POST FAQ Matcher API
/api/v1/ai/faq-match
Match a customer search query semantically against an FAQ Collection to retrieve the most relevant answer.
Architecture Role & AI Definition
FAQ Matcher API provides low-latency, deterministic REST execution for production engineering workflows. It processes structured payloads with strict schema validation, returns uniform JSON envelopes, and is secured via SHA-256 API key authentication with atomic credit pre-authorization locks.
Production Reliability Guidelines
Configure a hard client timeout of 5 to 8 seconds. If network latency spikes, cancel connection to avoid holding open sockets in worker pools.
Upon receiving 429 Rate Limit or transient 5xx, pause with exponential backoff:
wait = min(max_backoff, base * 2^attempt + jitter).
RSFlowHub acquires an atomic lock verifying base credits before model invocation. If validation fails, zero credits are deducted.
[!IMPORTANT]
Prerequisite: Before you can match queries, you must first upload your FAQs and create a collection using the FAQ Collections API to obtain a validcollection_id.
Request Body Parameters
| Field | Type | Required | Description |
|---|---|---|---|
collection_id |
integer | Required | The ID of the stored FAQ collection to search within. |
query |
string | Required | The user question or query to match (Max: 5,000 characters). |
min_confidence |
float | Optional | Optional confidence score threshold to filter out weak matches (0.0 to 1.0). Defaults to 0.3. |
top_n |
integer | Optional | Optional number of top matches to return (1 to 5). Defaults to 1. |
Pipeline & Adaptive Reranking
The FAQ Matcher API runs an advanced semantic pipeline to ensure high precision across multiple languages while maintaining low overhead:
- 🌐 Multilingual & Cross-Language Retrieval: If direct search confidence is low, the pipeline automatically detects the language (supporting Hindi, Hinglish, Gujarati, Bengali, Telugu, Marathi, Spanish, etc.) and normalizes it to a retrieval-oriented English intent query.
- 🤖 Adaptive AI Reranker: If candidates have close vector similarity scores (gap < 0.15) or low overall confidence (< 0.70), the matcher dynamically invokes an AI model reranker to determine the correct target FAQ by context. Clear winners bypass the reranker to optimize API cost.
- 🛡️ Robust Provider Fallback: If the primary AI reranker encounters failures (timeouts/rate-limits), the service catches the exception and falls back safely to vector-similarity ranks without disrupting the client.
Example Request
{
"collection_id": 12,
"query": "how to update my passcode?",
"min_confidence": 0.4,
"top_n": 1
}curl -X POST https://rsflowhub.com/api/v1/ai/faq-match \
-H "Authorization: Bearer YOUR_API_KEY" \
-H "Content-Type: application/json" \
-d '\''{
"collection_id": 12,
"query": "how to update my passcode?",
"min_confidence": 0.4,
"top_n": 1
}'\''Example Response
{
"success": true,
"request_id": "req_7f3a9c2d1e4b8a6f",
"data": {
"result": {
"matches": [
{
"id": "faq_1",
"question": "How do I reset my password?",
"answer": "Go to Account Settings > Security and click on 'Change Password'.",
"confidence": 0.9562,
"match_reason": "Semantic similarity match."
}
]
}
},
"meta": {
"version": "v1",
"timestamp": "2026-09-27T12:34:56+00:00"
}
}API Request Example
Use the cURL snippet below to test the endpoint.
curl -X POST 'https://rsflowhub.com/api/v1/ai/faq-match' \
-H 'Authorization: Bearer YOUR_API_KEY' \
-H 'Content-Type: application/json' \
-d '{
"collection_id": 12,
"query": "how to update my passcode?",
"min_confidence": 0.4,
"top_n": 1
}'Want to try it without code? Test this request in the live Playground.
Common Failure Modes & Troubleshooting Matrix
| HTTP Code | Error Code | Root Cause | Recommended Remediation |
|---|---|---|---|
| 401 | API_KEY_MISSING / INVALID_API_KEY |
No key sent, or the key was mistyped or regenerated. | Copy the current key from Dashboard → API Keys and send it as Authorization: Bearer (or X-API-Key). |
| 402 | INSUFFICIENT_CREDITS |
Your balance cannot cover the estimate for this call (base 2 credits plus input/output size). | Buy a plan in Dashboard → Billing. Nothing was charged. |
| 422 | VALIDATION_ERROR |
A field is missing, too long or the wrong type, or the body is not valid JSON. | Read error.details for the field messages and fix the request. Do not retry the same body. |
| 429 | RATE_LIMIT_EXCEEDED |
Your plan-wide or per-endpoint per-minute limit was reached. | Wait the seconds in the Retry-After header, then retry with jitter. Rate limits |
| 502 | AI_PROVIDER_ERROR |
The AI model and its backup both failed. | Retry with backoff and check the status page. Nothing was charged. |