Failed Precondition - Intent with ID not found among intents in environment

[Archetype90]

Environment details

1. Windows 10
2. Python 3.6.6
3. dialogflow version: 0.7.2`

Steps to reproduce

It is not clear to me what caused this issue to pop up. Existing code that has not changed gets this error intermittently starting about three days ago. I am roughly using the boilerplate stream audio and print the intermediate transcript code.

After I stop speaking, but before an intent is returned, I get the following exception:

2019-12-27 09:56:46 - INFO - Intermediate transcript: "turn the display".
2019-12-27 09:56:46 - ERROR - 400 Intent with id 'xxxxxxxxxxxxxxxxxxxxxxxxx' not found among intents in environment '' for agent with id 'xxxxxxxxxxxxxxxxxxxxxxxxx'.
com.google.apps.framework.request.StatusException: <eye3 title='FAILED_PRECONDITION'/> generic::FAILED_PRECONDITION: Intent with id 'xxxxxxxxxxxxxxxxxxxxxxxxx' not found among intents in environment '' for agent with id 'xxxxxxxxxxxxxxxxxxxxxxxxx'.
2019-12-27 09:56:46 - ERROR - Failed precondition - what does this mean?
====================
2019-12-27 09:56:46 - INFO - Query text:
2019-12-27 09:56:46 - INFO - Detected intent:  (confidence: 0.0)
2019-12-27 09:56:46 - INFO - Fulfillment text:

The error is very explicit, but the error is not consistent and the intents do exist in my project.

Stack trace

2019-12-27 09:52:13 - INFO - Intermediate transcript: "turn the display".
Exception in thread Thread-15:
Traceback (most recent call last):
  File "C:\Users\wbaker\AppData\Local\Programs\Python\Python36\lib\site-packages\google\api_core\grpc_helpers.py", line 79, in next
    return six.next(self._wrapped)
  File "C:\Users\wbaker\AppData\Local\Programs\Python\Python36\lib\site-packages\grpc\_channel.py", line 388, in __next__
    return self._next()
  File "C:\Users\wbaker\AppData\Local\Programs\Python\Python36\lib\site-packages\grpc\_channel.py", line 382, in _next
    raise self
grpc._channel._Rendezvous: <_Rendezvous of RPC that terminated with:
        status = StatusCode.FAILED_PRECONDITION
        details = "Intent with id 'xxxxxxxxxxxxxxxxxxxxxxxxx' not found among intents in environment '' for agent with id 'xxxxxxxxxxxxxxxxxxxxxxxxx'.
com.google.apps.framework.request.StatusException: <eye3 title='FAILED_PRECONDITION'/> generic::FAILED_PRECONDITION: Intent with id 'xxxxxxxxxxxxxxxxxxxxxxxxx' not found among intents in environment '' for agent with id 'xxxxxxxxxxxxxxxxxxxxxxxxx'."
        debug_error_string = "{"created":"@1577461933.949000000","description":"Error received from peer ipv4:172.217.14.170:443","file":"src/core/lib/surface/call.cc","file_line":1055,"grpc_message":"Intent with id 'xxxxxxxxxxxxxxxxxxxxxxxxx' not found among intents in environment '' for agent with id 'xxxxxxxxxxxxxxxxxxxxxxxxx'.\ncom.google.apps.framework.request.StatusException: <eye3 title='FAILED_PRECONDITION'/> generic::FAILED_PRECONDITION: Intent with id 'xxxxxxxxxxxxxxxxxxxxxxxxx' not found among intents in environment '' for agent with id 'xxxxxxxxxxxxxxxxxxxxxxxxx'.","grpc_status":9}"
>

I suspect there is a decent chance that I am doing something wrong here, but the fact that I had not seen this error for 1-2 months while using the same code has me concerned that this is some sort of bug.

[Syndace]

I'm experiencing the same issue. Some code that was working just a while ago now throws errors as described by @Archetype90 during intent classification. Note that I am doing text-based intent classification and not speech-based.

It might be important to add that I programmatically train the agent before trying to classify inputs.

I had a similar issue back when initially writing the code a few months ago, back then the issue was me not waiting for the training to finish. But now I am quite sure that I do in fact wait for the training to finish.

I found something weird in the web interface: When looking under "Intents", the web interface tells me "No regular intents yet. Create the first one.". This is expected, as I have removed all intents from the Agent. When clicking on "Validation" though, it shows a list of intent ids, saying "Intent 'xxx' contains no training phrases.". This makes me believe that the intent ids somehow remain in the system, even though the intents should have been completely removed.
Maybe this is related to the issues we are experiencing?

[nnegrey]

Hi folks, can you provide some more information about what you are doing?

Are you using V2 or V2Beta1?
Do you have environments configured or No?
Can you point to some code or share a bit of code to help re-produce the issue?

[Syndace]

Oh yeah, the API version would've been a good idea.

1. I'm on V2 (not beta).
2. Probably not, because I have no idea what you mean by "environment".
3. Following are a few quick code snippets I copied from my code. I don't really have the time to prepare a minimal example right now, sorry.

import dialogflow_v2

# Prepare the various clients (credentials are provided via the env var)
agents_client   = dialogflow_v2.AgentsClient()
intents_client  = dialogflow_v2.IntentsClient()
sessions_client = dialogflow_v2.SessionsClient()

# Create/configure the agent
agents_client.set_agent(dialogflow_v2.types.Agent(...))

# Create a few intents, each with multiple dialogflow_v2.types.Intent.TrainingPhrase
intent_instances = [ dialogflow_v2.types.Intent(...), ... ]

intent_batch = dialogflow_v2.types.IntentBatch(intents=intent_instances)

# Upload the intents
intents_client.batch_update_intents(..., intent_batch_inline=intent_batch).result()

# Train the agent
agents_client.train_agent(...).result()

for inputToClassify in inputsToClassify:
    # Prepare an input to classify
    text_input  = dialogflow_v2.types.TextInput(text=inputToClassify)
    query_input = dialogflow_v2.types.QueryInput(text=text_input)

    # Classify the input
    detect_intent_response = sessions_client.detect_intent(session, query_input)

When running the code like this, the call to detect_intent sometimes fails with the error message that OP posted.

[Archetype90]

1. V2 (not beta)
2. As above, I am not entirely clear what an environment is (can you link documentation?)
3. I am using fairly boilerplate code from your examples of detecting an intent with stream, with modifications around the intent detection. This can be found here: https://cloud.google.com/dialogflow/docs/detect-intent-stream

My try/except block (which generates the error in the OP) is this:

    print('=' * 20)
    try:
        for response in responses:
            self.logger.info('Intermediate transcript: "{}".'.format(
                response.recognition_result.transcript))
    except google_exceptions.FailedPrecondition as ex:
        self.logger.error(ex)
        self.logger.error("Failed precondition - what does this mean?")

[nnegrey]

On the environment front, I'll also assume no, I just asked cus I saw it pop up in your stack trace.

But it is a beta feature right now: https://cloud.google.com/dialogflow/docs/agents-versions

I haven't had as much time today, but I'll try to take a look more tomorrow.

[nnegrey]

Other questions for @Syndace, are you uploading new intents every time then trying to detect them?

[nnegrey]

@Archetype90, on our samples repo, there is a PR for a microphone streaming sample, is yours similar to this?
GoogleCloudPlatform/python-docs-samples#2566

[Syndace]

Yes, my flow is:

• create/set up the agent
• upload intents with example utterances
• train the agent
• classify some samples
• remove all intents

All of these steps happen right after the previous one is finished, one whole iteration of that process only takes a few seconds to a minute (depending on the number of training samples and the number of samples to classify).

[nnegrey]

What do you pass as the IntentID when using batchUpdate to create the intents?

[nnegrey]

Also @Syndace, since you seem to start from scratch every time, have you considered just using ImportAgentRequest?

[Syndace]

I don't think batch_update_intents takes any intent ids, does it?

ImportAgentRequest is not applicable, since the intents and utterances usually differ between runs, but thanks for the heads up.