step 7 of tutorial 1

[vinith-reddy25]

By running

"$ genie server --nlu-model file://OUTPUTDIR --thingpedia thingpedia.tt -l en-US"

output is coming as:

"
I 220118 16:31:25 transformer_lstm:79] Initializing encoder and decoder embeddings
[I 220118 16:31:26 transformer_lstm:108] Vocabulary has 28997 tokens
[I 220118 16:31:27 util:555] TransformerLSTM has 125,656,333 parameters

"
After that when "localhost:8400" is tried to open in browser then "{"error":"Invalid endpoint"}" is coming. Can someone tell what to do??

Will it take too much time to deploy?? or some issue is there in my side??

[sileix]

It won't take long to deploy.
The only endpoint opened for the nlu server is /en/query. For example, if you want to get the result for utterance "get restaurants", you would go to localhost:8400/en/query?q=get restaurants

[vinith-reddy25]

/en/query endpoint not working. Can you send any link where there is a discussion about api call

[sileix]

Here is some documentation about the API provided in our production server: https://wiki.genie.stanford.edu/en/api-references/nlp - it is the same as the self-served one here. And the endpoint is implemented here:

genie-toolkit/tool/server.ts

Line 222 in 235171b

app.post('/:locale/query', qv.validatePOST(QUERY_PARAMS, { accept: 'application/json' }), (req, res, next) => {

.

Both en and en-US should work as the locale. An example on our production server: https://nlp.genie.stanford.edu/en/query?q=get%20restaurants

[vinith-reddy25]

I was asking for http://localhost:8400 as it is giving error and not using https://nlp.genie.stanford.edu/

[sileix]

Both localhost and our production server share the same implementation of the API call. We don't have separate documentation for API for genie-server at the moment, but simply replace the URL from nlp.genie.stanford.edu to localhost:8400, everything else should be the same.