Docs: updating mcp tools reference page#226
Conversation
|
Preview deployment for your docs. Learn more about Mintlify Previews.
💡 Tip: Enable Workflows to automatically generate PRs for you. |
ayush-shah
left a comment
There was a problem hiding this comment.
Validated this reference update against the MCP implementation source of truth: openmetadata-mcp/src/main/resources/json/data/mcp/tools.json, DefaultToolContext.java, and the individual tool classes.
The consolidation is needed and directionally correct, but there are several schema/implementation mismatches that should be fixed before merge. I left the inline comments on the v1.12.x copy; please mirror the same content fixes into v1.13.x-SNAPSHOT/how-to-guides/mcp/reference.mdx because the two reference pages are currently identical in this PR.
| | `fields` | string | No | Comma-separated additional fields to include | | ||
| | `queryFilter` | string | No | Advanced: direct OpenSearch DSL JSON query | | ||
| | `includeDeleted` | boolean | No | Include deleted entities (default: false) | | ||
| | `includeAggregations` | boolean | No | Include aggregation/facet data (default: false) | |
There was a problem hiding this comment.
The implementation and schema also expose maxAggregationBuckets for search_metadata (tools.json and SearchMetadataTool). Since this page now claims to be the complete MCP tool reference, please add this optional parameter beside includeAggregations with the default/max behavior: default 10, max 50.
There was a problem hiding this comment.
Added maxAggregationBuckets to search_metadata with default/max behavior in both versions in commit 02426d6.
| ### semantic_search | ||
|
|
||
| **Description**: Retrieve detailed information about a specific entity using its fully qualified name. | ||
| **Description**: Find data assets by meaning using vector search ([setup guide](../../../deployment/semantic-search#enable-semantic-search-openmetadata-deployment-guide)). Use for exploratory queries where you don't know exact names. Returns conceptually related assets even when no keywords match. |
There was a problem hiding this comment.
The relative page path is right, but the hash looks stale. The target page heading is # Enable Semantic Search, so this should likely link to ../../../deployment/semantic-search#enable-semantic-search. Mintlify link checks often validate the page path without catching a bad fragment.
There was a problem hiding this comment.
Updated the semantic search setup link fragment to #enable-semantic-search in both pages in commit 02426d6.
| "arguments": { | ||
| "query": "revenue forecasting and financial reporting", | ||
| "filters": { | ||
| "entity_type": ["table", "dashboard"], |
There was a problem hiding this comment.
This filter key should be entityType, not entity_type. SemanticSearchTool passes filter keys through to VectorSearchQueryBuilder, and that builder only handles entityType; unknown keys are ignored, so this example would silently fail to filter by entity type.
There was a problem hiding this comment.
Updated semantic search filter docs/examples to use entityType (not entity_type) in both versions in commit 02426d6.
| "arguments": { | ||
| "entityType": "table", | ||
| "fqn": "mysql_prod.ecommerce.public.customer_orders", | ||
| "patch": [ |
There was a problem hiding this comment.
patch is documented here as a JSON array, but both tools.json and PatchEntityTool expect patch to be a JSON array string. PatchEntityTool casts params.get("patch") to String before parsing it with Json.createReader(...), so this example would fail as written. Please either stringify the patch in the example or update the schema/implementation to accept arrays.
There was a problem hiding this comment.
Updated patch_entity example so patch is a JSON array string as required by the current implementation, in both pages in commit 02426d6.
| | `name` | string | Yes | Name of the term | | ||
| | `description` | string | Yes | Definition of the term | | ||
| | `parentTerm` | string | No | FQN of the parent term (for nested/hierarchical terms) | | ||
| | `owners` | array | No | List of owner usernames or emails | |
There was a problem hiding this comment.
create_glossary_term also supports optional reviewers in both tools.json and GlossaryTermTool. Please add it to the parameter table so the consolidated reference stays aligned with the actual tool schema.
There was a problem hiding this comment.
Added optional reviewers to create_glossary_term parameter tables in both versions in commit 02426d6.
|
|
||
| **Description**: Create a new metric entity in OpenMetadata to track and standardize business KPIs. | ||
|
|
||
| > See the Sandbox/Collate tool schema for required parameters specific to your OpenMetadata version. |
There was a problem hiding this comment.
This placeholder leaves create_metric effectively undocumented. The actual schema has required name, metricExpressionLanguage, and metricExpressionCode, plus optional fields like metricType, granularity, unitOfMeasurement, owners, reviewers, relatedMetrics, tags, and domains. Please add a parameter table and at least one valid example instead of deferring to the tool schema.
There was a problem hiding this comment.
Replaced the placeholder with a full create_metric parameter table and valid example in both versions in commit 02426d6.
|
|
||
| **Description**: Create a data quality test case for a table or column. For column tests, ensure the column's data type is listed in the test definition's `supportedDataTypes`. | ||
|
|
||
| > See the Sandbox/Collate tool schema for the full parameter list, as parameters vary by test definition. |
There was a problem hiding this comment.
This section needs a real parameter table and example. tools.json exposes name, fqn, columnName, entityType, testDefinitionName, description, and parameterValues; clients see name, fqn, testDefinitionName, and parameterValues as required. Without those details, users cannot construct a valid create_test_case call from the reference page.
There was a problem hiding this comment.
Added a full create_test_case parameter table (including required fields) and a valid example in both versions in commit 02426d6.
|
|
||
| ### root_cause_analysis | ||
|
|
||
| **Description**: Trace a data quality failure back to its origin by traversing data quality lineage across pipeline hops. |
There was a problem hiding this comment.
root_cause_analysis is listed as an available tool but is missing its required and optional params. The schema requires fqn and entityType, and supports upstreamDepth, downstreamDepth, queryFilter, and includeDeleted. Please add a parameter table and example so this is usable as a reference.
There was a problem hiding this comment.
Added root_cause_analysis required/optional parameters and a usable example in both versions in commit 02426d6.
ayush-shah
left a comment
There was a problem hiding this comment.
Revalidated current head fd97a6a7.
Validation:
rtk mint broken-links: passed.git diff --check origin/main...refs/remotes/pr/226: reportsnew blank line at EOFin bothv1.12.x/how-to-guides/mcp/reference.mdxandv1.13.x-SNAPSHOT/how-to-guides/mcp/reference.mdx.
The earlier schema/implementation review on this same head still applies; please also remove the EOF blank lines before merge.
|
@copilot go through the open comments, validate them and fix those |
Agent-Logs-Url: https://github.com/open-metadata/docs-om/sessions/d5aa7e03-8401-49c5-b0a6-dbc438aac06b Co-authored-by: ayush-shah <40225091+ayush-shah@users.noreply.github.com>
Applied all requested schema/implementation alignment fixes and mirrored them to
Removed the EOF blank-line issue and addressed the earlier schema alignment points in both reference files in commit
Reviewed all open actionable comments, validated them, and applied the requested fixes in commit |
All MCP tools now on single docs page, added new tools, revised params to match