M4 Python API

The M4 Python API provides programmatic access to clinical datasets for code execution environments. It mirrors the MCP tools but returns native Python types (DataFrames, dicts) instead of formatted strings.

When to Use the API vs MCP Tools

Use the Python API when:

• Complex clinical analysis - Multi-step analyses that require intermediate results, joins across queries, or statistical computations
• Large result sets - Query results with thousands of rows can be stored in DataFrames without dumping into context
• Mathematical operations - Aggregations, percentile calculations, statistical tests, and counting that benefit from pandas/numpy
• Iterative exploration - Building up analysis through multiple queries where each step informs the next

Use MCP tools when:

• Simple one-off queries where the result fits comfortably in context
• Interactive exploration where you want to see results immediately

Required Workflow

You must follow this sequence:

1. set_dataset()

   - Select which dataset to query (REQUIRED FIRST)

2. get_schema()

   /

   get_table_info()

   - Explore available tables

3. execute_query()

   - Run SQL queries

from m4 import set_dataset, get_schema, get_table_info, execute_query

# Step 1: Always set dataset first
set_dataset("mimic-iv")  # or "mimic-iv-demo", "eicu", "mimic-iv-note"

# Step 2: Explore schema
schema = get_schema()
print(schema['tables'])  # List of table names

# Step 3: Inspect specific tables before querying
info = get_table_info("hosp_patients")
print(info['schema'])  # DataFrame with column names, types
print(info['sample'])  # DataFrame with sample rows

# Step 4: Execute queries
df = execute_query("SELECT gender, COUNT(*) as n FROM hosp_patients GROUP BY gender")
# Returns pd.DataFrame - use pandas operations freely

API Reference

Dataset Management

list_datasets()

list[str]

set_dataset(name)

str

get_active_dataset()

str

Tabular Data (requires TABULAR modality)

get_schema()

dict

{'backend_info': str, 'tables': list[str]}

get_table_info(table, show_sample=True)

dict

{'schema': DataFrame, 'sample': DataFrame}

execute_query(sql)

DataFrame

Clinical Notes (requires NOTES modality)

search_notes(query, note_type, limit, snippet_length)

dict

{'results': dict[str, DataFrame]}

get_note(note_id, max_length)

dict

{'text': str, 'subject_id': int, ...}

list_patient_notes(subject_id, note_type, limit)

dict

{'notes': dict[str, DataFrame]}

Error Handling

M4 uses a hierarchy of exceptions. Catch specific types to handle errors appropriately:

M4Error (base)
├── DatasetError      # Dataset doesn't exist or not configured
├── QueryError        # SQL syntax error, table not found, query failed
└── ModalityError     # Tool incompatible with dataset (e.g., notes on tabular-only)

Recovery patterns:

from m4 import execute_query, set_dataset, DatasetError, QueryError, ModalityError

try:
    df = execute_query("SELECT * FROM hosp_patients")
except DatasetError as e:
    # No dataset selected, or dataset not found
    # Recovery: call set_dataset() first, or check list_datasets()
    set_dataset("mimic-iv")
    df = execute_query("SELECT * FROM hosp_patients")
except QueryError as e:
    # SQL error or table not found
    # Recovery: check table name with get_schema(), fix SQL syntax
    print(f"Query failed: {e}")
except ModalityError as e:
    # Tried notes function on tabular-only dataset
    # Recovery: switch to dataset with NOTES modality
    set_dataset("mimic-iv-note")

Dataset State

Important: Dataset selection is module-level state that persists across function calls.

set_dataset("mimic-iv")
df1 = execute_query("SELECT COUNT(*) FROM hosp_patients")  # Uses mimic-iv

set_dataset("eicu")
df2 = execute_query("SELECT COUNT(*) FROM patient")        # Uses eicu

MCP Tool Equivalence

The Python API mirrors MCP tools but with better return types:

execute_query

execute_query()

pd.DataFrame

get_database_schema

get_schema()

dict

tables

get_table_info

get_table_info()

dict

schema

sample

Use the Python API when you need to:

• Chain queries in analysis pipelines
• Perform pandas operations on results
• Avoid parsing formatted output