Example

A demonstration server that shows how to run Tango end-to-end. It boots a YARPC/gRPC server on 127.0.0.1:8081, wiring together config parsing, storage, the repo manager, the orchestrator, and the controller. A companion CLI client calls the server's streaming RPCs, and a query-bench tool exercises the underlying Bazel query and target-hashing path without bringing up the server.

Configuration

The server reads tango-config.yaml. Top-level sections:

• storage — type: memory (default) or type: disk with a root_path.
• repository — remotes Tango is allowed to operate on, with default branch, excluded files, external-target handling, bzlmod, and per-query timeout.
• service — worker pool size, origin clone path, and per-worker checkout path. Both directories are created on start and removed on clean shutdown.

Running

Start the server:

make run-server

Query the target graph at HEAD:

make run-client-get-graph \
  REMOTE=https://github.com/uber/tango.git \
  BASE_SHA=HEAD

Diff targets between two revisions:

make run-client-changed-targets \
  REMOTE=https://github.com/uber/tango.git \
  BASE_SHA=872881fd~1 \
  NEW_BASE_SHA=872881fd

The client supports two methods (get-target-graph, get-changed-targets) and flags for limiting changed-target distance (-max-distance), cache bypass (-bypass-cache), output detail (-include-hashes, -include-tags, -include-attributes), and request URLs (-request-urls, -new-request-urls). Run with -h for the full list.

Benchmarking

The query-bench tool times the standard Tango query against a real Bazel workspace and reports per-stage timings (query, hashing, response conversion):

bazel run //example/cmd/query-bench -- --workspace /path/to/repo --runs 3