DocumentationAPI ReferenceTutorialsGitHub Code ExamplesDiscord Community

Chatbot Integration

It is easy and simple to imbue your chatbot with the NLP power that Haystack offers. If your chatbot is handling a question that is in the log tail of undefined intents, why not handle it using Haystack's question answering capabilities? Have a look at this guide to learn more.

Thanks to Haystack's REST API, chatbot frameworks can seamlessly communicate with your custom Haystack service. Below, we will trace through what it takes to set this up using the popular Rasa framework.


When a user speaks to a conversational AI program, each user message is classified and labelled with an intent. One approach to integrating Haystack would be to introduce a new intent, such as knowledge question, that triggers a call to a running Haystack API.

Alternatively, in cases where they are uncertain what intent fits, chatbots can route requests to a fallback action. Rasa uses the FallbackClassifier to make this happen.

In both cases, Rasa uses an action server to send a REST API request to Haystack. When it receives the result, it will base its next message to the user on the contents of this response.


How Rasa and Haystack interact.

Setting up Haystack with REST API

To set up a Haystack instance with REST API, have a look at REST API. By default, the API server runs on [<>](`).


Quick-start demo

Bundled with the Github repository is an example Haystack service with REST API that queries the Game of Thrones Wiki. Simply clone it and call the following in the root directory:

    cd haystack
    docker-compose pull
    docker-compose up

    # Or on a GPU machine: docker-compose -f docker-compose-gpu.yml up

Setting Up a Rasa Chatbot



For a bare bones example of a Rasa chatbot communicating with Haystack, have a look at this repo.

After installing Rasa and initializing a new project, there are a few steps that a developer needs to take in order to get Rasa to communicate with Haystack.

In data/nlu.yaml, you will want to define a new intent by providing example utterances of that intent:

- intent: knowledge_question
  examples: |
    - Can you look this up: what is ?
    - Look this up: what is?
    - Can you check: what is?
    - please check: What is ?
    - Can you look this up: where is ?
    - Look this up: where is?
    - Can you check: where is?
    - please check: where is ?

You will also want to define what action is taken when that intent is identified in the data/rules.yml file:

- rule: Query Haystack anytime the user has a direct question for your document base
  - intent: knowledge_question
  - action: call_haystack

If you want the Haystack API call to be triggered by a fallback instead of an intent, you will need to add the FallbackClassifier to the pipeline in config.yml:

   - name: FallbackClassifier
     threshold: 0.8
     ambiguity_threshold: 0.1

You will also need to add the following to data/rules.yml:

- rule: Query Haystack whenever they send a message with low NLU confidence
    - intent: nlu_fallback
    - action: call_haystack

By default the Rasa custom action server API is blocked so you will need to uncomment this line from endpoints.yml:

 url: "http://localhost:5055/webhook"

Finally, we can define an action function that calls the Haystack API and handles the response in actions/ Here is an example of what that might look like:

class ActionHaystack(Action):

    def name(self) -&gt; Text:
        return "call_haystack"

    def run(self, dispatcher: CollectingDispatcher,
            tracker: Tracker,
            domain: Dict[Text, Any]) -&gt; List[Dict[Text, Any]]:

        url = "http://localhost:8000/query"
        payload = {"query": str(tracker.latest_message["text"])}
        headers = {
            'Content-Type': 'application/json'
        response = requests.request("POST", url, headers=headers, json=payload).json()

        if response["answers"]:
            answer = response["answers"][0]["answer"]
            answer = "No Answer Found!"


        return []

You will also want to declare this new action in the domain.yml file:

- call_haystack

Running the Chatbot

Whenever changes are made to your Rasa project, you will want to retrain your chatbot via:

rasa train

You will need to start the Rasa action server using:

rasa run actions

To start interacting with the bot you have created, you can call

rasa shell

Now you can start talking to the Haystack enabled chatbot!


Example conversation with Haystack enabled chatbot.

Alternatively, you can use Rasa X which allows the chatbot to be run in a GUI. While interacting with you chatbot in this interface, you will also see the intent being assigned to each message as well as the action being taken.


Example conversation in Rasa X.