VeriLens API Overview
Welcome to the VeriLens API Developer Documentation. Integrate AI-driven Indonesian misinformation detection and evidence-grounded fact-checking directly into your systems.
What is VeriLens?
VeriLens is an AI-powered misinformation and hoax verification platform engineered specifically for Indonesian digital media and communication ecosystems. It combines a high-throughput IndoBERT model for instant syntactic claim classification with an autonomous 3-stage RAG (Retrieval-Augmented Generation) pipeline that crawls credible news outlets, fact-checking registries (e.g. TurnBackHoax, Kominfo Cek Fakta), and cross-references source evidence using Google Gemini.
Dual-Tier Verification Architecture
VeriLens provides two distinct analytical tiers depending on your latency and depth requirements: • **Fast Check (Free Tier)**: Sub-second inference powered by a fine-tuned IndoBERT transformer model. Evaluates stylistic, sensationalist, and linguistic markers characteristic of Indonesian hoaxes. Free to use with no credit cost. • **Deep Analysis (Premium Tier)**: Credit-gated 3-stage pipeline with live Server-Sent Events (SSE) streaming. Performs atomic claim extraction, real-time web context retrieval across reputable sources, and rigorous factual synthesis with confidence scoring, explanatory reasoning, and cited reference URLs.
Base Environments
All requests should be dispatched to one of the following root base URLs:
| Environment | Base URL | Protocol |
|---|---|---|
| Production | https://api.verilens.id | HTTPS (Encrypted) |
| Local Development | http://localhost:8080 | HTTP (Local) |
Quick Request Example
Here is an example of verifying an Indonesian text statement using the Free Tier endpoint:
curl -X POST https://api.verilens.id/api/v1/predict \-H "Authorization: Bearer YOUR_API_KEY" \-H "Content-Type: application/json" \-d '{"text": "Pemerintah secara resmi mengumumkan pembagian subsidi beras 10kg mulai besok."}'
Advisory Nature & Limitations
AI-generated evaluations examine semantic patterns, online registry records, and cross-referenced news reports. Confidence scores (ranging from 0.0 to 1.0) reflect statistical alignment with current accessible evidence.