Architecture &
Design
Built on a stateless, cloud-native stack designed for infinite horizontal scalability and millisecond latency.
System Topology v2.0
The Core Engine
Edge Layer
Global CDN & React Widget handling sub-30ms interactions.
HTTPSWSS
Active Process
Orchestrator
Node.js API handling auth, rate-limiting, and context retrieval.
$ verify_jwt(token)
$ query_vector_store()
Memory Core
pgvector instance storing millions of embeddings with tenant isolation.
PostgreSQLS3
Request Lifecycle
Trace a single message through our stack.
01
Ingestion
User sends a message. The widget encrypts the payload and attaches the session JWT.
02
Semantic Retrieval
System embeds the query and fetches the top-5 most relevant text chunks from the vector index, filtered by tenant_id.
03
Inference & Analysis
LLM generates natural response while parallel classifiers check for sentiment and lead signals.
04
Execution
Response streamed to user. If lead detected, data is pushed to CRM via Webhook immediately.