Notify LLM on endpoint create/update

[ruwinirathnamalala]

#990

Overview

Integrates LLM RAG (Retrieval-Augmented Generation) indexing into the endpoint lifecycle. When an endpoint is created or updated, its data is automatically sent to the LLM api-tools indexing service.

Changes

Liquibase

• 20260513120000_add-is-llm-synced-to-endpoints.xml/sql — New migration that adds an is_llm_synced BOOLEAN column to the endpoints table to track whether each endpoint has been successfully indexed by the LLM.
• ..._rollback.sql — Rollback script that drops the is_llm_synced column.

Resql

• get_endpoint_by_id.sql — New query that retrieves a single endpoint's details (name, description, selected URL, HTTP method, and parameters) by resolving the isSelected definition entry. Used by the LLM indexing flow to build the indexing payload.
• update_endpoint_llm_sync_status.sql — New query that updates the is_llm_synced flag on an endpoint to true (success) or false (failure) after an indexing attempt.

Ruuter

• index-llm.yml (new) — Standalone POST endpoint that orchestrates the full LLM indexing flow for a given endpointId:
  1. Validates the input.
  2. Fetches endpoint details via Resql.
  3. Calls LLM_RUUTER_PUBLIC /api-tools/index with the endpoint payload.
  4. Updates is_llm_synced based on the LLM response (success → true, failure/not-found → false).
  5. Returns appropriate HTTP status codes (200, 400, 404, 502).
• create-endpoint.yml — Updated to call the index-llm flow after a successful endpoint creation.
• update-endpoint.yml — Updated to call the index-llm flow after a successful endpoint update.

Configuration

• constants.ini — Added LLM_RUUTER_PUBLIC=http://ruuter-public:8086/rag-search to configure the LLM service base URL.

[ruwinirathnamalala]

The structuredClone approach failed with DataCloneError because flow nodes contain function references (onDelete, onEdit, setClickedNode, update), which cannot be cloned by the structured clone algorithm. The JSON.parse(JSON.stringify()) method is used here to strip functions from nodes before saving to history, which are then manually re-attached during undo/redo operations.

[ruwinirathnamalala]

The structuredClone approach failed with DataCloneError because flow nodes contain function references (onDelete, onEdit, setClickedNode, update), which cannot be cloned by the structured clone algorithm. The JSON.parse(JSON.stringify()) method is used here to strip functions from nodes before saving to history, which are then manually re-attached during undo/redo operations.

[sonarqubecloud]

Quality Gate passed

Issues
0 New issues
0 Accepted issues

Measures
0 Security Hotspots
0.0% Coverage on New Code
0.0% Duplication on New Code

See analysis details on SonarQube Cloud