. Types: before_after, placement_comparison, lr_symmetry, with_without_sub. Returns per-band analysis with improvement scores.
• rew_compare_to_target -- Compare a measurement against target curves: flat, rew_room_curve (LF rise + HF roll), harman, or custom. Returns deviation stats.

Optimization

• rew_optimize_room -- Iterative optimization guidance. Three actions:
  • get_recommendation -- Return the single highest-priority placement or settings recommendation. One at a time for scientific rigor.
  • validate_adjustment -- Compare pre/post measurement at a target frequency to classify result as success (>50% reduction), partial, unchanged, or worsened.
  • check_progress -- Evaluate zone-based success criteria (smoothness 40-200 Hz, L/R balance, sub integration). Returns should_stop flag when smoothness reaches "good" zone.

Data

• rew_ingest_measurement -- Parse REW text export (frequency response or impulse response) into internal store. Requires speaker_id (L/R/Sub/etc.) and condition metadata.
• rew_api_get_measurement -- Fetch a measurement from REW by UUID (not index -- indices shift).
• rew_api_list_measurements -- List all measurements in the connected REW instance.
• rew_api_import -- Import measurement data into REW. Actions: frequency_response_file, frequency_response_data, impulse_response_file, impulse_response_data, rta_file, sweep_recording.

GLM Integration

• rew_interpret_with_glm_context -- Interpret analysis results considering Genelec GLM capabilities and limitations. Classifies corrections as successfully applied, beyond scope, or residual. Supports GLM 3 and 4 version-specific behavior. Provides system readiness verdict.

Advanced

• rew_average_measurements -- Spatial averaging from multiple positions. Methods: RMS (incoherent, recommended for spatial averaging), Vector (coherent, requires phase), or hybrid.
• rew_api_spl_meter -- SPL meter control. Actions: start, stop, read, configure. Supports A/C/Z weighting, Slow/Fast/Impulse response.
• rew_api_rta -- Real-time analyzer. Actions: start, stop, capture, reset, configure, read_levels, read_captured, read_distortion.
• rew_api_eq -- Global EQ management. Actions: list_equalisers, list_manufacturers, get_defaults, set_defaults, get_house_curve, set_house_curve.
• rew_api_measurement_eq -- Per-measurement EQ. Actions: get_equaliser, set_equaliser, get_filters, set_filters, get_target, set_target, predicted_response, filter_response, match_target.
• rew_api_measurement_commands -- Per-measurement commands. Actions: list_commands, execute.
• rew_api_groups -- Measurement group management. Actions: list, create, get, update, delete, list_measurements, add_measurement, remove_measurement.

Workflow Sequences

1. Full Calibration (first-time setup)

The end-to-end workflow for a new studio or after major changes. Follow this exact sequence:

rew_api_connect
  -> rew_api_audio (list_devices, set_input, set_output)
  -> rew_api_check_levels
  -> rew_api_calibrate_spl (start -> check loop -> stop)
  -> rew_api_measurement_session (start_session)
  -> rew_api_measurement_session (measure: left)
  -> rew_api_measurement_session (measure: right)
  -> rew_api_measurement_session (measure: sub)
  -> rew_analyze_room (with L/R/Sub measurement IDs + dimensions)
  -> rew_interpret_with_glm_context (if Genelec monitors)
  -> rew_optimize_room (get_recommendation -> validate -> check_progress loop)

Pause for user input before playing audio, before each measurement, and before each physical adjustment. Invoke the rew_calibration_full prompt to activate this workflow with embedded guidance.

2. Quick Re-measure (after physical change)

Use when re-measuring after moving speakers, adding treatment, or changing position:

rew_api_connect (if not already connected)
  -> rew_api_measurement_session (start_session)
  -> rew_api_measurement_session (measure: left/right/sub as needed)
  -> rew_analyze_room
  -> rew_compare_measurements (before_after with previous session's measurements)
  -> rew_optimize_room (validate_adjustment on the specific issue addressed)

3. Analysis Only (existing measurements)

For offline analysis of previously exported REW data:

rew_ingest_measurement (for each exported file)
  -> rew_analyze_room (with ingested measurement IDs)
  -> rew_compare_to_target (against flat, room curve, or Harman)
  -> rew_interpret_with_glm_context (if applicable)

Or for measurements already in REW:

rew_api_connect
  -> rew_api_list_measurements
  -> rew_api_get_measurement (for each relevant measurement)
  -> rew_analyze_room

4. Optimization Cycle (iterative improvement)

The measure-adjust-validate loop. Never skip validation:

rew_optimize_room (action: get_recommendation)
  -> [User makes physical change]
  -> rew_api_measurement_session (measure the affected channel)
  -> rew_optimize_room (action: validate_adjustment, with pre/post IDs + target freq)
  -> rew_optimize_room (action: check_progress)
  -> [Repeat until should_stop or user satisfied]

Stop conditions: smoothness reaches "good" zone, user is satisfied, or diminishing returns detected.

MCP Prompts

Prompts provide pre-built conversational workflows. Invoke them to activate structured multi-step guidance with built-in checkpoints.

• rew_calibration_full -- Complete end-to-end calibration workflow. Optional args: target_spl_db (default 85), room_dimensions (format "LxWxH" in feet). Covers gain staging through optimization.
• rew_gain_staging -- Standalone SPL calibration using pink noise and SPL meter. Optional arg: target_spl_db. Does not require or create a measurement session.
• rew_measurement_workflow -- Session-aware L/R/Sub measurement sequence. Required arg: session_id. Embeds session state as a resource. Tracks remaining measurements and provides step-by-step guidance.
• rew_optimization_workflow -- Iterative placement optimization with measurement validation. Required arg: session_id (must have completed measurements). Embeds session state. Enforces scientific method: one recommendation, measure, validate, repeat.

MCP Resources

Resources provide read-only access to session and measurement state via URI templates.

• session://{session_id} -- Current session state: sequence step, completed measurements (with UUIDs, channels, timestamps), target SPL, notes.
• measurement://{measurement_id} -- Full measurement data including frequency response arrays (frequencies_hz, spl_db).
• recommendations://{session_id} -- Active recommendations for a session from analysis and optimization tools.
• history://{session_id} -- Measurement history and summaries for a session.

Access resources to maintain context across tool calls without re-running analysis. Session resources update automatically as measurements are taken.

Best Practices

Pre-Measurement Checklist

• Always call rew_api_connect first. Every other API tool depends on it and returns a connection error without it.
• Always run rew_api_check_levels before measuring. Do not proceed if the zone is CLIPPING or VERY_LOW.
• Calibrate SPL to 85 dB (broadcast reference) before any measurements. This ensures consistent level reference across sessions.
• Confirm the correct input device (measurement mic) and output device via rew_api_audio before measuring.

Measurement Discipline

• Measure L, R, and Sub separately -- never all at once. The session tool enforces left -> right -> sub ordering.
• Use UUID-based measurement references, not indices. Indices shift when measurements are added or removed in REW.
• Auto-naming uses {session_prefix}_{channel} format. Do not override unless the user requests custom naming.
• REW Pro license is required for automated sweep measurements. If a 403 error occurs, inform the user about the license requirement.

Analysis Strategy

• Use rew_analyze_room as the primary analysis entry point. It combines peaks/nulls, room modes, sub integration, L/R symmetry, and GLM comparison into a single prioritized result. Do not call individual analysis tools (analyze_room_modes, analyze_decay, etc.) unless investigating a specific isolated issue.
• Provide room dimensions when available -- they enable theoretical mode correlation, which significantly improves recommendation quality.
• If both pre-GLM and post-GLM measurements exist, pass both to analyze_room for full GLM comparison rather than heuristic mode.

Optimization Discipline

• Use rew_optimize_room for one-at-a-time recommendations. Never attempt to fix multiple issues simultaneously -- changes interact and prior measurements become invalid.
• Always validate adjustments with new measurements. Never assume an adjustment helped without measuring.
• Compare before/after for every physical change using rew_compare_measurements with before_after type.
• Prioritize placement and settings adjustments (free, high impact) before recommending acoustic treatment (cost, variable impact).
• Stop optimization when check_progress returns should_stop: true or when improvements fall below 1 dB per iteration.

Error Recovery

• Connection refused: REW is not running or API is not enabled. Direct the user to enable the API in REW Preferences or launch with the -api flag.
• Timeout: REW may be busy processing. Wait and retry once. If persistent, check if REW is frozen.
• Level clipping: Reduce mic gain or generator level before retrying. Never proceed with clipped input.
• Measurement fails with 403: REW Pro license required. Cannot be worked around for automated sweeps.
• Session not found: Sessions are in-memory only. If the server restarted, create a new session.

For detailed tool chaining patterns, parameter examples, and error recovery sequences, see references/tool-chaining.md.