kwwhat
kwwhat demo
Run the full kwwhat analytics pipeline locally — no cloud account needed.
Quickstart
# 1. Copy the environment template and add your API key
cp .env.example .env
# edit .env and set ANTHROPIC_API_KEY=...
# 2. Run the demo
./run-demo.sh
That’s it. The script will build the images, load the data, run the pipeline, and open the chat interface at http://localhost:5005.
Prerequisites
- Docker Desktop installed and running
This demo spins up three services via Docker Compose:
| Service | What it does |
|---|---|
duckdb-init |
Loads the sample OCPP log data into a local database |
dbt |
Runs the kwwhat dbt pipeline, transforming raw logs into analytics tables |
chat-bi |
Opens an AI chat interface so you can ask questions about the data |
- An Anthropic API key (powers the chat interface)
What happens under the hood
seeds/ demo/
ocpp_1_6_synthetic_logs_14d.csv ──▶ duckdb-init ──▶ raw.duckdb
ports.csv │
▼
dbt build
│
▼
analytics.duckdb
(fact_*, dim_*)
│
▼
nao chat
- duckdb-init reads the two CSV files in
seeds/and loads them intoraw.duckdb— a file that acts as the raw data warehouse. - dbt picks up from there, runs all the staging → intermediate → mart models, and writes the results to
analytics.duckdb. - chat-bi connects to both the analytics database and the dbt service, so you can ask plain-English questions like:
- “Report on reliability of my EV charging network”
- “What was the charge attempt success rate last week?”
- “Which chargers had the most downtime?”
- “Show me visit success by location.”
Sample data
The seeds/ folder contains synthetic OCPP 1.6 logs donated for this project:
| File | Contents |
|---|---|
ocpp_1_6_synthetic_logs_14d.csv |
14 days of synthetic EV charger messages (~2 MB) |
ports.csv |
Reference data: charger ports, connector types, and commission dates |
Manual commands
If you prefer to run services individually:
# Build images
docker compose build
# Load seed data (runs once and exits)
docker compose up duckdb-init
# Run dbt pipeline
docker compose up dbt
# Open chat interface (interactive) — available at http://localhost:5005
docker compose run --rm -p 5005:5005 -p 8005:8005 chat-bi
# Stop everything
docker compose down
# Stop and wipe the database volume
docker compose down -v
Running tests
Tests live in chat-bi/tests/ and run against the live chat server. With the demo already running, open a terminal and exec into the container:
docker exec -it $(docker ps -q --filter "publish=5005") bash
cd /app/kwwhat
nao test -m anthropic:claude-sonnet-4-6
You’ll be prompted to log in with the account you created at http://localhost:5005. Results (pass/fail, tokens, cost, latency) are saved to /app/kwwhat/tests/outputs/.
You should see
Test Results
┏━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━┳━━━━━━━━━┳━━━━━━━━┳━━━━━━━━━━━┳━━━━━━━━━━┳━━━━━━━┓
┃ Test ┃ Model ┃ Status ┃ Message ┃ Tokens ┃ Cost ┃ Time (s) ┃ Tools ┃
┡━━━━━━━━━━━━━╇━━━━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━╇━━━━━━━━━╇━━━━━━━━╇━━━━━━━━━━━╇━━━━━━━━━━╇━━━━━━━┩
│ total_ports │ anthropic:claude-sonnet-4-6 │ ✓ │ match │ 16253 │ 0.0246861 │ 12.2 │ 5 │
│ │ │ │ │ 16253 │ $0.0247 │ 12 │ 5 │
└─────────────┴─────────────────────────────┴────────┴─────────┴────────┴───────────┴──────────┴───────┘
To add more tests, create a YAML file in chat-bi/tests/:
name: my_test
prompt: "Your natural language question here"
sql: |
SELECT ... -- reference query whose results the agent must match
Project structure
demo/
├── README.md
├── docker-compose.yml
├── run-demo.sh
├── .env.example
├── seeds/ ← sample OCPP CSV data
├── duckdb-init/ ← Service 1: loads seeds into raw.duckdb
├── dbt/ ← Service 2: dbt build + MCP server
└── chat-bi/ ← Service 3: nao chat interface