Skip to main content
GoldenMatch includes a gold-themed terminal UI built with Textual. Launch it with goldenmatch interactive or goldenmatch dedupe (zero-config mode opens the TUI automatically).
goldenmatch interactive customers.csv
goldenmatch interactive customers.csv --config config.yaml

8 Tabs

The TUI has 8 tabs. The first 7 are accessible via keyboard shortcuts 1 through 7; use Tab to cycle through all of them, including Corrections.

Tab 1: Data

Data profiling view. Shows:
  • Record count, column names, data types
  • Null percentages, unique value counts
  • Sample values for each column
  • Auto-detected column types (name, email, phone, zip, address)

Tab 2: Config

Configuration editor. Shows:
  • Auto-detected matchkeys, blocking strategy, and golden rules
  • Edit thresholds, scorers, and weights inline
  • Save configuration to YAML
  • Run/Edit/Save buttons

Tab 3: Matches

Split-view match results:
  • Left panel: cluster list with IDs, sizes, and confidence scores
  • Right panel: golden record + individual member details for selected cluster
  • Live threshold slider: arrow keys adjust threshold in 0.05 increments with instant cluster count preview
  • Click any cluster to inspect member records and field-level scores

Tab 4: Golden

Golden record viewer:
  • Canonical merged records
  • Merge strategy applied to each field
  • Source provenance for each value
  • Export golden records to CSV

Tab 5: Boost

Active learning for accuracy improvement:
  • Shows borderline pairs (near the threshold)
  • Label with keyboard: y (match), n (no match), s (skip)
  • After labeling ~10 pairs, trains a LogisticRegression classifier
  • Reclassifies all borderline pairs with the trained model
  • Shows before/after accuracy comparison

Tab 6: Export

Output options:
  • Export golden records, duplicates, unique records
  • Choose format: CSV or Parquet
  • Export lineage JSON
  • Generate HTML report

Tab 7: Controller (v1.7-v1.12)

AutoConfigController telemetry from the most recent Ctrl+A invocation:
  • Health verdict badge (green / yellow / red)
  • StopReason with one-line hover explanation
  • Committed matchkeys with Path Y · N NE indicator + expandable NE field rows
  • Complexity profile cells (pairs, above-threshold, borderline, blocks, p99, clusters, transitivity)
  • Indicator column priors table (identity / corruption scores)
  • Refit decisions table (iter / rule / rationale / wall-clock)
Same JSON shape as the web /api/v1/controller/telemetry endpoint and the CLI goldenmatch autoconfig output.

Tab 8: Corrections

Learning Memory corrections browser:
  • Table of stored corrections (pair, decision, source, trust, dataset)
  • Live filter by dataset
  • r to refresh, d to delete the selected correction
  • Writes happen through the Golden and Matches correction modals

Keyboard shortcuts

KeyAction
17Jump to tab (Data, Config, Matches, Golden, Boost, Export, Controller)
F5Run the pipeline
Ctrl+AAuto-configure: run AutoConfigController; adopt config; switch to Controller tab
?Show all keyboard shortcuts
Ctrl+EExport results
Left/RightAdjust threshold (Matches tab)
yLabel pair as match (Boost tab)
nLabel pair as non-match (Boost tab)
sSkip pair (Boost tab)
qQuit

Pipeline progress

On first run, the TUI shows a full-screen progress view:
  • Stage tracker with status indicators (done / running / pending)
  • Progress bar for the current stage
  • Record count and timing
On re-runs (e.g., after threshold adjustment), progress appears in the footer bar.

Auto-config summary

After auto-configure runs (Ctrl+A), a review screen shows:
  • Detected columns and their inferred types
  • Suggested scorers for each column
  • Proposed blocking strategy
  • Three options: Run (accept and run), Edit (modify config), Save (write YAML)

TUI with config file

goldenmatch interactive customers.csv --config config.yaml
When a config file is provided, the TUI skips auto-detection and uses the specified matchkeys, blocking, and golden rules.

Screenshots

The TUI uses a dark theme with gold accents. Key views:
  • Data tab: column profiling with null rates and type detection
  • Matches tab: split-view cluster browser with threshold slider
  • Golden tab: merged records with merge provenance