Offline, approval-gated autonomous energy operations for African facilities on an 8 GB CPU laptop.
# SIMBA Autonomous Energy Operations Agent
SIMBA is an offline, approval-gated energy operations agent for African
campuses, hospitals, farms and small enterprises. An operator sets a demand
limit; SIMBA observes the facility network, forecasts a possible peak, ranks
safe non-critical interventions, asks for approval, applies approved actions to
a software plant, verifies the response and replans when conditions change.
The public prototype never switches real electrical equipment.
## What it does
`OBSERVE → FORECAST → PLAN → SAFETY CHECK → APPROVAL → EMULATE → VERIFY → REPLAN`
The included resilience demonstration deliberately creates one of two
complications:
- a selected device becomes unavailable; or
- an approved action produces less reduction than expected.
SIMBA blocks or detects the problem, excludes the failed resource, creates a
new safety-cleared plan, inherits approval only when the new actions remain
inside the approved limits, and verifies the final campus demand.
The default public scenario starts at 1,280.81 kVA with a 1,203.961 kVA limit.
Both resilience paths finish below the limit, record zero critical-load
actions, and keep live electrical control disabled.
## Offline African-language interface
The complete agent workflow, controls, safety boundary and mission outcome can
be viewed in:
- English (`en`)
- ChiShona (`sn`)
- isiNdebele (`nd`)
- Kiswahili (`sw`)
- isiZulu (`zu`)
Translations are bundled locally in `dashboard/static/i18n.js`. Selecting a
language makes no network request and does not use the LLM. This deterministic
localisation keeps safety wording stable on constrained, intermittently
connected sites.
## Run on Ubuntu 22.04
Target profile: x86-64, four CPU cores, 8 GB RAM, integrated graphics and no
cloud dependency during operation.
```bash
chmod +x setup_ubuntu.sh run_simba_demo.sh download_model.sh
./setup_ubuntu.sh
./run_simba_demo.sh
```
The launcher opens `
127.0.0.1`. Choose a resilience t …