Skip to main content
POST
/
api
/
v1
/
routing-preferences
/
suggest-rule
Suggest a Rule From Corrections
curl --request POST \
  --url https://api.example.com/api/v1/routing-preferences/suggest-rule \
  --header 'Content-Type: application/json' \
  --data '
{
  "expected_mode": "<string>"
}
'

Documentation Index

Fetch the complete documentation index at: https://docs.hitheo.ai/llms.txt

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

Generates a single suggested rule derived from the caller’s recent corrections for a given mode. Two-tier:
  1. Deterministic first. A pure helper extracts the highest-DF content tokens across the cluster’s prompts and assembles a safe \b(token1|token2|…)\b pattern. Vendor-name-scrubbed.
  2. LLM fallback. If the deterministic helper returns nothing (no surviving tokens), Theo asks an internal classifier model for a single regex that matches the sample prompts. Output is re-scanned for vendor names and compiled before being trusted. A fallback that fails either check returns 404 so no unsafe rule ever escapes.
The endpoint does not auto-append the rule. Callers PATCH the suggestion onto an existing preference (or add it to a new one) via PATCH /api/v1/routing-preferences/{id}.

​
Authentication

Requires a Bearer token with the billing API key scope.

​
Body

​
expected_mode
string
required
The Theo mode whose correction cluster you want a rule for.

​
Request Examples

curl
curl -X POST https://www.hitheo.ai/api/v1/routing-preferences/suggest-rule \
  -H "Authorization: Bearer $THEO_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{ "expected_mode": "think" }'

​
Response

{
  "suggested_rule": {
    "pattern": "\\b(clause|provision|indemnity)\\b",
    "target_mode": "think",
    "confidence": 0.88,
    "description": "Generated from 4 corrections that all routed to think."
  },
  "source_correction_ids": ["corr_a", "corr_b", "corr_c", "corr_d"],
  "confidence": 0.88
}
confidence is 0.88 when the deterministic helper produced the pattern, 0.78 when the LLM fallback did.

​
Errors

  • 400 routing_suggest_below_threshold — Cluster for the requested mode has fewer corrections than the cluster threshold.
  • 404 not_found — No cluster exists for expected_mode, or no safe rule could be derived from the available samples.