ZedLow/Constrained-Financial-RAG

Financial RAG Demo

This demo showcases a constrained Financial RAG pipeline designed to reduce hallucinations through explicit routing and hard constraints, not prompt tricks.

What this demo does

• —Routes queries based on detected company entities (Apple / Microsoft)
• —Prevents accidental cross-company document mixing
• —Processes financial tables as images to preserve structure
• —Explicitly rejects unsupported or ambiguous queries

How to test it

Try the following queries:

• —What was Apple’s total revenue in 2023?
• —What is Microsoft’s operating income?
• —Compare Apple and Microsoft revenues → rejected or limited
• —What was Google’s revenue in 2023? → rejected

The UI shows retrieved pages and scores to make the pipeline inspectable.

Important limitations

• —Explicit multi-company questions may trigger cross-entity reasoning
• —Source-constrained prompts are not strictly enforced
• —Dataset is intentionally small (demo-only)

For full technical details and design discussion, see the GitHub repository linked on the CV.