Skip to main content

Data Flow

This document describes the data flow for each core Evatar feature, helping developers understand the complete request path from client to storage.

Screenshot Sync Flow​

Screenshot sync is Evatar's most fundamental feature. The Android client scans screenshot files on the device, uploads them to the backend, which automatically triggers LLM analysis.

Chat Message Flow​

The chat system is based on the Agent architecture, supporting multi-turn dialogue and tool calling.

Memory Extraction Flow​

The memory system extracts user information from three sources: chat conversations, screenshot analysis, and reasoning articles.

Intent Reasoning Flow​

The intent reasoner (services/reasoner.py) is Evatar's "background thinking" module, running once per hour to analyze recent user activity and generate structured notes.

RAG Retrieval Flow​

When users ask questions in chat, the Agent uses RAG (Retrieval-Augmented Generation) to retrieve relevant information from the screenshot knowledge base.