when using keyword search use proper queries with /q #47
Conversation
| - Examples: "user_transactions", "revenue_2024", "customer_id" | ||
| - Will match: exact text appearances of these terms | ||
| - Structured full-text search - **always start queries with "/q "** | ||
| - Supports full boolean logic: AND (default), OR, NOT, parentheses, field searches |
There was a problem hiding this comment.
This would be good addition for simple search as well, once we establish this works as expected in majority of cases.
There was a problem hiding this comment.
what do you think the process o releasing it for the simple search should be? should we try updating it for the slackbot and run evals?
src/mcp_server_datahub/mcp_server.py
Outdated
| • "/q sales AND revenue" → both terms (explicit AND) | ||
| • "/q \"user data table\"" → exact phrase matching | ||
| • "/q (sales OR revenue) AND quarterly" → complex boolean combinations | ||
| • "/q name:dataset* AND platform:snowflake" → field-specific searches |
There was a problem hiding this comment.
What does * mean here ?
Also, Its not much likely for agent to know what fields (e.g. name, platform) to search on, therefore this example can cause issues.
There was a problem hiding this comment.
These field-specific filters have similar behavior to structured filters arg in search tool. I'm slightly concerned about behavior if both are used together ? Is one preferred over another Can you confirm this ?
There was a problem hiding this comment.
I think you are right I should remove this example - * would match with everything starting with dataset - I'm actually not sure if this is a realistic scenario here.
alexsku
changed the title
No description provided.