What's actual rate limits of filters?


I’m trying to figure out what’s limits of Streaming API’s filter method
Is there any maximum simultaneous connection limit?

Example Scenario:
I have 1000 user and want to watch 5 different keywords per user. So i will open 1000*5 connection to monitor tweets.

Is there a chance that any of that connection interrupts due to maximum tweet limit per search? If so, can i re-open connection?
Is there any maximum connection limit per user/per application? (e.g. maximum 3 keywords (connections) per user / maximum 3 keywords per API with 15 minute windows)
Any other limits could happen? and how can i avoid them?
Lastly, can i use Search API’s query operators? (https://dev.twitter.com/docs/using-search)



You need to rethink your approach. You should have a single open connection to stream.twitter.com representing your entire application – not one connection per user or even multiple connections for each user.

For all the terms your users want to track, you track those in a single stream and then divvy out the results with your own logic.

Connections should be long-lived and only reconnected when criteria change.

Check out [node:14935] for disucssion on this.


But what if user wants to filter results per keyword unlike one string, isn’t its impossible with one connection? And opening one connection for all users requires filtering returned tweets to determine which user to be warned, is there any easier way or should i stick with standard Search API (i will check keywords with 1-minute intervals)


You would have to do additional filtering application-side to accomplish those goals with the Streaming API (or reconnect each time a query was changing).

The Search API might be better given your use case – especially since you will have the user’s access token and be able to utilize it when making requests on their behalf.