Semantic Search (for RAG)
AI-generatedSummary
Sends a natural-language query to the selected CTERA expert’s semantic search tool and returns its text chunks for use in a retrieval-augmented generation workflow. You can filter by file modification dates and adjust the number of chunks and context window size.
Inputs
- Expert MCP Endpoint URL (required) — Per-expert MCP endpoint URL; the node’s List Experts operation can provide it.
- Query (required) — Natural-language question to search against the expert’s data.
- Additional Options — Optional filters and retrieval settings: Modified Date Start and Modified Date End filter files by modification date; Number of Results (k) sets the top-k chunk count (UI default 5, range 1–50); Window Size sets context around matches in KB (UI default 2, range 1–10).
Output shape
One n8n item per input item, containing the input JSON plus operation, toolName, and result fields.
The MCP response’s first content block is returned as parsed JSON when possible, otherwise as text; if it is not a text block, that block is returned directly. Date filters are converted to ISO strings. MCP errors fail the item unless Continue On Fail is enabled, in which case an error field is returned.