diff --git a/serverless/aws/README.md b/serverless/aws/README.md
index f11e701..7f44c5e 100644
--- a/serverless/aws/README.md
+++ b/serverless/aws/README.md
@@ -5,10 +5,12 @@
 
 Upload the resulting `lambda_function.zip` using the Lambda console.
 
-## Restrictions
 
-1. There is a limit of 1 MB for tiles served through Lambda@Edge.
-2. Lambda@Edge does not support layers, environment variables, or ARM functions.
+## Configuration
+
+* `BUCKET`: the S3 bucket name.
+* `PMTILES_PATH`
+* `TILE_PATH`
 
 ## AWS Notes
 
@@ -17,28 +19,19 @@ Upload the resulting `lambda_function.zip` using the Lambda console.
 
 ## Test Event
 
-CloudFront event:
-
-```json
-{
-  "Records": [
-    {
-      "cf": {
-        "request": {
-          "uri": "/my-tileset-name/0/0/0.pbf",
-          "method": "GET",
-          "headers": {}
-        }
-      }
-    }
-  ]
-}
-```
-
 API Gateway V2 / Lambda Function URLs:
 
 ```json
 {
   "rawPath": "/my-tileset-name/0/0/0.pbf"
 }
-```
\ No newline at end of file
+```
+
+
+### Monitoring
+
+
+
+### Lambda@Edge
+
+Lambda@Edge's multi-region features have little benefit when fetching data from S3 in a single region, and Lambda@Edge [doesn't support](https://docs.aws.amazon.com/AmazonCloudFront/latest/DeveloperGuide/edge-functions-restrictions.html) environment variables or responses over 1 MB. For globally distributed caching, use CloudFront in combination with Lambda Function URLs.
\ No newline at end of file
diff --git a/serverless/aws/create_lambda_function.py b/serverless/aws/create_lambda_function.py
index f95f89a..b3a07c1 100644
--- a/serverless/aws/create_lambda_function.py
+++ b/serverless/aws/create_lambda_function.py
@@ -1,7 +1,11 @@
 import zipfile
+import subprocess
+
+sha = subprocess.check_output(["git", "describe", "--always"]).strip()
 
 with zipfile.ZipFile("lambda_function.zip", "w", zipfile.ZIP_DEFLATED) as z:
     z.write("lambda_function.py")
     z.write("../../python/pmtiles/reader.py", "pmtiles/reader.py")
+    z.writestr("version",sha)
 
 print(f"created lambda_function.zip")
diff --git a/serverless/aws/lambda_function.py b/serverless/aws/lambda_function.py
index 8a644de..df03eb1 100644
--- a/serverless/aws/lambda_function.py
+++ b/serverless/aws/lambda_function.py
@@ -8,6 +8,7 @@ import re
 
 # Exists inside all lambda functions
 import boto3
+from botocore.exceptions import ClientError
 
 # create_lambda_function.py will vendor the relevant file
 from pmtiles.reader import Reader
@@ -16,7 +17,6 @@ Zxy = collections.namedtuple("Zxy", ["z", "x", "y"])
 
 s3 = boto3.client("s3")
 
-
 # Given a 512MB lambda function, use half of the memory for the cache,
 # assuming the average root/leaf/tile size is 512 KB
 @lru_cache(maxsize=500)
@@ -64,7 +64,7 @@ def lambda_handler(event, context):
     name, tile = parse_tile_path(os.environ.get("TILE_PATH"), uri)
 
     if not tile:
-        return {"statusCode": 400, "body": "Invalid Tile URL"}
+        return {"statusCode": 400, "body": "Invalid tile URL"}
 
     def get_bytes(offset, length):
         return get_object_bytes(
@@ -72,10 +72,22 @@ def lambda_handler(event, context):
         )
 
     reader = Reader(get_bytes)
-    tile_data = reader.get(tile.z, tile.x, tile.y)
-    if not tile_data:
+    minzoom = int(reader.header().metadata["minzoom"])
+    maxzoom = int(reader.header().metadata["maxzoom"])
+    if tile.z < minzoom or tile.z > maxzoom:
         return {"statusCode": 404, "body": "Tile not found"}
 
+    try:
+        tile_data = reader.get(tile.z, tile.x, tile.y)
+        if not tile_data:
+            return {"statusCode": 204}
+    except ClientError as e:
+        error_code = e.response["Error"]["Code"]
+        if error_code == "AccessDenied":
+            return {"statusCode": 404, "body": "Archive not found"}
+        else:
+            raise e
+
     # CloudFront requires decompressed responses from lambda
     # in order to implement the Compressed CacheOptimized policy correctly
     # as well as Brotli support