Get prices from your own sources¶

microgrid now values your plant at this moment, reading each input from the provider you name in a sources/1 file. You need the sources extra (Install), a plant/2 file for your plant (Describe your plant), and a source for each input.

1. Name a source for each input¶

Put sources.toml beside your plant.toml. Each input names its provider: a tool's integration in Home Assistant, or any entity whose attribute holds the forecast. This one reads prices from an entity, the PV forecast from Solcast's integration and the load forecast from EMHASS:

schema = "sources/1"
plant = "plant.toml"

[home_assistant]
url = "http://homeassistant.local:8123"

[prices]
from = "entity"
entity = "sensor.electricity_rates"
attribute = "rates"
keys = { import = "buy", export = "sell" }
unit = "EUR/kWh"

[pv_forecast]
from = "solcast"
entities = [
    "sensor.solcast_pv_forecast_forecast_today",
    "sensor.solcast_pv_forecast_forecast_tomorrow",
    "sensor.solcast_pv_forecast_forecast_day_3",
]
last_polled = "sensor.solcast_pv_forecast_api_last_polled"

[load_forecast]
from = "emhass"

Use your own entity names: none is assumed. Put a long-lived access token in the environment, never in the file:

export MICROGRID_HA_TOKEN=...      # or the variable [home_assistant] token_env names

2. Check what it reads¶

uv run microgrid sources --sources sources.toml

sources.toml: the plant plant.toml, the engine defaults
  yes  load_forecast  emhass
  yes  prices         entity
  yes  pv_forecast    solcast
  no   soc            no source configured

One line per input: yes and its provider, or no and why. This plant has a battery, so it needs the SOC too, which no forecast gives: that takes a [measured] section.

3. Add what is measured¶

The SOC, and whether the plant runs as modelled, come from the plant's monitoring. sources/1 reads them from Home Assistant:

[measured]
from = "home_assistant"
soc = "sensor.battery_state_of_charge"
dispatch_as_modelled = true

dispatch_as_modelled = true says the plant always runs its own dispatch. If it switches modes, name the mode entity instead:

dispatch = "select.inverter_mode"
modelled_modes = ["Self consumption"]

4. Swap a provider¶

Each input takes whichever provider you have: prices from TariffKit's integration (from = "tariffkit") or an entity; the PV forecast from Solcast's integration or an entity; the load forecast from EMHASS, Predbat or an entity. Configure a source covers each one.

5. Value now¶

uv run microgrid now --sources sources.toml --slots 3

With step 1 alone, for a plant with a battery:

not read: soc: no source configured
valued at 2026-10-17 21:17 CEST: fallback_import, 288 slots, degraded: battery_soc, work_mode
slot                     import   export  marginal  surplus kW  regime
2026-10-17 21:10 CEST    0.3000   0.0700    0.3000       0.000  import
2026-10-17 21:20 CEST    0.3000   0.0700    0.3000       0.000  import
2026-10-17 21:30 CEST    0.3000   0.0700    0.3000       0.000  import

The input that could not be read is named on standard error, and the valuation falls back to the import price (fallback_import), naming its causes (battery_soc, and work_mode for dispatch as modelled) rather than guessing. With [measured] in place, the same command gives the model's valuation (stored_value_v1), printed as value prints one (Run your first valuation).

To keep what was read, add --save DIR: it writes the inputs as a dataset with the plant's and engine's files. Re-run it with microgrid value --data DIR --at T, where T is the time of the read (the prices entry's received_at).

Next¶

For the scoring commands, microgrid fetch writes what was measured over a span as a dataset (the commands).
To publish prices continuously: Run it continuously.