OmniReconcile: Local Agentic AI for Financial Ops

OmniReconcile is a decentralized agentic framework designed to solve the "data privacy vs. automation" dilemma in fintech. Built with a high-performance Go (Fiber) backend, the system acts as a secure gateway for massive multi-part file ingestions. The core innovation lies in its Self-Correcting Execution Pipeline: Schema Mapping: It uses excelize to stream-read headers without memory bloat. Local Brain: It leverages Ollama to host local models (Gemma/Llama), which interpret natural language queries to generate targeted Pandas/Polars scripts. Secure Sandbox: A specialized Go-orchestrated Python virtual environment executes these scripts in isolation, returning structured JSON insights. This architecture eliminates the need for expensive, non-private SaaS tools. By keeping the entire lifecycle—from file upload to LLM inference and code execution—on bare-metal AMD hardware, OmniReconcile provides banking-grade security with the flexibility of modern Generative AI.