Technical Roadmap¶

This document tracks the future technical roadmap and planned improvements for My Independent AI.

Intent-Driven Query Orchestration¶

Smart Query Router¶

Transition from basic RAG to a ReAct-based Dispatcher that classifies user intent (Analytical vs. Semantic) in real-time.

Query Decomposition¶

Ability to break complex prompts (e.g., "List my cameras and why my Paris photos are blurry") into parallel sub-tasks.

Hybrid Intelligence Engine¶

Text-to-SQL Sidecar¶

A pluggable PostgreSQL/SQLite module for local metadata (EXIF, WhatsApp stats) to handle "Global/Count" queries that Vector search misses.

Knowledge Graph Integration (GraphRAG)¶

Implementing Neo4j/Memgraph to map relationships between entities (People, Places, Gear) for deep relational insights.

Privacy & Security Hardening¶

Edge-First PII Redaction¶

Local ingestion pipeline using Presidio or custom Regex to scrub sensitive data before any cloud-based embedding.

CMEK Layer¶

Standardizing Customer-Managed Encryption Keys for all modular data stores (Vector, SQL, and Object Storage).

Modular Data Importers¶

Extended Exporters¶

New pluggable connectors for Synology Photos (Face Tags), Gmail (Metadata-only).

OCR for Scanned PDFs (Docling)¶

Replace the current pypdf text extractor with IBM Docling to support image-based PDFs (e.g. scanned contracts, receipts). Docling runs fully locally, handles complex layouts (tables, multi-column), and can replace pypdf entirely — making the WARNING: No text extracted from PDF a thing of the past.