Add get_asset_history tool #145
Conversation
modelcontextprotocol/tools/assets.py
Outdated
| "count": len(entity_audits), | ||
| "totalCount": response.total_count, |
There was a problem hiding this comment.
what is the difference between count and totalCount?
There was a problem hiding this comment.
Count is the number of audit entries that are now being sent by the MCP, which should be actually equal to the number of entries requested by the user in the request. And total_count is the total number of entries that can be obtained from Atlas.
rahul-madaan
force-pushed
the
audit_history
branch
from
5f8fbd8 to
df3e9e3
Compare
rahul-madaan
force-pushed
the
audit_history
branch
from
df3e9e3 to
85fc754
Compare
| guid=None, qualified_name=None, type_name=None, size=10, sort_order="DESC" | ||
| ): | ||
| """ | ||
| Get the audit history of an asset by GUID or qualified name. |
There was a problem hiding this comment.
Lets add information about how does the LLM fetch the guid or qualified name of the asset. Also this should be Get the audit history of an asset by GUID or qualified name/typename.
There was a problem hiding this comment.
Also do we need to support qualifiedname/typename or just guid should be enough?
There was a problem hiding this comment.
I believe guid should be enough because there would be a very rare case when customer directly provides qualifiedname/typename and even if they do, claude can make search call to find the guid.
I will remove this.
modelcontextprotocol/server.py
Outdated
| Returns: | ||
| Dict[str, Any]: Dictionary containing: | ||
| - entityAudits: List of audit entries with details about each change |
There was a problem hiding this comment.
can we follow a proper pattern. The argument names are snake case but dictionary keys are camelcase
| request = create_audit_search_request( | ||
| guid, qualified_name, type_name, size, sort_item | ||
| ) | ||
| response = client.audit.search(criteria=request, bulk=False) |
There was a problem hiding this comment.
We should catch exceptions and return them as errors since you have defined the tool in that way
There was a problem hiding this comment.
exception handling is properly implemented here, it is in try catch block.
modelcontextprotocol/tools/assets.py
Outdated
| - errors: List of any errors encountered | ||
| Raises: | ||
| Exception: If there's an error retrieving the asset history |
There was a problem hiding this comment.
in cases when API key does not have enough permissions.
modelcontextprotocol/tools/assets.py
Outdated
| client = get_atlan_client() | ||
|
|
||
| # Create sort item | ||
| sort_item = create_sort_item(sort_order) |
There was a problem hiding this comment.
Do we need separate functions for these?
There was a problem hiding this comment.
no, I will remove this.
modelcontextprotocol/server.py
Outdated
| Either guid or qualified_name must be provided. | ||
| type_name (str, optional): Type name of the asset (required when using qualified_name). | ||
| Examples: "Table", "Column", "DbtModel", "AtlasGlossary" | ||
| size (int): Number of history entries to return. Defaults to 10. |
| Dictionary with basic audit information | ||
| """ | ||
| return { | ||
| "entityQualifiedName": getattr(result, "entity_qualified_name", None), |
There was a problem hiding this comment.
Do we need to send qf, guid, typename? considering we are fetching it for that entity only. It will just use up token
There was a problem hiding this comment.
removing qf and typename
| "timestamp": getattr(result, "timestamp", None), | ||
| "action": getattr(result, "action", None), | ||
| "user": getattr(result, "user", None), | ||
| "created": getattr(result, "created", None), |
| """ | ||
| try: | ||
| # Extract basic audit information | ||
| audit_entry = extract_basic_audit_info(result) |
There was a problem hiding this comment.
Move this to the main function only
|
|
||
| return audit_entry | ||
|
|
||
| except Exception as e: |
There was a problem hiding this comment.
When will this exception be raised?
There was a problem hiding this comment.
No, I checked it is all handled. I will remove this.
3 participants
This PR adds a new tool get_asset_history for the following usecases
Tested for prompts like:
"when was the description updated and who updated it for x asset?"
"who was the owner of this x asset on y date"
"what was the description of this asset before it was last updated?"
"who changes the description from x to y for z asset?"