​
One batch

bash scripts/run_pipeline.sh -i input/batch1 -o outputs/batch1
The default is: 
--pipeline s_core_relations_smoke_drug
--writers sqlite
--offline
The plain run writes one database file. It does not create per-note CSV/XMI/HTML folders unless those writers are requested.

​
Multiple batches

If the input folder looks like this: 
input/
  batch1/
  batch2/
  batch3/
run: 
bash scripts/run_input.sh --input-root input --output-root outputs/site_run --foreground
That writes: 
outputs/site_run/batch1/
outputs/site_run/batch2/
outputs/site_run/batch3/
run_input.sh defaults to background mode. Use --foreground when a clinician or reviewer wants to watch the run in the terminal.

​
Normal database run

bash scripts/run_pipeline.sh -i input/batch1 -o outputs/batch1
Output: 
outputs/batch1/annotations.sqlite
outputs/batch1/run_manifest.json
outputs/batch1/run_pipeline.log
outputs/batch1/run_pipeline.effective.piper
outputs/batch1/run_dictionary.effective.xml

​
CSV and XMI run

Use this only when someone specifically asks for folder outputs: 
bash scripts/run_pipeline.sh -i input/batch1 -o outputs/batch1_standard --writers standard
Output: 
outputs/batch1_standard/concepts/
outputs/batch1_standard/rxnorm/
outputs/batch1_standard/cui_counts/
outputs/batch1_standard/xmi/
outputs/batch1_standard/run_manifest.json
standard keeps CSV and XMI on, but keeps HTML off.

​
Database plus CSV/XMI

bash scripts/run_pipeline.sh -i input/batch1 -o outputs/batch1_standard_sqlite --writers standard --sqlite
That keeps the CSV/XMI folders and also writes: 
outputs/batch1_standard_sqlite/annotations.sqlite

​
HTML views

HTML is off by default. Use this only for a small debug or demo run: 
bash scripts/run_pipeline.sh -i input/batch1 -o outputs/batch1_html --writers all
That writes html/ along with XMI and CSV.

​
Stable run id

bash scripts/run_pipeline.sh \
  -i input/batch1 \
  -o outputs/batch1 \
  --run-id star-siteA-batch1-20260709
run_id lands in: 
run_manifest.json
annotations.sqlite: run_metadata.run_id
annotations.sqlite: documents.run_id
annotations.sqlite: concepts.run_id
annotations.sqlite: rxnorm.run_id
annotations.sqlite: cui_counts.run_id
If no --run-id is passed, the wrapper generates one from the run time and run inputs. That generated id is an audit id for this execution, not the reproducibility fingerprint by itself.

​
Custom dictionary XML

bash scripts/run_pipeline.sh \
  -i input/batch1 \
  -o outputs/batch1_custom_dict \
  --dict /path/to/custom_dictionary.xml
Use this only when the custom dictionary XML has already been tested. The wrapper will still snapshot the effective XML into: 
outputs/batch1_custom_dict/run_dictionary.effective.xml

​
Smaller memory

If Java asks for too much memory: 
export CTAKES_AUTOSCALE_MEM_PERCENT=60
bash scripts/run_pipeline.sh -i input/batch1 -o outputs/batch1
Or set a fixed heap: 
bash scripts/run_pipeline.sh -i input/batch1 -o outputs/batch1 --xmx 4096
4096 means 4096 MB.

​
Show the write plan

bash scripts/run_pipeline.sh -i input/batch1 -o outputs/batch1 --print-write-plan
That prints the output folder, work folder, log path, manifest path, effective pipeline snapshot path, effective dictionary snapshot path, and SQLite path without running the notes.