livekit

mirror of https://github.com/hanzoai/livekit.git synced 2026-10-30 18:29:34 +00:00

Files

T

History

zeekay e9292da479 refactor: own the module path, and move the KV client to hanzokv/go

This fork lived at github.com/hanzoai/livekit but still declared
`module github.com/livekit/livekit-server`, so all 505 self-references pointed
at upstream. A fork that does not own its module path is the exact thing the
owned-module-paths rule exists to prevent: every import line in every consumer
carries someone else's brand, and `go get` resolves to their code, not ours.

  module + 505 self-imports  github.com/livekit/livekit-server -> github.com/hanzoai/livekit
  KV client                  github.com/redis/go-redis/v9      -> github.com/hanzokv/go/v9
  message bus                github.com/livekit/psrpc          -> github.com/hanzoai/psrpc

psrpc had to be forked (hanzoai/psrpc, Apache-2.0) because its bus is typed
against the KV client — the same shape redsync had for hanzoai/git.

DELIBERATELY NOT FORKED: github.com/livekit/{protocol,mediatransportutil,
mageutil} keep their addresses. Only `protocol` still reaches go-redis, through
a single file (redis/redis.go) out of 305. Forking an 11.5MB core wire-protocol
library that this server must track in lockstep — and that upstream pushed
yesterday — to change one file trades a tidy dependency graph for permanent
divergence on the fastest-moving dependency here. So one indirect
github.com/redis/go-redis entry remains, via livekit/protocol/redis, and that
is the correct place for it to live.

2026-07-26 08:14:23 -07:00

buffer_base.go

refactor: own the module path, and move the KV client to hanzokv/go

2026-07-26 08:14:23 -07:00

buffer_test.go

Do not restart receiver on codec change mid-session. (#4225 )

2026-01-08 02:35:32 +05:30

buffer.go

refactor: own the module path, and move the KV client to hanzokv/go

2026-07-26 08:14:23 -07:00

datastats_test.go

Adding license notices (#1913 )