Skip to content

readme.md

Latest commit

 

History

History
85 lines (57 loc) · 3.81 KB

readme.md

File metadata and controls

85 lines (57 loc) · 3.81 KB
Raw

⚠️ Archived — moved to monorepo.

This package now lives at packages/server/ inside arcnode-io/ems-analyst. History preserved via git subtree. This repo is read-only.

EMS Analyst Server 🛰️

The running FastAPI process that composes analyst-model and analyst-agent behind a single HTTP interface

Architecture

Principle: the agent is just another client of the server, same as the HMI. Both speak the same REST contracts. Any data path added for the HMI is available to the agent as a tool, and vice versa.

API Endpoints

GET /measurements

Hourly-bucketed gap-filled timeseries from the canonical measurements table.

Params: device_id (str), measurement (str), start (ISO-8601), end (ISO-8601), aggregation (mean|max|min|last, default mean). Response: { site_id, device_id, measurement, unit, points: [{ ts, value|null }] }.

GET /description

Inventory of distinct (device, measurement) pairs in the measurements table, with sample counts. The historian-side discovery surface — agents call this to learn exact measurement names (incl. non-device series like market price feeds) before querying. Distinct from the device-api DTM, which describes installed equipment.

Response: { site_id, pairs: [{ device_id, measurement, samples }] }.

GET /forecast

Query the forecasts table (populated by ems-analyst-model's scoring step).

Params: measurement (str), start (ISO-8601), end (ISO-8601). Response: { site_id, measurement, unit, model_name, model_version, points: [{ forecast_for, value }] }.

POST /analyst/chat

Multi-turn analyst chat. Delegates to ems_analyst_agent.Agent.chat_turn; persists thread in Postgres. See HMI handoff for body shape.

Project Structure

├── pyproject.toml              # Dependencies and build config
├── src/
│   ├── main.py                 # FastAPI application entry point
│   ├── app.py                  # Application factory and configuration
│   ├── models.py               # Pydantic request/response models
│   ├── database.py             # Database connections
│   ├── historical/
│   │   ├── historical_controller.py    # Historical data endpoints
│   │   └── historical_service.py       # TimescaleDB service
│   ├── predictions/
│   │   └── predictions_controller.py   # Prediction endpoints (MLflow integration)
│   └── chat/
│       └── chat_controller.py          # Chat completion endpoints (agent integration)
├── tests/
│   └── test_integration.py     # FastAPI TestClient integration tests
└── README.md                   # This file

Configuration

Environment Variables

  • TIMESCALEDB_URL: TimescaleDB connection string

  • MLFLOW_TRACKING_URI: MLflow server URL

  • AGENT_CONFIG: Agent package configuration

Dependencies

  • energy-analyst-agent: Installed as Python package for chat completions

  • FastAPI: Web framework with automatic OpenAPI documentation

  • TimescaleDB: Time-series database for historical data

  • MLflow: Model registry and serving