diff --git a/CHANGELOG.md b/CHANGELOG.md
index 2e17cd60..c86761c4 100644
--- a/CHANGELOG.md
+++ b/CHANGELOG.md
@@ -12,6 +12,7 @@ and this project adheres to [Semantic Versioning](https://semver.org/spec/v2.0.0
 * Web API for adding & querying build logs
 * CLI commands to access build logs API
 * Cron build logs are uploaded to above API
+* Proper ASCII table output in CLI
 
 ### Changed
 
@@ -20,6 +21,7 @@ and this project adheres to [Semantic Versioning](https://semver.org/spec/v2.0.0
 * Official Arch packages are now split between `vieter` & `vieter-git`
     * `vieter` is the latest release
     * `vieter-git` is the latest commit on the dev branch
+* Full refactor of Docker socket code
 
 ## [0.3.0-alpha.1](https://git.rustybever.be/vieter/vieter/src/tag/0.3.0-alpha.1)
 
diff --git a/src/build/build.v b/src/build/build.v
index 41f68e28..2784c260 100644
--- a/src/build/build.v
+++ b/src/build/build.v
@@ -6,10 +6,13 @@ import time
 import os
 import db
 import client
+import strings
+import util
 
-const container_build_dir = '/build'
-
-const build_image_repo = 'vieter-build'
+const (
+	container_build_dir = '/build'
+	build_image_repo    = 'vieter-build'
+)
 
 // create_build_image creates a builder image given some base image which can
 // then be used to build & package Arch images. It mostly just updates the
@@ -17,6 +20,12 @@ const build_image_repo = 'vieter-build'
 // makepkg with. The base image should be some Linux distribution that uses
 // Pacman as its package manager.
 pub fn create_build_image(base_image string) ?string {
+	mut dd := docker.new_conn()?
+
+	defer {
+		dd.close() or {}
+	}
+
 	commands := [
 		// Update repos & install required packages
 		'pacman -Syu --needed --noconfirm base-devel git'
@@ -46,14 +55,15 @@ pub fn create_build_image(base_image string) ?string {
 	image_tag := if image_parts.len > 1 { image_parts[1] } else { 'latest' }
 
 	// We pull the provided image
-	docker.pull_image(image_name, image_tag)?
+	dd.pull_image(image_name, image_tag)?
 
-	id := docker.create_container(c)?
-	docker.start_container(id)?
+	id := dd.create_container(c)?.id
+	// id := docker.create_container(c)?
+	dd.start_container(id)?
 
 	// This loop waits until the container has stopped, so we can remove it after
 	for {
-		data := docker.inspect_container(id)?
+		data := dd.inspect_container(id)?
 
 		if !data.state.running {
 			break
@@ -67,8 +77,8 @@ pub fn create_build_image(base_image string) ?string {
 	// TODO also add the base image's name into the image name to prevent
 	// conflicts.
 	tag := time.sys_mono_now().str()
-	image := docker.create_image_from_container(id, 'vieter-build', tag)?
-	docker.remove_container(id)?
+	image := dd.create_image_from_container(id, 'vieter-build', tag)?
+	dd.remove_container(id)?
 
 	return image.id
 }
@@ -85,6 +95,12 @@ pub:
 // provided GitRepo. The base image ID should be of an image previously created
 // by create_build_image. It returns the logs of the container.
 pub fn build_repo(address string, api_key string, base_image_id string, repo &db.GitRepo) ?BuildResult {
+	mut dd := docker.new_conn()?
+
+	defer {
+		dd.close() or {}
+	}
+
 	build_arch := os.uname().machine
 
 	// TODO what to do with PKGBUILDs that build multiple packages?
@@ -112,27 +128,31 @@ pub fn build_repo(address string, api_key string, base_image_id string, repo &db
 		user: 'builder:builder'
 	}
 
-	id := docker.create_container(c)?
-	docker.start_container(id)?
+	id := dd.create_container(c)?.id
+	dd.start_container(id)?
 
-	mut data := docker.inspect_container(id)?
+	mut data := dd.inspect_container(id)?
 
 	// This loop waits until the container has stopped, so we can remove it after
 	for data.state.running {
 		time.sleep(1 * time.second)
 
-		data = docker.inspect_container(id)?
+		data = dd.inspect_container(id)?
 	}
 
-	logs := docker.get_container_logs(id)?
+	mut logs_stream := dd.get_container_logs(id)?
 
-	docker.remove_container(id)?
+	// Read in the entire stream
+	mut logs_builder := strings.new_builder(10 * 1024)
+	util.reader_to_writer(mut logs_stream, mut logs_builder)?
+
+	dd.remove_container(id)?
 
 	return BuildResult{
 		start_time: data.state.start_time
 		end_time: data.state.end_time
 		exit_code: data.state.exit_code
-		logs: logs
+		logs: logs_builder.str()
 	}
 }
 
@@ -150,7 +170,14 @@ fn build(conf Config, repo_id int) ? {
 	res := build_repo(conf.address, conf.api_key, image_id, repo)?
 
 	println('Removing build image...')
-	docker.remove_image(image_id)?
+
+	mut dd := docker.new_conn()?
+
+	defer {
+		dd.close() or {}
+	}
+
+	dd.remove_image(image_id)?
 
 	println('Uploading logs to Vieter...')
 	c.add_build_log(repo.id, res.start_time, res.end_time, build_arch, res.exit_code,
diff --git a/src/cron/daemon/daemon.v b/src/cron/daemon/daemon.v
index ade8fcbf..da3b46e2 100644
--- a/src/cron/daemon/daemon.v
+++ b/src/cron/daemon/daemon.v
@@ -253,14 +253,21 @@ fn (mut d Daemon) rebuild_base_image() bool {
 fn (mut d Daemon) clean_old_base_images() {
 	mut i := 0
 
+	mut dd := docker.new_conn() or {
+		d.lerror('Failed to connect to Docker socket.')
+		return
+	}
+
+	defer {
+		dd.close() or {}
+	}
+
 	for i < d.builder_images.len - 1 {
 		// For each builder image, we try to remove it by calling the Docker
 		// API. If the function returns an error or false, that means the image
 		// wasn't deleted. Therefore, we move the index over. If the function
 		// returns true, the array's length has decreased by one so we don't
 		// move the index.
-		if !docker.remove_image(d.builder_images[i]) or { false } {
-			i += 1
-		}
+		dd.remove_image(d.builder_images[i]) or { i += 1 }
 	}
 }
diff --git a/src/docker/README.md b/src/docker/README.md
new file mode 100644
index 00000000..4cc8971a
--- /dev/null
+++ b/src/docker/README.md
@@ -0,0 +1,3 @@
+This module implements part of the Docker Engine API v1.41
+([documentation](https://docs.docker.com/engine/api/v1.41/)) using socket-based
+HTTP communication.
diff --git a/src/docker/containers.v b/src/docker/containers.v
index 14ac12d9..0bc59bb3 100644
--- a/src/docker/containers.v
+++ b/src/docker/containers.v
@@ -3,17 +3,10 @@ module docker
 import json
 import net.urllib
 import time
+import net.http { Method }
 
-struct Container {
-	id    string   [json: Id]
-	names []string [json: Names]
-}
-
-// containers returns a list of all currently running containers
-pub fn containers() ?[]Container {
-	res := request('GET', urllib.parse('/v1.41/containers/json')?)?
-
-	return json.decode([]Container, res.text) or {}
+struct DockerError {
+	message string
 }
 
 pub struct NewContainer {
@@ -26,27 +19,37 @@ pub struct NewContainer {
 }
 
 struct CreatedContainer {
-	id string [json: Id]
+pub:
+	id       string   [json: Id]
+	warnings []string [json: Warnings]
 }
 
-// create_container creates a container defined by the given configuration. If
-// successful, it returns the ID of the newly created container.
-pub fn create_container(c &NewContainer) ?string {
-	res := request_with_json('POST', urllib.parse('/v1.41/containers/create')?, c)?
+// create_container creates a new container with the given config.
+pub fn (mut d DockerConn) create_container(c NewContainer) ?CreatedContainer {
+	d.send_request_with_json(Method.post, urllib.parse('/v1.41/containers/create')?, c)?
+	head, res := d.read_response()?
 
-	if res.status_code != 201 {
-		return error('Failed to create container.')
+	if head.status_code != 201 {
+		data := json.decode(DockerError, res)?
+
+		return error(data.message)
 	}
 
-	return json.decode(CreatedContainer, res.text)?.id
+	data := json.decode(CreatedContainer, res)?
+
+	return data
 }
 
-// start_container starts a container with a given ID. It returns whether the
-// container was started or not.
-pub fn start_container(id string) ?bool {
-	res := request('POST', urllib.parse('/v1.41/containers/$id/start')?)?
+// start_container starts the container with the given id.
+pub fn (mut d DockerConn) start_container(id string) ? {
+	d.send_request(Method.post, urllib.parse('/v1.41/containers/$id/start')?)?
+	head, body := d.read_response()?
 
-	return res.status_code == 204
+	if head.status_code != 204 {
+		data := json.decode(DockerError, body)?
+
+		return error(data.message)
+	}
 }
 
 struct ContainerInspect {
@@ -67,16 +70,18 @@ pub mut:
 	end_time   time.Time [skip]
 }
 
-// inspect_container returns the result of inspecting a container with a given
-// ID.
-pub fn inspect_container(id string) ?ContainerInspect {
-	res := request('GET', urllib.parse('/v1.41/containers/$id/json')?)?
+// inspect_container returns detailed information for a given container.
+pub fn (mut d DockerConn) inspect_container(id string) ?ContainerInspect {
+	d.send_request(Method.get, urllib.parse('/v1.41/containers/$id/json')?)?
+	head, body := d.read_response()?
 
-	if res.status_code != 200 {
-		return error('Failed to inspect container.')
+	if head.status_code != 200 {
+		data := json.decode(DockerError, body)?
+
+		return error(data.message)
 	}
 
-	mut data := json.decode(ContainerInspect, res.text)?
+	mut data := json.decode(ContainerInspect, body)?
 
 	data.state.start_time = time.parse_rfc3339(data.state.start_time_str)?
 
@@ -87,31 +92,31 @@ pub fn inspect_container(id string) ?ContainerInspect {
 	return data
 }
 
-// remove_container removes a container with a given ID.
-pub fn remove_container(id string) ?bool {
-	res := request('DELETE', urllib.parse('/v1.41/containers/$id')?)?
+// remove_container removes the container with the given id.
+pub fn (mut d DockerConn) remove_container(id string) ? {
+	d.send_request(Method.delete, urllib.parse('/v1.41/containers/$id')?)?
+	head, body := d.read_response()?
 
-	return res.status_code == 204
+	if head.status_code != 204 {
+		data := json.decode(DockerError, body)?
+
+		return error(data.message)
+	}
 }
 
-// get_container_logs retrieves the logs for a Docker container, both stdout &
-// stderr.
-pub fn get_container_logs(id string) ?string {
-	res := request('GET', urllib.parse('/v1.41/containers/$id/logs?stdout=true&stderr=true')?)?
-	mut res_bytes := res.text.bytes()
+// get_container_logs returns a reader object allowing access to the
+// container's logs.
+pub fn (mut d DockerConn) get_container_logs(id string) ?&StreamFormatReader {
+	d.send_request(Method.get, urllib.parse('/v1.41/containers/$id/logs?stdout=true&stderr=true')?)?
+	head := d.read_response_head()?
 
-	// Docker uses a special "stream" format for their logs, so we have to
-	// clean up the data.
-	mut index := 0
+	if head.status_code != 200 {
+		content_length := head.header.get(http.CommonHeader.content_length)?.int()
+		body := d.read_response_body(content_length)?
+		data := json.decode(DockerError, body)?
 
-	for index < res_bytes.len {
-		// The reverse is required because V reads in the bytes differently
-		t := res_bytes[index + 4..index + 8].reverse()
-		len_length := unsafe { *(&u32(&t[0])) }
-
-		res_bytes.delete_many(index, 8)
-		index += int(len_length)
+		return error(data.message)
 	}
 
-	return res_bytes.bytestr()
+	return d.get_stream_format_reader()
 }
diff --git a/src/docker/docker.v b/src/docker/docker.v
index 305e8253..ccc6bedd 100644
--- a/src/docker/docker.v
+++ b/src/docker/docker.v
@@ -1,97 +1,137 @@
 module docker
 
 import net.unix
-import net.urllib
+import io
 import net.http
+import strings
+import net.urllib
 import json
+import util
 
-const socket = '/var/run/docker.sock'
+const (
+	socket               = '/var/run/docker.sock'
+	buf_len              = 10 * 1024
+	http_separator       = [u8(`\r`), `\n`, `\r`, `\n`]
+	http_chunk_separator = [u8(`\r`), `\n`]
+)
 
-const buf_len = 1024
+pub struct DockerConn {
+mut:
+	socket &unix.StreamConn
+	reader &io.BufferedReader
+}
 
-// send writes a request to the Docker socket, waits for a response & returns
-// it.
-fn send(req &string) ?http.Response {
-	// Open a connection to the socket
-	mut s := unix.connect_stream(docker.socket) or {
-		return error('Failed to connect to socket ${docker.socket}.')
+// new_conn creates a new connection to the Docker daemon.
+pub fn new_conn() ?&DockerConn {
+	s := unix.connect_stream(docker.socket)?
+
+	d := &DockerConn{
+		socket: s
+		reader: io.new_buffered_reader(reader: s)
 	}
 
-	defer {
-		// This or is required because otherwise, the V compiler segfaults for
-		// some reason
-		// https://github.com/vlang/v/issues/13534
-		s.close() or {}
-	}
+	return d
+}
 
-	// Write the request to the socket
-	s.write_string(req) or { return error('Failed to write request to socket ${docker.socket}.') }
+// close closes the underlying socket connection.
+pub fn (mut d DockerConn) close() ? {
+	d.socket.close()?
+}
 
-	s.wait_for_write()?
+// send_request sends an HTTP request without body.
+pub fn (mut d DockerConn) send_request(method http.Method, url urllib.URL) ? {
+	req := '$method $url.request_uri() HTTP/1.1\nHost: localhost\n\n'
 
-	mut c := 0
-	mut buf := []u8{len: docker.buf_len}
+	d.socket.write_string(req)?
+
+	// When starting a new request, the reader needs to be reset.
+	d.reader = io.new_buffered_reader(reader: d.socket)
+}
+
+// send_request_with_body sends an HTTP request with the given body.
+pub fn (mut d DockerConn) send_request_with_body(method http.Method, url urllib.URL, content_type string, body string) ? {
+	req := '$method $url.request_uri() HTTP/1.1\nHost: localhost\nContent-Type: $content_type\nContent-Length: $body.len\n\n$body\n\n'
+
+	d.socket.write_string(req)?
+
+	// When starting a new request, the reader needs to be reset.
+	d.reader = io.new_buffered_reader(reader: d.socket)
+}
+
+// send_request_with_json<T> is a convenience wrapper around
+// send_request_with_body that encodes the input as JSON.
+pub fn (mut d DockerConn) send_request_with_json<T>(method http.Method, url urllib.URL, data &T) ? {
+	body := json.encode(data)
+
+	return d.send_request_with_body(method, url, 'application/json', body)
+}
+
+// read_response_head consumes the socket's contents until it encounters
+// '\r\n\r\n', after which it parses the response as an HTTP response.
+// Importantly, this function never consumes the reader past the HTTP
+// separator, so the body can be read fully later on.
+pub fn (mut d DockerConn) read_response_head() ?http.Response {
 	mut res := []u8{}
 
-	for {
-		c = s.read(mut buf) or { return error('Failed to read data from socket ${docker.socket}.') }
-		res << buf[..c]
+	util.read_until_separator(mut d.reader, mut res, docker.http_separator)?
 
-		if c < docker.buf_len {
-			break
-		}
-	}
-
-	// After reading the first part of the response, we parse it into an HTTP
-	// response. If it isn't chunked, we return early with the data.
-	parsed := http.parse_response(res.bytestr()) or {
-		return error('Failed to parse HTTP response from socket ${docker.socket}.')
-	}
-
-	if parsed.header.get(http.CommonHeader.transfer_encoding) or { '' } != 'chunked' {
-		return parsed
-	}
-
-	// We loop until we've encountered the end of the chunked response
-	// A chunked HTTP response always ends with '0\r\n\r\n'.
-	for res.len < 5 || res#[-5..] != [u8(`0`), `\r`, `\n`, `\r`, `\n`] {
-		// Wait for the server to respond
-		s.wait_for_write()?
-
-		for {
-			c = s.read(mut buf) or {
-				return error('Failed to read data from socket ${docker.socket}.')
-			}
-			res << buf[..c]
-
-			if c < docker.buf_len {
-				break
-			}
-		}
-	}
-
-	// Decode chunked response
 	return http.parse_response(res.bytestr())
 }
 
-// request_with_body sends a request to the Docker socket with the given body.
-fn request_with_body(method string, url urllib.URL, content_type string, body string) ?http.Response {
-	req := '$method $url.request_uri() HTTP/1.1\nHost: localhost\nContent-Type: $content_type\nContent-Length: $body.len\n\n$body\n\n'
+// read_response_body reads `length` bytes from the stream. It can be used when
+// the response encoding isn't chunked to fully read it.
+pub fn (mut d DockerConn) read_response_body(length int) ?string {
+	if length == 0 {
+		return ''
+	}
 
-	return send(req)
+	mut buf := []u8{len: docker.buf_len}
+	mut c := 0
+	mut builder := strings.new_builder(docker.buf_len)
+
+	for builder.len < length {
+		c = d.reader.read(mut buf) or { break }
+
+		builder.write(buf[..c])?
+	}
+
+	return builder.str()
 }
 
-// request sends a request to the Docker socket with an empty body.
-fn request(method string, url urllib.URL) ?http.Response {
-	req := '$method $url.request_uri() HTTP/1.1\nHost: localhost\n\n'
+// read_response is a convenience function which always consumes the entire
+// response & returns it. It should only be used when we're certain that the
+// result isn't too large.
+pub fn (mut d DockerConn) read_response() ?(http.Response, string) {
+	head := d.read_response_head()?
 
-	return send(req)
+	if head.header.get(http.CommonHeader.transfer_encoding) or { '' } == 'chunked' {
+		mut builder := strings.new_builder(1024)
+		mut body := d.get_chunked_response_reader()
+
+		util.reader_to_writer(mut body, mut builder)?
+
+		return head, builder.str()
+	}
+
+	content_length := head.header.get(http.CommonHeader.content_length)?.int()
+	res := d.read_response_body(content_length)?
+
+	return head, res
 }
 
-// request_with_json<T> sends a request to the Docker socket with a given JSON
-// payload
-pub fn request_with_json<T>(method string, url urllib.URL, data &T) ?http.Response {
-	body := json.encode(data)
+// get_chunked_response_reader returns a ChunkedResponseReader using the socket
+// as reader.
+pub fn (mut d DockerConn) get_chunked_response_reader() &ChunkedResponseReader {
+	r := new_chunked_response_reader(d.reader)
 
-	return request_with_body(method, url, 'application/json', body)
+	return r
+}
+
+// get_stream_format_reader returns a StreamFormatReader using the socket as
+// reader.
+pub fn (mut d DockerConn) get_stream_format_reader() &StreamFormatReader {
+	r := new_chunked_response_reader(d.reader)
+	r2 := new_stream_format_reader(r)
+
+	return r2
 }
diff --git a/src/docker/images.v b/src/docker/images.v
index 2e873fa0..6161565a 100644
--- a/src/docker/images.v
+++ b/src/docker/images.v
@@ -1,6 +1,6 @@
 module docker
 
-import net.http
+import net.http { Method }
 import net.urllib
 import json
 
@@ -9,26 +9,53 @@ pub:
 	id string [json: Id]
 }
 
-// pull_image pulls tries to pull the image for the given image & tag
-pub fn pull_image(image string, tag string) ?http.Response {
-	return request('POST', urllib.parse('/v1.41/images/create?fromImage=$image&tag=$tag')?)
-}
+// pull_image pulls the given image:tag.
+pub fn (mut d DockerConn) pull_image(image string, tag string) ? {
+	d.send_request(Method.post, urllib.parse('/v1.41/images/create?fromImage=$image&tag=$tag')?)?
+	head := d.read_response_head()?
 
-// create_image_from_container creates a new image from a container with the
-// given repo & tag, given the container's ID.
-pub fn create_image_from_container(id string, repo string, tag string) ?Image {
-	res := request('POST', urllib.parse('/v1.41/commit?container=$id&repo=$repo&tag=$tag')?)?
+	if head.status_code != 200 {
+		content_length := head.header.get(http.CommonHeader.content_length)?.int()
+		body := d.read_response_body(content_length)?
+		data := json.decode(DockerError, body)?
 
-	if res.status_code != 201 {
-		return error('Failed to create image from container.')
+		return error(data.message)
 	}
 
-	return json.decode(Image, res.text) or {}
+	// Keep reading the body until the pull has completed
+	mut body := d.get_chunked_response_reader()
+
+	mut buf := []u8{len: 1024}
+
+	for {
+		body.read(mut buf) or { break }
+	}
 }
 
-// remove_image removes the image with the given ID.
-pub fn remove_image(id string) ?bool {
-	res := request('DELETE', urllib.parse('/v1.41/images/$id')?)?
+// create_image_from_container creates a new image from a container.
+pub fn (mut d DockerConn) create_image_from_container(id string, repo string, tag string) ?Image {
+	d.send_request(Method.post, urllib.parse('/v1.41/commit?container=$id&repo=$repo&tag=$tag')?)?
+	head, body := d.read_response()?
 
-	return res.status_code == 200
+	if head.status_code != 201 {
+		data := json.decode(DockerError, body)?
+
+		return error(data.message)
+	}
+
+	data := json.decode(Image, body)?
+
+	return data
+}
+
+// remove_image removes the image with the given id.
+pub fn (mut d DockerConn) remove_image(id string) ? {
+	d.send_request(Method.delete, urllib.parse('/v1.41/images/$id')?)?
+	head, body := d.read_response()?
+
+	if head.status_code != 200 {
+		data := json.decode(DockerError, body)?
+
+		return error(data.message)
+	}
 }
diff --git a/src/docker/stream.v b/src/docker/stream.v
new file mode 100644
index 00000000..001f4b3e
--- /dev/null
+++ b/src/docker/stream.v
@@ -0,0 +1,135 @@
+module docker
+
+import io
+import util
+import encoding.binary
+import encoding.hex
+
+// ChunkedResponseReader parses an underlying HTTP chunked response, exposing
+// it as if it was a continuous stream of data.
+struct ChunkedResponseReader {
+mut:
+	reader              io.BufferedReader
+	bytes_left_in_chunk u64
+	started             bool
+}
+
+// new_chunked_response_reader creates a new ChunkedResponseReader on the heap
+// with the provided reader.
+pub fn new_chunked_response_reader(reader io.BufferedReader) &ChunkedResponseReader {
+	r := &ChunkedResponseReader{
+		reader: reader
+	}
+
+	return r
+}
+
+// read satisfies the io.Reader interface.
+pub fn (mut r ChunkedResponseReader) read(mut buf []u8) ?int {
+	if r.bytes_left_in_chunk == 0 {
+		// An io.BufferedReader always returns none if its stream has
+		// ended.
+		r.bytes_left_in_chunk = r.read_chunk_size()?
+	}
+
+	mut c := 0
+
+	// Make sure we don't read more than we can safely read. This is to avoid
+	// the underlying reader from becoming out of sync with our parsing:
+	if buf.len > r.bytes_left_in_chunk {
+		c = r.reader.read(mut buf[..r.bytes_left_in_chunk])?
+	} else {
+		c = r.reader.read(mut buf)?
+	}
+
+	r.bytes_left_in_chunk -= u64(c)
+
+	return c
+}
+
+// read_chunk_size advances the reader & reads the size of the next HTTP chunk.
+// This function should only be called if the previous chunk has been
+// completely consumed.
+fn (mut r ChunkedResponseReader) read_chunk_size() ?u64 {
+	if r.started {
+		mut buf := []u8{len: 2}
+
+		// Each chunk ends with a `\r\n` which we want to skip first
+		r.reader.read(mut buf)?
+	}
+
+	r.started = true
+
+	mut res := []u8{}
+	util.read_until_separator(mut r.reader, mut res, http_chunk_separator)?
+
+	// The length of the next chunk is provided as a hexadecimal
+	mut num_data := hex.decode(res#[..-2].bytestr())?
+
+	for num_data.len < 8 {
+		num_data.insert(0, 0)
+	}
+
+	num := binary.big_endian_u64(num_data)
+
+	// This only occurs for the very last chunk, which always reports a size of
+	// 0.
+	if num == 0 {
+		return none
+	}
+
+	return num
+}
+
+// StreamFormatReader parses an underlying stream of Docker logs, removing the
+// header bytes.
+struct StreamFormatReader {
+mut:
+	reader              ChunkedResponseReader
+	bytes_left_in_chunk u32
+}
+
+// new_stream_format_reader creates a new StreamFormatReader using the given
+// reader.
+pub fn new_stream_format_reader(reader ChunkedResponseReader) &StreamFormatReader {
+	r := &StreamFormatReader{
+		reader: reader
+	}
+
+	return r
+}
+
+// read satisfies the io.Reader interface.
+pub fn (mut r StreamFormatReader) read(mut buf []u8) ?int {
+	if r.bytes_left_in_chunk == 0 {
+		r.bytes_left_in_chunk = r.read_chunk_size()?
+	}
+
+	mut c := 0
+
+	if buf.len > r.bytes_left_in_chunk {
+		c = r.reader.read(mut buf[..r.bytes_left_in_chunk])?
+	} else {
+		c = r.reader.read(mut buf)?
+	}
+
+	r.bytes_left_in_chunk -= u32(c)
+
+	return c
+}
+
+// read_chunk_size advances the reader & reads the header bytes for the length
+// of the next chunk.
+fn (mut r StreamFormatReader) read_chunk_size() ?u32 {
+	mut buf := []u8{len: 8}
+
+	r.reader.read(mut buf)?
+
+	num := binary.big_endian_u32(buf[4..])
+
+	if num == 0 {
+		return none
+	}
+
+	return num
+}
diff --git a/src/env/README.md b/src/env/README.md
new file mode 100644
index 00000000..135e8fa9
--- /dev/null
+++ b/src/env/README.md
@@ -0,0 +1,7 @@
+This module provides a framework for parsing a configuration, defined as a
+struct, from both a TOML configuration file & environment variables. Some
+notable features are:
+
+* Overwrite values in config file using environment variables
+* Allow default values in config struct
+* Read environment variable value from file
diff --git a/src/env/env.v b/src/env/env.v
index d1459311..5ed19552 100644
--- a/src/env/env.v
+++ b/src/env/env.v
@@ -3,13 +3,19 @@ module env
 import os
 import toml
 
-// The prefix that every environment variable should have
-const prefix = 'VIETER_'
-
-// The suffix an environment variable in order for it to be loaded from a file
-// instead
-const file_suffix = '_FILE'
+const (
+	// The prefix that every environment variable should have
+	prefix      = 'VIETER_'
+	// The suffix an environment variable in order for it to be loaded from a file
+	// instead
+	file_suffix = '_FILE'
+)
 
+// get_env_var tries to read the contents of the given environment variable. It
+// looks for either `${env.prefix}${field_name.to_upper()}` or
+// `${env.prefix}${field_name.to_upper()}${env.file_suffix}`, returning the
+// contents of the file instead if the latter. If both or neither exist, the
+// function returns an error.
 fn get_env_var(field_name string) ?string {
 	env_var_name := '$env.prefix$field_name.to_upper()'
 	env_file_name := '$env.prefix$field_name.to_upper()$env.file_suffix'
diff --git a/src/main.v b/src/main.v
index 41d0d331..db6d5ef9 100644
--- a/src/main.v
+++ b/src/main.v
@@ -31,7 +31,7 @@ fn main() {
 			logs.cmd(),
 		]
 	}
-
 	app.setup()
 	app.parse(os.args)
+	return
 }
diff --git a/src/util/README.md b/src/util/README.md
new file mode 100644
index 00000000..529e412e
--- /dev/null
+++ b/src/util/README.md
@@ -0,0 +1,2 @@
+This module defines a few useful functions used throughout the codebase that
+don't specifically fit inside a module.
diff --git a/src/util/stream.v b/src/util/stream.v
new file mode 100644
index 00000000..06397aa8
--- /dev/null
+++ b/src/util/stream.v
@@ -0,0 +1,95 @@
+// Functions for interacting with `io.Reader` & `io.Writer` objects.
+module util
+
+import io
+import os
+
+// reader_to_writer tries to consume the entire reader & write it to the writer.
+pub fn reader_to_writer(mut reader io.Reader, mut writer io.Writer) ? {
+	mut buf := []u8{len: 10 * 1024}
+
+	for {
+		bytes_read := reader.read(mut buf) or { break }
+		mut bytes_written := 0
+
+		for bytes_written < bytes_read {
+			c := writer.write(buf[bytes_written..bytes_read]) or { break }
+
+			bytes_written += c
+		}
+	}
+}
+
+// reader_to_file writes the contents of a BufferedReader to a file
+pub fn reader_to_file(mut reader io.BufferedReader, length int, path string) ? {
+	mut file := os.create(path)?
+	defer {
+		file.close()
+	}
+
+	mut buf := []u8{len: reader_buf_size}
+	mut bytes_left := length
+
+	// Repeat as long as the stream still has data
+	for bytes_left > 0 {
+		// TODO check if just breaking here is safe
+		bytes_read := reader.read(mut buf) or { break }
+		bytes_left -= bytes_read
+
+		mut to_write := bytes_read
+
+		for to_write > 0 {
+			// TODO don't just loop infinitely here
+			bytes_written := file.write(buf[bytes_read - to_write..bytes_read]) or { continue }
+			// file.flush()
+
+			to_write = to_write - bytes_written
+		}
+	}
+}
+
+// match_array_in_array<T> returns how many elements of a2 overlap with a1. For
+// example, if a1 = "abcd" & a2 = "cd", the result will be 2. If the match is
+// not at the end of a1, the result is 0.
+pub fn match_array_in_array<T>(a1 []T, a2 []T) int {
+	mut i := 0
+	mut match_len := 0
+
+	for i + match_len < a1.len {
+		if a1[i + match_len] == a2[match_len] {
+			match_len += 1
+		} else {
+			i += match_len + 1
+			match_len = 0
+		}
+	}
+
+	return match_len
+}
+
+// read_until_separator consumes an io.Reader until it encounters some
+// separator array. The data read is stored inside the provided res array.
+pub fn read_until_separator(mut reader io.Reader, mut res []u8, sep []u8) ? {
+	mut buf := []u8{len: sep.len}
+
+	for {
+		c := reader.read(mut buf)?
+		res << buf[..c]
+
+		match_len := match_array_in_array(buf[..c], sep)
+
+		if match_len == sep.len {
+			break
+		}
+
+		if match_len > 0 {
+			match_left := sep.len - match_len
+			c2 := reader.read(mut buf[..match_left])?
+			res << buf[..c2]
+
+			if buf[..c2] == sep[match_len..] {
+				break
+			}
+		}
+	}
+}
diff --git a/src/util/util.v b/src/util/util.v
index 66026215..266bcb5b 100644
--- a/src/util/util.v
+++ b/src/util/util.v
@@ -1,13 +1,13 @@
 module util
 
 import os
-import io
 import crypto.md5
 import crypto.sha256
 
-const reader_buf_size = 1_000_000
-
-const prefixes = ['B', 'KB', 'MB', 'GB']
+const (
+	reader_buf_size = 1_000_000
+	prefixes        = ['B', 'KB', 'MB', 'GB']
+)
 
 // Dummy struct to work around the fact that you can only share structs, maps &
 // arrays
@@ -23,34 +23,6 @@ pub fn exit_with_message(code int, msg string) {
 	exit(code)
 }
 
-// reader_to_file writes the contents of a BufferedReader to a file
-pub fn reader_to_file(mut reader io.BufferedReader, length int, path string) ? {
-	mut file := os.create(path)?
-	defer {
-		file.close()
-	}
-
-	mut buf := []u8{len: util.reader_buf_size}
-	mut bytes_left := length
-
-	// Repeat as long as the stream still has data
-	for bytes_left > 0 {
-		// TODO check if just breaking here is safe
-		bytes_read := reader.read(mut buf) or { break }
-		bytes_left -= bytes_read
-
-		mut to_write := bytes_read
-
-		for to_write > 0 {
-			// TODO don't just loop infinitely here
-			bytes_written := file.write(buf[bytes_read - to_write..bytes_read]) or { continue }
-			// file.flush()
-
-			to_write = to_write - bytes_written
-		}
-	}
-}
-
 // hash_file returns the md5 & sha256 hash of a given file
 // TODO actually implement sha256
 pub fn hash_file(path &string) ?(string, string) {