How the Obridge AI Screener Works
The AI screener is the most technically complex feature in Obridge. This post explains how it works — what data it uses, how the score is calculated, and why we built it the way we did.
The AI View Score (1–10)
Every stock analysis in Obridge ends with an AI View Score between 1 and 10. This is not a price target. It is a signal strength — a measure of how aligned the technical, fundamental, and momentum indicators are in a single direction.
Bearish
Neutral
Bullish
Strong
What data goes in
For every analysis request, the screener receives a structured payload of market data:
Technical indicators
RSI (14-period), MACD (12/26/9), EMA-20, EMA-50, SMA-200, Bollinger Band position, pivot levels (S1/S2/R1/R2). These tell us momentum, trend direction, and overbought/oversold conditions.
Fundamental data
P/E ratio, EPS (trailing), promoter holding %, debt-to-equity, revenue growth (YoY), profit margins. These tell us whether the company behind the stock is healthy.
Price context
Current LTP, 52-week high/low position, volume vs. 20-day average, and exchange (NSE/BSE). This grounds the analysis in current market conditions.
How the LLM produces the score
All this data is assembled into a structured prompt and sent to our AI service (hosted on our own servers — your device never touches an AI API directly). The model is instructed to:
- Identify alignment or divergence between technical and fundamental signals.
- Assess momentum strength and directional conviction.
- Produce a structured JSON response with a
viewScore(1–10),viewLabel,recommendation(BUY/SELL/HOLD), a list of strengths (pros), and a list of risks (cons).
The model does not have access to news sentiment or social media — it works purely from quantitative inputs. This keeps the output reproducible and avoids amplifying market rumours.
What the screener explicitly does not do
Our AI guardrails are strict. The screener will never give a price target (“RELIANCE will hit ₹3,200”), a guaranteed return, or advice framed as “you must buy this now.” It provides a directional signal and reasoned pros/cons — the decision remains yours.
This is by design. SEBI regulations on AI-generated investment advice are evolving rapidly, and we believe the right approach is to inform rather than direct.
Token quotas (free vs. Premium)
AI analyses are compute-intensive. Free users receive 5 screener analyses per day. Premium subscribers get unlimited analyses. Quotas reset at midnight IST. The quota counter is tracked server-side — it cannot be bypassed by the client.