CLI & Automation¶

The openmed console script mirrors the Python APIs so you can analyze snippets, process batches, inspect registry metadata, manage profiles, and tweak configuration without writing code. It is perfect for demos, notebooks, or CI smoke tests.

Installation¶

The CLI is installed automatically with the base package:

uv pip install .
# or pip install openmed

Commands Overview¶

PII Detection & De-identification (v0.5.0)¶

Extract and de-identify PII entities from clinical text:

# Extract PII entities
openmed pii extract \
  --text "Patient: John Doe, DOB: 01/15/1970, SSN: 123-45-6789" \
  --model pii_detection_superclinical \
  --confidence-threshold 0.6

# Extract from file
openmed pii extract \
  --input-file patient_note.txt \
  --output results.json

# De-identify with different methods
openmed pii deidentify \
  --text "Patient: John Doe, DOB: 01/15/1970" \
  --method mask \
  --output deidentified.txt

# De-identify with date shifting
openmed pii deidentify \
  --input-file note.txt \
  --method shift_dates \
  --date-shift-days 180

# Batch PII extraction
openmed pii batch-extract \
  --input-dir ./patient_notes \
  --output-dir ./pii_results

# Batch de-identification
openmed pii batch-deidentify \
  --input-dir ./notes \
  --output-dir ./deidentified \
  --method mask

De-identification methods: - mask — Replace with placeholders [NAME], [DATE], etc. - remove — Remove PII entities completely - replace — Replace with synthetic data - hash — Cryptographic hashing for linking - shift_dates — Shift dates while preserving temporal relationships

Smart Entity Merging (default): Prevents fragmentation of dates, SSN, phone numbers by merging tokenized fragments into complete entities.

See PII Detection & Smart Merging for comprehensive documentation and PII notebook for examples.

Analyze text from the terminal¶

openmed analyze \
  --model disease_detection_superclinical \
  --text "Imatinib inhibits BCR-ABL in chronic myeloid leukemia." \
  --output-format json \
  --confidence-threshold 0.6 \
  --group-entities

Flags:

• --model: registry key or HF id.
• Provide either --text or --input-file.
• --output-format: dict, json, html, or csv.
• --group-entities: toggles adjacent span merging.
• --no-confidence: omit confidence values from output.
• --use-medical-tokenizer / --no-medical-tokenizer: force on/off medical token remapping in the output (defaults to config/on).
• --medical-tokenizer-exceptions: comma-separated terms to keep intact when remapping.

Discover models¶

openmed models list --include-remote   # adds Hugging Face search
openmed models info pharma_detection_superclinical

models list prints registry models (and optionally HF remote ones). models info <key> dumps the curated metadata so you can confirm categories, entity types, recommended confidence, and model sizes.

Manage configuration¶

openmed config show
openmed config set device cuda
openmed config set default_org OpenMed
openmed config set hf_token xxx --unset    # remove a setting

• Values are stored in the CLI config file (see openmed/core/config.py for defaults). Use --config-path to point at an alternate location when scripting CI/CD workflows.
• The CLI reuses OpenMedConfig, so any changes you make here apply to both CLI runs and Python imports (assuming they load the same file).

Automation tips¶

• Wrap CLI calls in shell scripts for smoke tests. Example:

openmed models list >/tmp/models.txt
openmed analyze --model disease_detection_superclinical --text "$SAMPLE_NOTE" --output-format json > /tmp/result.json

• Use uv run or pipx run to avoid polluting system environments when scripting release pipelines.
• In GitHub Actions, you can execute the CLI after installing .[hf] to ensure models load before publishing.

For more complex workflows (structured logging, streaming), prefer the Python APIs covered in Analyze Text Helper and ModelLoader & Pipelines.

Batch processing¶

Process multiple texts or files in one command:

# Process inline texts
openmed batch --texts "Text one" "Text two" "Text three" --model disease_detection_superclinical

# Process files
openmed batch --input-files note1.txt note2.txt --output-format json

# Process a directory
openmed batch --input-dir /path/to/notes --pattern "*.txt" --recursive

# Save results to file
openmed batch --input-dir ./notes --output results.json --output-format json

# Quiet mode (suppress progress)
openmed batch --input-dir ./notes --quiet

See Batch Processing for full documentation.

Profile management¶

Manage configuration profiles from the command line:

# List all available profiles
openmed config profiles

# Show settings for a profile
openmed config profile-show dev

# Apply a profile to your configuration
openmed config profile-use prod

# Show config with a profile applied (without saving)
openmed config show --profile dev

# Save current configuration as a new profile
openmed config profile-save myprofile

# Delete a custom profile
openmed config profile-delete myprofile

See Configuration Profiles for full documentation.

Interactive TUI¶

Launch the interactive terminal interface for visual NER analysis:

# Basic launch
openmed tui

# With custom model and threshold
openmed tui --model disease_detection_superclinical --confidence-threshold 0.6

The TUI provides: - Real-time text analysis with Ctrl+Enter - Color-coded entity highlighting - Sortable entity table with confidence bars - Keyboard-driven workflow

See TUI - Interactive Terminal for full documentation.