rag_search
Search Deriva documentation, catalog schema, and data semantically to locate tables, features, vocabularies, datasets, and API guidance.
Instructions
Search Deriva documentation, catalog schema, and catalog data using semantic similarity.
Searches across three categories of indexed content:
Catalog schema (
doc_type="catalog-schema"): The connected catalog's structure, including:Tables with columns, types, nullability, and comments
Foreign key relationships between tables
Feature definitions with term columns, asset columns, and value columns
Vocabulary terms with names, descriptions, and synonyms Use this to understand what data exists, how tables relate, what features are defined, and what vocabulary values are available.
Catalog data (
doc_type="catalog-data"): User-specific records from the connected catalog, including:Datasets with types, versions, and descriptions
Executions with workflow names, status, and inputs Use this to find specific datasets by purpose or trace experiment provenance.
Documentation (
doc_type="user-guide","api-reference","sdk-reference"): Indexed docs from Deriva ecosystem repositories (deriva-ml, ermrest, chaise, deriva-py). Use this for API usage questions and how-to guidance.
Prefer this tool over reading raw resources for catalog exploration.
Use doc_type to focus results on the category you need.
Args:
query: Natural language search query (e.g., "how to create a dataset")
limit: Maximum number of results to return (default 10)
source: Filter by source name (e.g., "deriva-ml-docs", "ermrest-docs")
doc_type: Filter by document type. Key values:
- "catalog-schema": Tables, columns, FKs, features, vocab terms
- "catalog-data": Datasets and executions in the catalog
- "user-guide": DerivaML and Chaise documentation
- "api-reference": ERMrest API documentation
- "sdk-reference": deriva-py SDK documentation
include_schema: If True (default), include catalog schema results when
connected. Set to False to search only documentation.
include_data: If True (default), include per-user data index results
(datasets, executions) when connected. Set to False to
exclude user data from search results.
Returns: Dict with search results including text snippets, relevance scores, source metadata, and GitHub URLs.
Examples: # Explore catalog structure — tables, columns, relationships rag_search("Image tables and features", doc_type="catalog-schema")
# Find vocabulary terms by meaning
rag_search("classification categories", doc_type="catalog-schema")
# Find feature definitions and their columns
rag_search("diagnosis label confidence", doc_type="catalog-schema")
# Find datasets by description, type, or purpose
rag_search("training split labeled", doc_type="catalog-data")
# Find executions by workflow or status
rag_search("training experiment results", doc_type="catalog-data")
# Search Deriva API documentation
rag_search("how to create a dataset", include_schema=False, include_data=False)
# Search everything (docs + schema + data) — the default
rag_search("how are images classified")
Input Schema
| Name | Required | Description | Default |
|---|---|---|---|
| limit | No | ||
| query | Yes | ||
| source | No | ||
| doc_type | No | ||
| include_data | No | ||
| include_schema | No |