Skip to main content

Documentation Index

Fetch the complete documentation index at: https://docs.orinadus.com/llms.txt

Use this file to discover all available pages before exploring further.

​
Schema

{
  "name": "urchin_search",
  "inputSchema": {
    "type": "object",
    "required": ["query"],
    "properties": {
      "query": { "type": "string" },
      "hours": { "type": "number" },
      "limit": { "type": "number" }
    },
    "additionalProperties": false
  }
}

​
Arguments

​
query
string
required
Substring to search for. Case-insensitive. Matched against the content field of each event. Must not be blank.
​
hours
number
default:"168"
Look back this many hours. Default is 168 (1 week). Events older than the window are excluded.
​
limit
number
default:"20"
Maximum number of results to return. Applied after time and content filtering.

​
Behaviour

  1. Reads all events from the journal.
  2. Filters to events within the hours window.
  3. Filters to events where content contains query (case-insensitive).
  4. Returns the most recent limit matching events, newest first.
The search is a simple str::to_lowercase().contains() — no tokenisation, no stemming, no relevance ranking. For semantic/conceptual search, use urchin_semantic_search instead.

​
Success response

One line per matching event: 
[2026-05-04T20:00:00Z] claude — Use file-backed ephemeral flag for cross-process suppression
[2026-05-03T14:20:00Z] copilot — ephemeral mode: suppress writes during sensitive session
Returns an empty string if no events match.

​
Example calls

{
  "name": "urchin_search",
  "arguments": {
    "query": "ephemeral"
  }
}

​
Error response

{
  "content": [{"type": "text", "text": "query must not be empty"}],
  "isError": true
}
Also returns isError: true if the journal cannot be read. urchin_search is a literal substring match — exact and fast. Use it when you know the specific word or phrase. urchin_semantic_search uses token-cosine similarity (or vector embeddings if URCHIN_EMBEDDER_URL is set). Use it for conceptual queries like “decisions about authentication” where the exact words may not appear in the content.