custom document processing pipe without touching norch?

[Espen Klem]

I was thinking that it could be good to choose different custom document processors when adding documents. This could be done by (almost) not touching norch code.

Could be done by adding the document processors as dependencies in package.json and then calling them from command line:

Here a twitter document processor is inserted before it's added to the index.

curl -X POST -p twitterPipeline -d @justTen.str http://localhost:3030/add

[Fergus McDowall]

It should be possible to use unix pipes:

cat data.json | twitterPipeline | curl -H "Content-Type: application/json" -X POST -d @- http://localhost:3030/add

[Espen Klem]

Ah, nice!
A little unclear issue. I was also thinking of when the post request comes from i.e. Zapier and I can only rely on the regular /add functionality. A way to get Norch to choose one of a set of custom pipelines is what I'm looking for.

[Fergus McDowall]

Yup- I see what you mean- what should the API look like?

[Espen Klem]

Hmm, maybe something could be sent as header key/value pair:
customPipeline | [source]

[Espen Klem]

I need to fix a bug in a demo, but I can see what this looks like on the norch server side and get back to you.

[Espen Klem]

Could be done through the JSON, but that's a bit ugly.