rhino-sdk-write（rhino SDK write 命令行）

Rhino Health SDK — Code Generator

Generate production-ready rhino-health Python SDK code from natural language descriptions.

Context Loading

Before generating code, read all of these reference files:

1. API Reference — ../../context/sdk_reference.md Endpoint classes, methods, enums, CreateInput summaries, dataclass fields, import paths.
2. Patterns & Gotchas — ../../context/patterns_and_gotchas.md Auth patterns, resource lookup, metrics execution, filtering, code objects, async, and pitfalls.
3. Metrics Reference — ../../context/metrics_reference.md All 40+ federated metrics with parameters, import paths, and decision guide.
4. Example Index — ../../context/examples/INDEX.md Mapping of use cases to working example files with key methods and difficulty levels.

Example Matching

After loading context, check the example index for a matching use case. If one exists, read the full example file from ../../context/examples/<filename> and follow its patterns. The examples are verified working code from the official Rhino GitHub repository.

Code Template

Every generated script must follow this structure:

# --- Imports ---
import rhino_health as rh
from getpass import getpass
# ... additional imports (metrics, dataclasses, enums) ...

# --- Authentication ---
session = rh.login(username="my_email@example.com", password=getpass())

# --- Configuration ---
PROJECT_NAME = "My Project"
DATASET_UIDS = ["uid-1", "uid-2"]  # Replace with actual UIDs

# --- Resource Lookup ---
project = session.project.get_project_by_name(PROJECT_NAME)
if project is None:
    raise ValueError(f"Project '{PROJECT_NAME}' not found")

# --- Core Logic ---
# ... SDK calls ...

# --- Result Handling ---
print(result)

Template Rules

• Authentication: Always use getpass(). Never hardcode passwords. Support MFA with otp_code parameter.
• Imports: Place all imports at the top. Use exact paths from the Import Path Reference table in sdk_reference.md.
• Resource lookup: Use get_*_by_name() for human-friendly lookups. Always check for None returns.
• Constants: Define project names, UIDs, and configuration values as named constants near the top.
• Type hints: Add type hints to function signatures when generating functions or classes.

Validation Checklist

Run through every item before returning generated code. Flag violations and fix them.

Endpoint Accessors

Verify the correct accessor is used for each operation:

Import Paths

Verify every import against the Import Path Reference in sdk_reference.md. Common mistakes:

Metric Calls

• aggregate_dataset_metric takes List[str] of UIDs: [str(d.uid) for d in datasets]
• get_dataset_metric takes a single dataset_uid: str
• joined_dataset_metric takes query_datasets and optional filter_datasets as List[str]
• Metric configuration objects require data_column (not column or field)
• FilterVariable dicts use keys: data_column, filter_column, filter_value, filter_type

CreateInput Alias Fields

Several CreateInput classes use Pydantic aliases. Pass the alias name, not the field name:

Nested Structures

• CodeObjectRunInput.input_dataset_uids is List[List[str]]: [[uid1, uid2]]
• output_dataset_uids is triply nested: access via .root[0].root[0].root[0]
• group_by parameter format: {"groupings": [{"data_column": "col"}]}
• data_filters list: [FilterVariable(data_column="col", filter_column="col", filter_value="val", filter_type=FilterType.EQUALS)]

Async Operations

• Call wait_for_build() after creating Generalized Compute code objects
• Call wait_for_completion() after run_code_object(), run_data_harmonization(), and run_sql_query()
• Both methods block until the operation finishes or times out

None Checks

Every get_*_by_name() call must be followed by a None check:

dataset = project.get_dataset_by_name("Name")
if dataset is None:
    raise ValueError("Dataset not found")

Output Format

Return a single, complete, runnable .py script. Include:

• All necessary imports at the top
• Authentication block
• Constants for configurable values (project names, UIDs, column names)
• Inline comments explaining non-obvious SDK behavior (e.g., why UIDs are stringified, why None-checks are needed)
• A brief header comment describing what the script does

Do not split the code across multiple blocks. The user should be able to copy the entire output into a .py file and run it (after replacing placeholder values).

Additional Guidance

Choosing Between Per-Site, Aggregated, and Joined Metrics

Refer to patterns_and_gotchas.md section 4 for the decision:

• Per-site (get_dataset_metric): results from one dataset/site
• Aggregated (aggregate_dataset_metric): combined results across multiple datasets
• Federated join (joined_dataset_metric): SQL-like join across distributed datasets

Choosing the Right Metric

Consult the Quick Decision Guide in metrics_reference.md:

• "How many..." -> Count
• "Average/mean..." -> Mean
• "Survival time..." -> KaplanMeier or Cox
• "Correlation..." -> Pearson, Spearman, ICC
• "Compare groups..." -> TTest, OneWayANOVA, ChiSquare
• "Risk/odds..." -> TwoByTwoTable, OddsRatio, RiskRatio
• "ROC curve..." -> RocAuc, RocAucWithCI