Chorus is a nostr relay.
It is pretty fast: docs/PERFORMANCE.md.
It can work as your personal relay (or as an open relay): docs/PERSONAL_RELAY.md
To deploy chorus, read docs/DEPLOYING.md
To configure chorus read docs/CONFIG.md
To undertand the relay's behavior, read docs/BEHAVIOR.md
To understand command line tools, read docs/TOOLS.md
To understand the management API, read docs/MANAGEMENT.md
Chorus does not have any provisions for charging users.
Chorus does not have any provisions for synchronizing events with other relays outside of the nostr protocol.
strfry is a more mature relay that offers additional features including:
- Synchronizing events with other relays efficiently (negentropy)
- Zero-downtime restarts
- Websocket permessage-deflate
- Plugins for event sifting
However:
- Chorus is probably faster (more efficient)
- Chorus has personal relay rules by default
- Chorus has extensive IP banning and rate limiting to protect your relay from abuse
- Chorus supports NIP-42 (AUTH), NIP-59 (GiftWrap), NIP-65 (Relay Lists) and PR 1030 and PR 1325.
- Chorus has a moderation cmd line tool and a moderation API (PR 1325)
Use the branch latest.
Do not run off of the master branch. Master is updated with breaking changes that are
not only unstable, but which may require you to update your configuration. I will not
announce upgrade instructions until release.
- NEW CONFIG:
throttling_bytes_per_secondhow many bytes are allowed per second - NEW CONFIG:
throttling_bursthow many bytes can be accepted per connection in a burst
- FIX: large non-utf8 messages were attempted to be logged causing a panic
- FIX: parse errors were not being punished so a nasty client could do a DoS of chorus with simple junk.
- BREAKING: If you run chorus behind a proxy like nginx, you MUST set the new
chorus_is_behind_a_proxyconfig variable to true, and your proxy MUST set theX-Real-IPheader. If the header is missing, connections will not be served. If you fail to set thechorus_is_behind_a_proxysetting, the proxy IP address will be used directly, generally causing all connections to quickly become banned due to the bad behavior of just one client, or due to too many connections from a single IP. - NEW CONFIG:
chorus_is_behind_a_proxy(please set to true or false) - NEW CONFIG:
max_connections_per_ip(defaults to 5) - NEW CONFIG:
moderator_hex_keys(see next bullet point) - A rudimentary Management API is now available using nostr-protocol/nips#1325
To use management front ends against chorus, you must add hex pubkeys to
moderator_hex_keys. - Errors about DM kinds are much less common now, as we don't explicitly error unless they specify some set of kinds (we implicitly filter out the DMs still)
- Accurate count of bytes sent/received (SSL header data is now counted)
- Kind 10050 dm relay list events are now treated the same way as kind 10002 relay list events.
- Error message detail (e.g. source code line numbers) is now no longer sent to clients.
- Some mild errors are now swallowed.
- Updates of many dependencies, some updates were very large jumps and may change network/http behavior.
- Receipt of a deleted EVENT now returns OK false (was OK true)
- Origin header logged
- CLOSED: auth-required sent if DMs requested and not authenticated yet
- config: minimum_ban_seconds, timeout_seconds, enable_ip_blocking
- default for allow_scrape_if_max_seconds raised from 3600 to 7200
- default for max_subscriptions raised from 32 to 128
- timeouts no longer affect ban seconds
- Internal: switched to pocket for the backend storage
- creates lmdb subdirectory if missing
- several bugfixes: filter parsing, empty tags, event ordering, chorus_compress was fixed
chorus_moderatetool to approve events or pubkeyschorus_dump_approvalstool to view prior approvalschorus_compressto backup and rebuild your data, compressing as it goes.- NIP-40 event expiration support
- extended NIP-11 document, with updates to supported NIPs
- Traffic bytes counted and logged on exit (or HUP)
- Logging improvements
- Data migrations. Less space consumed.
- performance improved via 8-byte alignment of events
- Config option to run as an open relay
- Config variable for scraper behavior
- More efficient scraping
- Reload config on HUP without dropping connections
- IP addresses no longer logged. They are hashed and tracked by their hash.
- FIX some LIMIT related bugs
- FIX some indexing bugs that cuased some events that should have been returned to not be.
- Allow scrape up to limit=100, or time range of 1 hour
- Update docs
- BREAKING: Config changed to TOML format
- Logging improvements
- Error handling improvements
- Ban time improvements
- Documentation updates
- Restructured into a
binand alib - Tool to dump all events as JSON to stdout
- FIX: REQ errors now properly CLOSED
- Logging improvements
- Persistent IP reputation data for adjusting ban times
- Config settings for
serve_ephmeralandserve_relay_lists - Uses 'X-Real-Ip' header if behind a proxy
- FIX: significant performance problem addressed
- Logging improvements
- EVENTs submitted that are known to be deleted are now responded to with 'blocked:'
- Longer filters now accepted
- FIX: after EOSE, outbound events were not being screened properly
- FIX: shutdown issue
- NIP-11 updated
- Idle connections with no REQs now timed out at 5 seconds
- FIX: limits were not working properly
- Update IP banning logic and timing
- Improve logging
- Allow more REQ filters (with low limits)
- Better performance when filters are more open
- FIX: deadlock (nested transaction issue)
- Logging changes
- Documentation updated
- FIX: AUTH was failing due to a bad time difference comparison
- AUTH failures now include detail as to why
- documentation updated
- Initial release.
- Works as a personal relay
- Supported NIPs: 1, 4, 9, 11, 42, 59, 65
- Bans IPs to prevent abuse
- Limits to 32 subscriptions by default, configurable