SmartDocs Multillingual Agentic Rag

Multilingual Agentic RAG engineered for India's linguistic reality, processing legal, GST, and insurance documents natively across 22 languages without intermediate translation layers. Standard RAG architectures break down on real-world Indian data. This system explicitly engineers solutions for four root failures: * Translation-Induced Semantic Loss: Eliminates the intermediate Query → English → LLM → Target Language translation step. Legal and financial terminology (e.g., "Input Tax Credit," "धारा 16") degrades irreversibly across translation boundaries. * Devanagari Processing Noise: Mitigates Zero-Width Joiner (ZWJ) artifacts, 16 Unicode space variants, and OCR extraction anomalies prior to embedding — failures invisible in dev but catastrophic in production on government-issued PDFs. * Retrieval Failure on Alphanumerics: Solves the dense embedding blindspot for exact-match terms (GSTINs, section codes, specific tax amounts) using Reciprocal Rank Fusion (RRF) with a BM25 sparse index trained on Indic-normalized text. * Language Detection Instability: Replaces statistically unstable detection models — which misclassify "transformer kya hai" as Norwegian — with a deterministic 7-step script-and-lexicon execution tree seeded for reproducibility. SmartDocs Multillingual Agentic Rag is the open-source project sahilalaknur21/SmartDocs-Multillingual-Agentic-Rag (https://github.com/sahilalaknur21/SmartDocs-Multillingual-Agentic-Rag) by sahilalaknur21, released under MIT. Ion packages it unmodified from commit ddb9f4c52a53, adding only a manifest, an icon and a start script.

Written by sahilalaknur21, published to Oikou by Ben Grande.

• ★ 36 stars on GitHub
• Licence: MIT
• Built from commit ddb9f4c52a53

View the repository on GitHub

Not rated. Nobody has said who this app is suitable for, so Personal Clouds and the Oikou app treat it as 18+.

Internet

It asks to reach nothing on the internet.

Personal Cloud tools

• Keep a database of its own on your Personal Cloud ion.db — its database uses pgvector
• Run a Redis server of its own on your Personal Cloud (for its task queue and cache) ion.redis

Notifications

It does not send push notifications.

Storage

Up to 1 GB on your Personal Cloud.

Sign in to add this web app to a Personal Cloud of yours. It runs there, on your own computer, and you open it from there; nothing is installed on your phone.

Sign in to add