Sanitize an embedding chunk with AI

Sends the chunk text to the institution’s summary model for AI-powered cleanup. Removes noise (navigation, boilerplate, encoding artifacts), normalizes whitespace, and fixes broken formatting — without summarizing or shortening the content.

The sanitized text is returned for preview but not persisted. To save the cleaned text, call PUT /api/user/embedding/{id} with the returned sanitizedText.

Token usage is automatically tallied to the parent Upload’s tokens_used counter.

POST /api/user/embedding/{id}/sanitize

Sanitize an embedding chunk with AI

cURL

curl --request POST \
  --url https://pria.praxislxp.com/api/user/embedding/{id}/sanitize \
  --header 'x-access-token: <api-key>'

200

{
  "success": true,
  "sanitizedText": "This is the cleaned paragraph with noise removed and formatting normalized...",
  "tokensUsed": 142,
  "model": "anthropic.claude-3-haiku-20250514-v1:0"
}

Authorizations

x-access-token · string · header · required

JWT token passed in x-access-token header

Path Parameters

id · string · required

Embedding chunk ID to sanitize

Response

200

application/json

Segment sanitized successfully (preview only — not saved)

success · boolean

Example:

true

sanitizedText · string

The AI-cleaned segment text. Not persisted — use PUT /embedding/{id} to save.

Example:

"This is the cleaned paragraph with noise removed and formatting normalized..."

tokensUsed · integer

Token count consumed by the sanitization LLM call (tallied to parent Upload)

Example:

142

model · string

The LLM model used for sanitization

Example:

"anthropic.claude-3-haiku-20250514-v1:0"