EMS Analyst API 📊🤖

Umbrella repo grouping the three analyst subsystems — model, agent, and server — into a single architectural unit

Projects

• ems-analyst-model - ML forecasting models with automatic retraining
• ems-analyst-agent - RAG-enabled energy analyst agent with knowledge graph
• ems-analyst-server - FastAPI service providing unified data and AI interface

Architecture

rectangle analyst_api {
  rectangle model
  rectangle agent
  rectangle server
}

rectangle domain_mcp_server
database timeseries
database vector_db
database graph_db
cloud ercot_api
cloud agent_apis
rectangle ems_hmi

domain_mcp_server -u-> vector_db: SQL
domain_mcp_server -u-> graph_db: Cypher
agent -u-> domain_mcp_server: MCP
model -u-> timeseries: SQL
model -l-> ercot_api: HTTP
model -d-> server
agent -d-> server
agent -r-> agent_apis: HTTP

ems_hmi -u-> server: HTTP

Server Capabilities

The server provides a unified FastAPI interface for:

Endpoint	Purpose	Data Source
/historical	Historical timeseries data	TimescaleDB
/predictions	ML forecasting	Model service
/chat	AI energy assistant	Agent service

Data Flow

Unified API Access

participant ems_hmi
participant server
database timeseries
participant model
participant agent

ems_hmi -> server: GET /historical
server -> timeseries: query historical data
timeseries -> server: timeseries results
server -> ems_hmi: historical data

ems_hmi -> server: GET /predictions
server -> model: request forecast
model -> server: ML predictions
server -> ems_hmi: forecast data

ems_hmi -> server: POST /chat
server -> agent: process query
agent -> server: AI response
server -> ems_hmi: chat response

This repo is the architectural grouping — it has no runtime process of its own. The running service is ems-analyst-server, which composes model and agent into a single FastAPI process.