Add UC function + VS index constructors for MCP adapters #234
Conversation
nisha2003
changed the title
aravind-segu
left a comment
aravind-segu
left a comment
There was a problem hiding this comment.
Looks good overall, can we also add one for Agents SDK. Its here: https://github.com/databricks/databricks-ai-bridge/blob/main/integrations/openai/src/databricks_openai/agents/mcp_server.py
bbqiu
left a comment
bbqiu
left a comment
There was a problem hiding this comment.
this looks great! happy to approve after we add an agents example as well
| ) | ||
|
|
||
| @classmethod | ||
| def from_uc_function( |
There was a problem hiding this comment.
can we add some docstrings to these functions?
including some sample usage with and without the multiserver mcp client would be helpful!
There was a problem hiding this comment.
Added w/ + w/o multiserver for the six constructors added!
| result = await Runner.run(agent, user_messages) | ||
| return result | ||
| Using with DatabricksMultiServerMCPClient to manage multiple servers: |
There was a problem hiding this comment.
i don't think we have a multi mcp client for the openai agents sdk, only langchain sorry
There was a problem hiding this comment.
Hm ok. I saw some multi mcp client usage here: notebook by aravind. Is this deprecated/am I missing smth?
There was a problem hiding this comment.
i don't think it made it to master unless i'm missing something: https://github.com/search?q=repo:databricks/databricks-ai-bridge%20DatabricksMultiServerMCPClient&type=code
@aravind-segu i believe we ended up not shipping this for openai agents right? but it is something that we should do in the future
There was a problem hiding this comment.
Ah gotcha ok -- I can remove that example
bbqiu
left a comment
bbqiu
left a comment
There was a problem hiding this comment.
lgtm! feel free to merge after adding type hints
| index_name: str = None, | ||
| workspace_client: WorkspaceClient = None, | ||
| **kwargs, | ||
| ): |
|
jenkins merge |
Introduces two new class methods that make it easier to create MCP server connections for Unity Catalog functions and vector search indexes without manually constructing URLs.
Tested manually via Databricks notebooks:
for langchain
for openai
for agents sdk