Rhino Health SDK — Error Debugger
Diagnose errors in rhino-health Python SDK code (v2.1.x) and provide corrected, runnable fixes.
Context Loading
Before diagnosing, read these reference files:
- Patterns & Gotchas —
../../context/patterns_and_gotchas.mdFocus on §11 (Common Import Paths) and §12 (Gotchas/Pitfalls) for the most frequent error causes. - API Reference —
../../context/sdk_reference.mdCorrect method signatures, endpoint accessors, CreateInput field aliases, and the Import Path Reference table.
Error-to-Fix Reference
Use this table to quickly identify the root cause. Then confirm against the context files before responding.
| Error pattern | Root cause | Fix |
|---|---|---|
NotAuthenticatedError / HTTP 401 | Token expired, wrong credentials, or MFA required | Re-login with rh.login(); pass otp_code param if MFA is enabled |
AttributeError: 'NoneType' has no attribute | get_*_by_name() returned None because the resource was not found | Add a None check immediately after every get_*_by_name() call |
ValidationError (pydantic) | Wrong field names or types in a CreateInput class — often alias confusion | Use Pydantic alias names: project not project_uid, workgroup not workgroup_uid. Check CreateInput summaries in sdk_reference.md |
TypeError in metric configuration | Passed a raw string where a FilterVariable dict or metric object is expected | Use FilterVariable(data_column=..., filter_column=..., filter_value=..., filter_type=FilterType.EQUALS) |
ImportError / ModuleNotFoundError | Wrong import path — SDK uses deep paths that differ from guesses | from rhino_health.lib.metrics import X (NOT rhino_health.metrics); from rhino_health.lib.endpoints.X.X_dataclass import Y (NOT rhino_health.endpoints.X) |
TypeError: aggregate_dataset_metric() argument | Passing List[Dataset] objects instead of List[str] UIDs | Convert: [str(d.uid) for d in datasets] |
IndexError on output_dataset_uids | Accessing as flat list — structure is triply nested | Use .root[0].root[0].root[0] to access the first output dataset UID |
TimeoutError / operation hangs | Default timeout_seconds too low for large datasets | Increase timeout_seconds in wait_for_completion() or wait_for_build() |
TypeError: input_dataset_uids | Passing List[str] instead of List[List[str]] | CodeObjectRunInput.input_dataset_uids must be double-nested: [[uid1, uid2]] |
KeyError / unexpected None in metric results | Wrong data_column name or missing column in dataset | Verify column name matches the dataset schema exactly (case-sensitive) |
Diagnostic Process
When the user provides an error or traceback:
- Identify the exception class — match it to the table above.
- Locate the failing line — find the SDK call that triggered the error.
- Cross-reference with context — check the correct signature/pattern in sdk_reference.md or patterns_and_gotchas.md.
- Check for compound errors — one fix may reveal a second issue (e.g., fixing an import may expose an alias field error).
Response Format
Structure every response in three parts:
1. What went wrong
Identify the specific error and the SDK call that caused it. Quote the relevant line from the traceback.
2. Why
Explain the root cause with reference to SDK behavior. For example: "get_dataset_by_name() returns None when no dataset matches — it does not raise an exception. The subsequent .uid access fails because None has no uid attribute."
3. Corrected code
Provide the complete corrected code block with:
- Fixed imports (exact paths from the Import Path Reference table)
- The fix applied with a brief inline comment explaining what changed
- Any defensive checks added (None guards, type conversions)
# WRONG
from rhino_health.metrics import Mean
# CORRECT
from rhino_health.lib.metrics import MeanCommon Multi-Error Patterns
Watch for these combinations that frequently appear together:
- Import error + alias error: User copied code from an older example. Fix both the import path and the CreateInput field names.
- NoneType + no error handling: User chained calls without checking
get_*_by_name()returns. Add None checks for every lookup. - Metric TypeError + wrong execution method: User confused
get_dataset_metric(per-site, single UID) withaggregate_dataset_metric(cross-site, list of UIDs).
Import Path Quick Reference
When the error is an ImportError, consult this cheatsheet:
# Metrics (CORRECT)
from rhino_health.lib.metrics import Mean, Count, Cox, KaplanMeier, RocAuc
# Dataclasses (CORRECT pattern)
from rhino_health.lib.endpoints.project.project_dataclass import ProjectCreateInput
from rhino_health.lib.endpoints.dataset.dataset_dataclass import DatasetCreateInput
from rhino_health.lib.endpoints.code_object.code_object_dataclass import CodeObjectCreateInput, CodeObjectRunInput, CodeTypes
from rhino_health.lib.endpoints.syntactic_mapping.syntactic_mapping_dataclass import DataHarmonizationRunInput
# Enums (CORRECT)
from rhino_health.lib.endpoints.endpoint import VersionMode, NameFilterMode
from rhino_health.lib.metrics import FilterType, FilterVariableFor the full list, see the Import Path Reference table at the end of sdk_reference.md.