Template of a custom feed generator service for the Bluesky network in Ruby
1require 'json' 2require 'sinatra/activerecord' 3require 'skyfall' 4 5require_relative 'config' 6require_relative 'models/feed_post' 7require_relative 'models/post' 8require_relative 'models/subscription' 9 10class FirehoseStream 11 attr_accessor :show_progress, :log_status, :log_posts, :save_posts, :replay_events 12 13 DEFAULT_RELAY = 'bsky.network' 14 15 def initialize(service = nil) 16 @env = (ENV['APP_ENV'] || ENV['RACK_ENV'] || :development).to_sym 17 @service = service || DEFAULT_RELAY 18 19 @show_progress = (@env == :development) ? true : false 20 @log_status = true 21 @log_posts = (@env == :development) ? :matching : false 22 @save_posts = (@env == :development) ? :all : :matching 23 @replay_events = (@env == :development) ? false : true 24 25 @feeds = BlueFactory.all_feeds.select(&:is_updating?) 26 end 27 28 def start 29 return if @sky 30 31 last_cursor = load_or_init_cursor 32 cursor = @replay_events ? last_cursor : nil 33 34 @sky = sky = Skyfall::Firehose.new(@service, :subscribe_repos, cursor) 35 36 @sky.on_message do |m| 37 process_message(m) 38 end 39 40 if @log_status 41 @sky.on_connecting { |u| log "Connecting to #{u}..." } 42 @sky.on_connect { 43 @replaying = !!(cursor) 44 log "Connected ✓" 45 } 46 @sky.on_disconnect { 47 log "Disconnected." 48 save_cursor(sky.cursor) 49 } 50 @sky.on_reconnect { log "Connection lost, reconnecting..." } 51 @sky.on_error { |e| log "ERROR: #{e.class} #{e.message}" } 52 end 53 54 @sky.connect 55 end 56 57 def stop 58 @sky&.disconnect 59 @sky = nil 60 end 61 62 def load_or_init_cursor 63 if sub = Subscription.find_by(service: @service) 64 sub.cursor 65 else 66 Subscription.create!(service: @service, cursor: 0) 67 nil 68 end 69 end 70 71 def save_cursor(cursor) 72 Subscription.where(service: @service).update_all(cursor: cursor) 73 end 74 75 def process_message(msg) 76 if msg.type == :info 77 # AtProto error, the only one right now is "OutdatedCursor" 78 log "InfoMessage: #{msg}" 79 elsif msg.type == :handle 80 # use these events if you want to track handle changes: 81 # log "Handle change: #{msg.repo} => #{msg.handle}" 82 elsif msg.is_a?(Skyfall::Firehose::UnknownMessage) 83 log "Unknown message type: #{msg.type} (#{msg.seq})" 84 end 85 86 return unless msg.type == :commit 87 88 if @replaying 89 log "Replaying events since #{msg.time.getlocal} -->" 90 @replaying = false 91 end 92 93 if msg.seq % 10 == 0 94 save_cursor(msg.seq) 95 end 96 97 msg.operations.each do |op| 98 case op.type 99 when :bsky_post 100 process_post(msg, op) 101 102 when :bsky_like, :bsky_repost 103 # if you want to use the number of likes and/or reposts for filtering or sorting: 104 # add a likes/reposts column to feeds, then do +1 / -1 here depending on op.action 105 106 when :bsky_follow 107 # if you want to make a personalized feed that needs info about given user's follows/followers: 108 # add a followers table, then add/remove records here depending on op.action 109 110 else 111 # other types like :bsky_block, :bsky_profile (includes profile edits) 112 end 113 end 114 end 115 116 def process_post(msg, op) 117 if op.action == :delete 118 if post = Post.find_by(repo: op.repo, rkey: op.rkey) 119 post.destroy 120 end 121 end 122 123 return unless op.action == :create 124 125 begin 126 if op.raw_record.nil? 127 log "Error: missing expected record data in operation: #{op.uri} (#{msg.seq})" 128 return 129 end 130 rescue CBOR::UnpackError => e 131 log "Error: couldn't decode record data for #{op.uri} (#{msg.seq}): #{e}" 132 return 133 end 134 135 # ignore posts with past date from Twitter etc. imported using some kind of tool 136 begin 137 post_time = Time.parse(op.raw_record['createdAt']) 138 return if post_time < msg.time - 86400 139 rescue StandardError => e 140 log "Skipping post with invalid timestamp: #{op.raw_record['createdAt'].inspect} (#{op.repo}, #{msg.seq})" 141 return 142 end 143 144 text = op.raw_record['text'] 145 146 # to save space, delete redundant post text and type from the saved data JSON 147 trimmed_record = op.raw_record.dup 148 trimmed_record.delete('$type') 149 trimmed_record.delete('text') 150 trimmed_json = JSON.generate(trimmed_record) 151 152 # tip: if you don't need full record data for debugging, delete the data column in posts 153 post = Post.new( 154 repo: op.repo, 155 time: msg.time, 156 text: text, 157 rkey: op.rkey, 158 data: trimmed_json, 159 record: op.raw_record 160 ) 161 162 matched = false 163 164 @feeds.each do |feed| 165 if feed.post_matches?(post) 166 FeedPost.create!(feed_id: feed.feed_id, post: post, time: msg.time) unless !@save_posts 167 matched = true 168 end 169 end 170 171 if @log_posts == :all || @log_posts && matched 172 puts 173 puts text 174 end 175 176 post.save! if @save_posts == :all 177 178 print '.' if @show_progress && @log_posts != :all 179 rescue StandardError => e 180 log "Error in #process_post: #{e}" 181 182 unless e.message == "nesting of 100 is too deep" 183 log msg.inspect 184 log e.backtrace.reject { |x| x.include?('/ruby/') } 185 end 186 end 187 188 def log(text) 189 puts if @show_progress 190 puts "[#{Time.now}] #{text}" 191 end 192 193 def inspect 194 vars = instance_variables - [:@feeds, :@timer] 195 values = vars.map { |v| "#{v}=#{instance_variable_get(v).inspect}" }.join(", ") 196 "#<#{self.class}:0x#{object_id} #{values}>" 197 end 198end