Mozci looks for a configuration file in your user config dir (e.g ~/.config/mozci/config.toml):

The config is a TOML file, which looks something like:

verbose = 1

List of Options

The following keys are valid config options.


This value allows you to set up a cache to store the results for future use. This avoids the penalty of hitting expensive data sources.

The mozci module uses cachy to handle caching. Therefore the following stores are supported:

  • database
  • file system
  • memcached
  • redis

To enable caching, you’ll need to configure at least one store using the cache.stores key. Follow cachy’s configuration format identically. In addition to the options cachy supports, you can set the mozci.cache.retention key to the time in minutes before stored queries are invalidated.

For example:

retention = 10080  # minutes

file = { driver = "file", path = "/path/to/dir/to/keep/cache" }

In addition, mozci defines several custom cache stores:

  • a seeded-file store. This is the same as the “file system” store, except you can specify a URL to initially seed your cache on creation:
file = {
    driver = "seeded-file",
    path = "/path/to/dir/to/keep/cache",
    url = ""

Supported archive formats include .zip, .tar, .tar.gz, .tar.bz2 and .tar.zst.

The config also accepts a reseed_interval (in minutes) which will re-seed the cache after the interval expires. This assumes the URL is automatically updated by some other process.

As well as an archive_relpath config, which specifies the path to the cache data “within” the archive. Otherwise the cache data is assumed to be right at the root of the archive.

  • a renewing-file store. This is the same as the “file system” store, except it renews items in the cache when they are retrieved.
  • an s3 store, which allows caching items in a S3 bucket. With this store, items are renewed on access like renewing-file. It’s suggested to use a S3 Object Expiration policy to clean up items which are not accessed for a long time. Example configuration:
s3 = {
    driver = "s3",
    bucket = "myBucket",
    prefix = "data/mozci_cache/"


Mozci can retrieve data from many different sources, e.g treeherder, taskcluster,, etc. Often these sources can provide the same data, but may have different runtime characteristics. For example, some may not have realtime data, might require authentication or might take a really long time.

You can choose which sources you want to use with this key. For example:

data_sources = ["treeherder_client", "taskcluster"]

The above will first try to fulfill any data requirements using the treeherder_client source. But if that source is unable to fulfill the contract, the taskcluster source will be used as a backup.

Available sources are defined in the DataHandler class.


Enable verbose logging (default: 0). Setting this to 1 enables debug logging, while setting it to 2 enables trace logging.


Mozci controls which push classification results can be automatically processed by third-party tools (like Treeherder), using a feature flag and a set of filters for test names.

The feature can be fully disabled by setting enabled to false.

enabled = true
test-suite-names = [

Each value in the list of test-suite-names support [fnmatch]( syntax to allow glob-like syntax (using * for wildcard and ? for single characters).

The configuration above will enable autoclassification for tests matching test-linux64-*/opt-mochitest-* or *wpt*

Finally, the JSON classification output is extended to have an autoclassify boolean flag on each failure details payload, to check if this specific result should be processed.