I was wondering whether there was any consideration given to enabling the whole build/deploy process outside the planner service. It would be nice if i could write and test my code in my own environment before uploading and distributing it to the imp.
Additionally, are you looking at publishing the server API?
I did this myself by building squirrel command line (myself) and write my own imp/server objects to simulate those calls. I can run my code 100% locally before putting it on the imp. It isn’t perfect, since it is hard to do the timing, etc, exactly right, but for the most part, I can simulate most of the environment.
Right now, the only way to run squirrel outside the website is to download squirrel and have some hardware stubs to run against (sounds like cnr1089 has done this).
@cnr1089 - maybe you could share your stubs?
As for the server API, if you mean the API the imps talk to the server with then no - this channel is only open to imps. If you have other devices that want to talk to your imps, you can implement any API you fancy in the companion agent and offer it externally.
Thanks for the response.
@cnr1089 Do you still have to copy the code into the planner for deploying to the imp?
@hugo I’m referring to the former. For instance is there a way i can upload my code to the imp via an API call? Can i access the data written to permanent storage via an API call?
There’s no current API for uploading code direct, no, though we have better integration with things like github on the todo list.
For external access to imp-accessible storage, you will be able to do this by implementing your own API with an agent. The agent storage isn’t seamlessly shared with the imp for many reasons, but you should be able to achieve what you need to with an agent.
Is there specific documentation on ‘agents’ I have seen it mentioned on the list but can’t seem to locate any documentation?
So, just to clarify, the only way to deploy code to an imp is via the planner.
What’s the priority on github intergration?
P.S. Have a look at the process Stoic uses to garner community support for features (http://stoic.com/wishlist.html). It provides great visibility to the community,.
No, there is no agent documentation available until agents go on general release. Correct on the deployment right now. Github is “later”; it’s something we want to do but it’s behind features requested by commercial customers which we need to address first.
A user-driven wishlist is nice, but a significant number of requests we deal with are related to commercial users who don’t necessarily want their specific use cases broadcast in advance so it’s a bit harder to for us to “develop in public” in that way.