Skip to main content
The /answer endpoint processes support tickets through the RAG system, combining ML triage classification with retrieval-augmented generation to provide comprehensive, citation-backed answers.

​
Endpoint

POST /api/v1/answer

​
Request Body

​
subject
string
required
The subject line of the support ticket
​
body
string
required
The full body/description of the support ticket
​
user_question
string
Optional specific question from the user. If not provided, the system will generate an answer based on the subject and body.

​
Response

​
triage
object
required
ML triage classification results
​
draft_reply
string
required
AI-generated draft response to send to the customer
​
internal_next_steps
array
required
List of recommended next steps for internal support team
​
citations
array
required
Source documents and snippets used to generate the answer
​
needs_human_review
boolean
required
Flag indicating whether the ticket requires human review before responding

​
Example Request

curl -X POST "http://localhost:8000/api/v1/answer" \
  -H "Content-Type: application/json" \
  -d '{
    "subject": "Refund issue",
    "body": "I was charged twice",
    "user_question": "Can I get a refund?"
  }'

​
How It Works

The /answer endpoint performs the following operations:
  1. Triage Classification - Predicts the ticket category and priority using ML models
  2. Document Retrieval - Searches the knowledge base for relevant information
  3. Answer Generation - Uses an LLM to generate a structured response with:
    • Customer-facing draft reply
    • Internal next steps for support agents
    • Citations from source documents
    • Human review flag based on confidence scores
Ensure documents have been ingested into the knowledge base using the /ingest endpoint before querying for answers.

Build docs developers (and LLMs) love

Get started for freeTalk to us