diff --git a/README.md b/README.md
index 9b81888..52bc349 100644
--- a/README.md
+++ b/README.md
@@ -34,7 +34,7 @@ doing inference of this value.
 
 The approach this project takes is to use pytorch to capture the relationships in the [Globe At Night
 dataset](https://globeatnight.org/maps-data/) and use that to predict sky brightness for H3
-cells at a configured H3 resoultion (default `0`).
+cells at a configured [H3 resoultion](https://h3geo.org/docs/core-library/restable/) (default `0`).
 
 ## running with docker
 
@@ -77,9 +77,8 @@ stars visible)
 
 ## changing the resolution
 
-By default, the producer runs over [resolution 0 cells](https://h3geo.org/docs/core-library/restable/),
-but this can be adjusted by setting the environment of the `producer` container in
-`docker-compose.yaml`:
+By default, the producer runs over all resolution 0 cells, but this can be adjusted
+by setting the environment of the `producer` container in `docker-compose.yaml`:
 
 ```yaml
 producer:
diff --git a/api/README.md b/api/README.md
index 04d9322..fc28052 100644
--- a/api/README.md
+++ b/api/README.md
@@ -27,14 +27,3 @@ with grpc.insecure_channel(f"{host}:{port}") as channel:
     print(response)
 
 ```
-
-## model
-
-### building and training
-
-The api depends on a model being trained from csv data.
-
-The following commands will generate a new `model.pth` (i.e. the learned parameters):
-
-- `python -m api.model.build` to write the csv that the model trains on
-- `python -m api.model.train` to train on the data in the csv
diff --git a/api/api/config.py b/api/api/config.py
index bb60134..66d973c 100644
--- a/api/api/config.py
+++ b/api/api/config.py
@@ -1,4 +1,3 @@
 import os
 
-log_level = int(os.getenv("LOG_LEVEL", 20))
 service_port = int(os.getenv("SERVICE_PORT", 50051))
diff --git a/api/api/main.py b/api/api/main.py
index ea9e530..2f2d87b 100644
--- a/api/api/main.py
+++ b/api/api/main.py
@@ -3,17 +3,18 @@
 
 import grpc
 
+from .config import service_port
 from .stubs import brightness_service_pb2_grpc
 from .service.brightness_servicer import BrightnessServicer
-from .config import log_level, service_port
 
 log_format = "%(asctime)s [%(levelname)s] %(message)s"
-logging.basicConfig(level=log_level, format=log_format)
+logging.basicConfig(level=logging.INFO, format=log_format)
 
 log = logging.getLogger(__name__)
 
 
 def serve():
+    log.info(f"starting up gRPC server on port {service_port}")
     server = grpc.server(futures.ThreadPoolExecutor(max_workers=10))
     brightness_service_pb2_grpc.add_BrightnessServiceServicer_to_server(BrightnessServicer(), server)
     server.add_insecure_port(f"[::]:{service_port}")
@@ -22,5 +23,4 @@ def serve():
 
 
 if __name__ == "__main__":
-    log.info(f"starting up gRPC server on port {service_port}")
     serve()
diff --git a/api/model.md b/api/model.md
new file mode 100644
index 0000000..ebb9c82
--- /dev/null
+++ b/api/model.md
@@ -0,0 +1,8 @@
+# model
+
+## building and training
+
+The api depends on a model being trained from csv data.
+
+1. use `python -m api.model.build` to write the csv that the model trains on
+2. use `python -m api.model.train` to generate a new `model.pth` (i.e. to learn new parameters)
diff --git a/docker-compose.yml b/docker-compose.yml
index ec2f3be..5cc99ad 100644
--- a/docker-compose.yml
+++ b/docker-compose.yml
@@ -36,9 +36,6 @@ services:
     build: ./api
     ports:
       - "50051:50051"
-    environment:
-      API_VERSION: "v1"
-      LOG_LEVEL: 30
     restart: on-failure
     depends_on:
       openmeteo: