Skip to content

Conversation

Orbital-Web
Copy link
Contributor

@Orbital-Web Orbital-Web commented Jun 10, 2025

Description

IMPORTANT MAYBE?: Run kg_rs_vespa inside KG Dev Assistant if you have already extracted & clustered some kg stuff

Updated how kg relationships and entities are stored/filtered in vespa

How Has This Been Tested?

Locally, with these questions:
Filtered Search Entity

  • Which linear ticket talks about confluence permission syncing (LINEAR::*)
  • Which linear or jira tickets talks about login issues due to user migration error. EN2
    Filtered Search Relations
  • Which linear tickets by Chris talks about confluence permission syncing
  • Which linear tickets assigned to employee Chris talks about confluence permission syncing
  • Which JIRA tickets under EN-13 talks about performance testing? (did not go SEARCH route, which I guess is fine)

Made sure normal search works fine

Backporting (check the box to trigger backport action)

Note: You have to check that the action passes, otherwise resolve the conflicts manually and tag the patches.

  • This PR should be backported (make sure to check that the backport attempt succeeds)
  • [Optional] Override Linear Check

@Orbital-Web Orbital-Web requested a review from a team as a code owner June 10, 2025 16:31
Copy link

vercel bot commented Jun 10, 2025

The latest updates on your projects. Learn more about Vercel for Git ↗︎

Name Status Preview Comments Updated (UTC)
internal-search ✅ Ready (Inspect) Visit Preview 💬 Add feedback Jun 13, 2025 8:13pm

Copy link
Contributor

@greptile-apps greptile-apps bot left a comment

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

PR Summary

Refactors knowledge graph (KG) storage and filtering in Vespa, significantly changing how relationships and entities are handled to support more precise querying patterns.

  • Changes KG storage format in backend/onyx/document_index/vespa/app_config/schemas/danswer_chunk.sd.jinja from weightedset to array types with structured relationship storage
  • Implements fail-loud error handling in backend/onyx/agents/agent_search/kb_search/nodes/a3_generate_simple_sql.py by raising exceptions instead of silently continuing on SQL failures
  • Removes entity/relationship generalization (e.g., 'LINEAR::*') from backend/onyx/document_index/vespa/kg_interactions.py in favor of exact matching
  • Adds sophisticated KG filtering in backend/onyx/document_index/vespa/shared_utils/vespa_request_builders.py with support for wildcards and relationship-based queries

5 files reviewed, 2 comments
Edit PR Review Bot Settings | Greptile

@Orbital-Web
Copy link
Contributor Author

Merging with the subtype pr

@Orbital-Web Orbital-Web deleted the kg-vespa-new-schema branch June 17, 2025 17:03
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Labels
None yet
Projects
None yet
Development

Successfully merging this pull request may close these issues.

2 participants