diff --git a/.github/workflows/main.yml b/.github/workflows/main.yml
index 2b8a46a..8c8eb06 100644
--- a/.github/workflows/main.yml
+++ b/.github/workflows/main.yml
@@ -36,6 +36,7 @@ jobs:
           - playlist_manager
           - chat_archiver
           - zulip_bots
+          - bus_analyzer
     steps:
       - name: Check out repo
         uses: actions/checkout@v2
diff --git a/bus_analyzer/Dockerfile b/bus_analyzer/Dockerfile
new file mode 100644
index 0000000..9f643b9
--- /dev/null
+++ b/bus_analyzer/Dockerfile
@@ -0,0 +1,14 @@
+FROM alpine:3.14
+# dependencies needed for compiling c extensions
+# also busybox-extras for telnet for easier use of backdoor
+RUN apk --update add py3-pip g++ python3-dev libffi-dev musl-dev file make busybox-extras
+
+# Install gevent so that we don't need to re-install it when common changes
+RUN pip install gevent
+
+# Actual application
+COPY bus_analyzer /tmp/bus_analyzer
+RUN pip install /tmp/bus_analyzer && rm -r /tmp/bus_analyzer
+
+LABEL org.opencontainers.image.source https://github.com/dbvideostriketeam/wubloader
+ENTRYPOINT ["python3", "-m", "bus_analyzer"]
diff --git a/bus_analyzer/bus_analyzer/__init__.py b/bus_analyzer/bus_analyzer/__init__.py
new file mode 100644
index 0000000..e69de29
diff --git a/bus_analyzer/bus_analyzer/__main__.py b/bus_analyzer/bus_analyzer/__main__.py
new file mode 100644
index 0000000..c722e54
--- /dev/null
+++ b/bus_analyzer/bus_analyzer/__main__.py
@@ -0,0 +1,16 @@
+
+import gevent.monkey
+gevent.monkey.patch_all()
+
+import logging
+import os
+
+import argh
+
+from .main import cli
+
+LOG_FORMAT = "[%(asctime)s] %(levelname)8s %(name)s(%(module)s:%(lineno)d): %(message)s"
+
+level = os.environ.get('WUBLOADER_LOG_LEVEL', 'INFO').upper()
+logging.basicConfig(level=level, format=LOG_FORMAT)
+cli()
diff --git a/bus_analyzer/bus_analyzer/extract.py b/bus_analyzer/bus_analyzer/extract.py
new file mode 100644
index 0000000..b5a89e5
--- /dev/null
+++ b/bus_analyzer/bus_analyzer/extract.py
@@ -0,0 +1,3 @@
+
+def extract_segment(path):
+	raise NotImplementedError
diff --git a/bus_analyzer/bus_analyzer/main.py b/bus_analyzer/bus_analyzer/main.py
new file mode 100644
index 0000000..0f2a509
--- /dev/null
+++ b/bus_analyzer/bus_analyzer/main.py
@@ -0,0 +1,177 @@
+
+import datetime
+import logging
+import os
+import signal
+import traceback
+
+import argh
+import gevent.event
+
+from common import database
+from common.segments import parse_segment_path
+
+from .extract import extract_segment
+
+
+cli = argh.EntryPoint()
+
+
+@cli
+@argh.named("extract-segment")
+def do_extract_segment(*segment_paths):
+	"""Extract info from individual segments and print them"""
+	for segment_path in segment_paths:
+		odometer = extract_segment(segment_path)
+		print(f"{segment_path} {odometer}")
+
+
+@cli
+@argh.named("analyze-segment")
+def do_analyze_segment(dbconnect, *segment_paths, base_dir='.'):
+	"""Analyze individual segments and write them to the database"""
+	dbmanager = database.DBManager(dsn=dbconnect)
+	conn = dbmanager.get_conn()
+
+	for segment_path in segment_paths:
+		analyze_segment(conn, segment_path)
+
+
+def analyze_segment(conn, segment_path, check_segment_name=None):
+	segment_info = parse_segment_path(segment_path)
+	segment_name = '/'.join(segment_path.split('/')[-4:]) # just keep last 4 path parts
+	if check_segment_name is not None:
+		assert segment_name == check_segment_name
+
+	try:
+		odometer = extract_segment(segment_path)
+	except Exception:
+		logging.warning(f"Failed to extract segment {segment_path!r}", exc_info=True)
+		odometer = None
+		error = traceback.format_exc()
+	else:
+		logging.info(f"Got odometer = {odometer} for segment {segment_path!r}")
+		error = None
+
+	database.query(
+		conn,
+		"""
+			INSERT INTO bus_data (channel, timestamp, segment, error, odometer)
+			VALUES (%(channel)s, %(timestamp)s, %(segment)s, %(error)s, %(odometer)s)
+			ON CONFLICT (channel, timestamp, segment) DO UPDATE
+				SET error = %(error)s,
+					odometer = %(odometer)s
+		""",
+		channel=segment_info.channel,
+		timestamp=segment_info.start,
+		segment=segment_name,
+		error=error,
+		odometer=odometer,
+	)
+
+
+def analyze_hour(conn, existing_segments, base_dir, channel, quality, hour):
+	hour_path = os.path.join(base_dir, channel, quality, hour)
+	try:
+		segments = os.listdir(hour_path)
+	except FileNotFoundError:
+		logging.info(f"No such hour {hour_path!r}, skipping")
+		return
+
+	logging.info("Found {} segments for hour {!r}".format(len(segments), hour_path))
+	segments_to_do = []
+	for segment in segments:
+		# Format as relative path from basedir, this is the format the DB expects.
+		segment_name = os.path.join(channel, quality, hour, segment)
+		if segment_name in existing_segments:
+			continue
+
+		segment_path = os.path.join(base_dir, segment_name)
+		assert segment_path == os.path.join(hour_path, segment)
+
+		segments_to_do.append((segment_path, segment_name))
+
+	logging.info("Found {} segments not already existing".format(len(segments_to_do)))
+	for segment_path, segment_name in segments_to_do:
+		analyze_segment(conn, segment_path, segment_name)
+
+
+def parse_hours(s):
+	try:
+		return int(s)
+	except ValueError:
+		return s.split(",")
+
+
+@cli
+@argh.arg("--hours", type=parse_hours, help="If integer, watch the most recent N hours. Otherwise, comma-seperated list of hours.")
+def main(
+	dbconnect,
+	*channels,
+	base_dir='.',
+	quality='source',
+	hours=2,
+	run_once=False,
+	overwrite=False,
+):
+	CHECK_INTERVAL = 2
+
+	stopping = gevent.event.Event()
+
+	gevent.signal_handler(signal.SIGTERM, stopping.set)
+
+	db_manager = database.DBManager(dsn=dbconnect)
+	conn = db_manager.get_conn()
+
+	logging.info("Started")
+
+	while not stopping.is_set():
+		start_time = datetime.datetime.utcnow()
+
+		# If we aren't using a hard-coded hours list, work out hours based on current time
+		if isinstance(hours, int):
+			do_hours = [
+				(start_time - datetime.timedelta(hours=hours_ago)).strftime("%Y-%m-%dT%H")
+				for hours_ago in range(hours)
+			]
+		else:
+			do_hours = hours
+
+		# Unless we're overwriting, fetch a list of existing segments from the database.
+		# We can optimize a little here by restricting to the channels and hour range we need.
+		if overwrite:
+			existing_segments = set()
+		else:
+			start = datetime.datetime.strptime(min(do_hours), "%Y-%m-%dT%H")
+			end = datetime.datetime.strptime(max(do_hours), "%Y-%m-%dT%H")
+			logging.info("Fetching existing segments from {} to {} for {}".format(
+				start,
+				end,
+				", ".join(channels),
+			))
+			result = database.query(conn, """
+				SELECT segment
+				FROM bus_data
+				WHERE channel IN %(channels)s
+					AND timestamp >= %(start)s::timestamp
+					AND timestamp < %(end)s::timestamp + interval '1 hour'
+					AND segment IS NOT NULL
+			""", channels=channels, start=start, end=end)
+			existing_segments = {segment for (segment,) in result.fetchall()}
+			logging.info("Found {} existing segments".format(len(existing_segments)))
+
+		for channel in channels:
+			for hour in do_hours:
+				analyze_hour(conn, existing_segments, base_dir, channel, quality, hour)
+
+		if run_once:
+			logging.info("Requested to only run once, stopping")
+			return
+
+		elapsed = (datetime.datetime.utcnow() - start_time).total_seconds()
+		remaining = CHECK_INTERVAL - elapsed
+		if remaining > 0:
+			logging.info(f"Sleeping {remaining} until next check")
+			stopping.wait(remaining)
+
+	logging.info("Gracefully stopped")
diff --git a/bus_analyzer/setup.py b/bus_analyzer/setup.py
new file mode 100644
index 0000000..a9101bd
--- /dev/null
+++ b/bus_analyzer/setup.py
@@ -0,0 +1,18 @@
+from setuptools import setup, find_packages
+
+setup(
+	name='bus_analyzer',
+	version='0.0.1',
+	author='DB Video Strike Team',
+	author_email='dbvideostriketeam@gmail.com',
+	description='',
+	packages=find_packages(),
+	install_requires=[
+		"argh==0.28.1",
+		"gevent",
+        "psycogreen",
+        "psycopg2",
+		"python-dateutil",
+		"wubloader-common",
+	],
+)
diff --git a/postgres/setup.sh b/postgres/setup.sh
index 793e0b2..ab31c08 100644
--- a/postgres/setup.sh
+++ b/postgres/setup.sh
@@ -156,17 +156,20 @@ CREATE TABLE playlists (
 -- The "error" column records a free-form human readable message about why a value could not
 -- be determined.
 -- The odometer column is in miles. The game shows the odometer to the 1/10th mile precision.
+-- The segment may be NULL, which indicates a manually-inserted value.
+-- The primary key serves two purposes:
+--   It provides an index on channel, followed by a range index on timestamp
+--   It provides a unique constraint on the same segment and timestamp
+-- Note that multiple manual records may exist for the same channel and timestamp
+-- as all NULL values are considered distinct, so the unique constraint does not hold.
 CREATE TABLE bus_data (
-	timestamp TIMESTAMP NOT NULL,
 	channel TEXT NOT NULL,
+	timestamp TIMESTAMP NOT NULL,
 	segment TEXT,
 	error TEXT,
 	odometer DOUBLE PRECISION,
+	PRIMARY KEY (channel, timestamp, segment)
 );
-
--- Range index on timestamp as we will often want the closest timestamp to a requested point.
--- Note btree is the default anyway but we use it explicitly here as we want the range behaviour.
-CREATE INDEX bus_data_timestamp ON bus_data USING btree (timestamp);
 EOSQL
 
 if [ -a /mnt/wubloader/nodes.csv ]; then