Observability AI Assistant

edit

Observability AI Assistant

edit

The AI Assistant uses generative AI to provide:

  • Contextual insights — open prompts throughout Observability that explain errors and messages and suggest remediation.
  • Chat — have conversations with the AI Assistant. Chat uses function calling to request, analyze, and visualize your data.
Observability AI assistant preview

The AI Assistant integrates with your large language model (LLM) provider through our supported Elastic connectors:

The AI Assistant is powered by an integration with your large language model (LLM) provider. LLMs are known to sometimes present incorrect information as if it’s correct. Elastic supports configuration and connection to the LLM provider and your knowledge base, but is not responsible for the LLM’s responses.

Also, the data you provide to the Observability AI assistant is not anonymized, and is stored and processed by the third-party AI provider. This includes any data used in conversations for analysis or context, such as alert or event data, detection rule configurations, and queries. Therefore, be careful about sharing any confidential or sensitive details while using this feature.

Requirements

edit

The AI assistant requires the following:

  • Elastic Stack version 8.9 and later.
  • An Enterprise subscription.
  • An account with a third-party generative AI provider that supports function calling. The Observability AI Assistant supports the following providers:

    • OpenAI gpt-4+.
    • Azure OpenAI Service gpt-4(0613) or gpt-4-32k(0613) with API version 2023-07-01-preview or more recent.
    • AWS Bedrock, specifically the Anthropic Claude models.
  • The knowledge base requires a 4 GB machine learning node.

The free tier offered by third-party generative AI provider may not be sufficient for the proper functioning of the AI assistant. In most cases, a paid subscription to one of the supported providers is required.

The Observability AI assistant doesn’t support connecting to a private LLM. Elastic doesn’t recommend using private LLMs with the Observability AI assistant.

In Elastic Cloud or Elastic Cloud Enterprise, if you have Machine Learning autoscaling enabled, Machine Learning nodes will be started when using the knowledge base and AI Assistant. Therefore using these features will incur additional costs.

Your data and the AI Assistant

edit

Elastic does not use customer data for model training. This includes anything you send the model, such as alert or event data, detection rule configurations, queries, and prompts. However, any data you provide to the AI Assistant will be processed by the third-party provider you chose when setting up the OpenAI connector as part of the assistant setup.

Elastic does not control third-party tools, and assumes no responsibility or liability for their content, operation, or use, nor for any loss or damage that may arise from your using such tools. Please exercise caution when using AI tools with personal, sensitive, or confidential information. Any data you submit may be used by the provider for AI training or other purposes. There is no guarantee that the provider will keep any information you provide secure or confidential. You should familiarize yourself with the privacy practices and terms of use of any generative AI tools prior to use.

Set up the AI Assistant

edit

To set up the AI Assistant:

  1. Create an authentication key with your AI provider to authenticate requests from the AI Assistant. You’ll use this in the next step. Refer to your provider’s documentation for information about creating authentication keys:

  2. From Stack ManagementConnectors in Kibana, create an OpenAI or Amazon Bedrock connector.
  3. Authenticate communication between Observability and the AI provider by providing the following information:

    1. In the URL field, enter the AI provider’s API endpoint URL.
    2. Under Authentication, enter the API key or access key/secret you created in the previous step.

Add data to the AI Assistant knowledge base

edit

If you started using the AI Assistant in technical preview, any knowledge base articles you created before 8.12 will have to be reindexed or upgraded before they can be used. Knowledge base articles created before 8.12 use ELSER v1. In 8.12, knowledge base articles must use ELSER v2. You can either:

  • Clear all old knowledge base articles manually and reindex them.
  • Upgrade all knowledge base articles indexed with ELSER v1 to ELSER v2 using a Python script.

The AI Assistant uses ELSER, Elastic’s semantic search engine, to recall data from its internal knowledge base index to create retrieval augmented generation (RAG) responses. Adding data such as Runbooks, GitHub issues, internal documentation, and Slack messages to the knowledge base gives the AI Assistant context to provide more specific assistance.

Your AI provider may collect telemetry when using the AI Assistant. Contact your AI provider for information on how data is collected.

You can add information to the knowledge base by asking the AI Assistant to remember something while chatting (for example, "remember this for next time"). The assistant will create a summary of the information and add it to the knowledge base.

You can also add external data to the knowledge base either in Kibana using the Stack Management UI or using the Elasticsearch Index API.

Use the UI

edit

To add external data to the knowledge base in Kibana:

  1. Go to Stack Management.
  2. In the Kibana section, click AI Assistants.
  3. Then select the Elastic AI Assistant for Observability.
  4. Switch to the Knowledge base tab.
  5. Click the New entry button, and choose either:

    • Single entry: Write content for a single entry in the UI.
    • Bulk import: Upload a newline delimited JSON (ndjson) file containing a list of entries to add to the knowledge base. Each object should conform to the following format:

      {
        "id": "a_unique_human_readable_id",
        "text": "Contents of item",
      }

Use the Elasticsearch Index API

edit
  1. Ingest external data (GitHub issues, Markdown files, Jira tickets, text files, etc.) into Elasticsearch using the Elasticsearch Index API.
  2. Reindex your data into the AI Assistant’s knowledge base index by completing the following query in ManagementDev Tools in Kibana. Update the following fields before reindexing:

    • InternalDocsIndex — name of the index where your internal documents are stored.
    • text_field — name of the field containing your internal documents' text.
    • timestamp — name of the timestamp field in your internal documents.
    • public — (true or false) if true, the document is available to users in the space defined in the following space field or in all spaces if no space is defined. If false, the document is restricted to the user indicated in the following user.name field.
    • space — (can be null) if defined, restricts the internal document’s availability to a specific Kibana space.
    • user.name — (can be null) if defined, restricts the internal document’s availability to a specific user.
    • You can add a query filter to index specific documents.
POST _reindex
{
    "source": {
        "index": "<InternalDocsIndex>",
        "_source": [
            "<text_field>",
            "<timestamp>",
            "namespace",
            "is_correction",
            "public",
            "confidence"
        ]
    },
    "dest": {
        "index": ".kibana-observability-ai-assistant-kb-000001",
        "pipeline": ".kibana-observability-ai-assistant-kb-ingest-pipeline"
    },
    "script": {
        "inline": "ctx._source.text = ctx._source.remove(\"<text_field>\");ctx._source.namespace=\"<space>\";ctx._source.is_correction=false;ctx._source.public=<public>;ctx._source.confidence=\"high\";ctx._source['@timestamp'] = ctx._source.remove(\"<timestamp>\");ctx._source['user.name'] = \"<user.name>\""
    }
}

Interact with the AI Assistant

edit

You can chat with the AI Assistant or interact with contextual insights located throughout Observability. See the following sections for more on interacting with the AI Assistant.

After every answer the LLM provides, let us know if the answer was helpful. Your feedback helps us improve the AI Assistant!

Chat with the assistant

edit

Click AI Assistant in the upper-right corner of any Observability application to start the chat:

Observability AI assistant preview

This opens the AI Assistant flyout, where you can ask the assistant questions about your instance:

Observability AI assistant chat

Asking questions about your data requires function calling, which enables LLMs to reliably interact with third-party generative AI providers to perform searches or run advanced functions using customer data.

When the Observability AI Assistant performs searches in the cluster, the queries are run with the same level of permissions as the user.

Suggest functions

edit

This functionality is in beta and is subject to change. The design and code is less mature than official GA features and is being provided as-is with no warranties. Beta features are not subject to the support SLA of official GA features.

The AI Assistant uses functions to include relevant context in the chat conversation through text, data, and visual components. Both you and the AI Assistant can suggest functions. You can also edit the AI Assistant’s function suggestions and inspect function responses.

You can suggest the following functions:

alerts

Get alerts for Observability.

elasticsearch

Call Elasticsearch APIs on your behalf.

kibana

Call Kibana APIs on your behalf.

summarize

Summarize parts of the conversation.

visualize_query

Visualize charts for ES|QL queries.

Additional functions are available when your cluster has APM data:

get_apm_correlations

Get field values that are more prominent in the foreground set than the background set. This can be useful in determining which attributes (such as error.message, service.node.name, or transaction.name) are contributing to, for instance, a higher latency. Another option is a time-based comparison, where you compare before and after a change point.

get_apm_downstream_dependencies

Get the downstream dependencies (services or uninstrumented backends) for a service. Map the downstream dependency name to a service by returning both span.destination.service.resource and service.name. Use this to drill down further if needed.

get_apm_error_document

Get a sample error document based on the grouping name. This also includes the stacktrace of the error, which might hint to the cause.

get_apm_service_summary

Get a summary of a single service, including the language, service version, deployments, the environments, and the infrastructure that it is running in. For example, the number of pods and a list of their downstream dependencies. It also returns active alerts and anomalies.

get_apm_services_list

Get the list of monitored services, their health statuses, and alerts.

get_apm_timeseries

Display different APM metrics (such as throughput, failure rate, or latency) for any service or all services and any or all of their dependencies. Displayed both as a time series and as a single statistic. Additionally, the function returns any changes, such as spikes, step and trend changes, or dips. You can also use it to compare data by requesting two different time ranges, or, for example, two different service versions.

Use contextual prompts

edit

AI Assistant contextual prompts throughout Observability provide the following information:

  • Universal Profiling — explains the most expensive libraries and functions in your fleet and provides optimization suggestions.
  • Application performance monitoring (APM) — explains APM errors and provides remediation suggestions.
  • Infrastructure Observability — explains the processes running on a host.
  • Logs — explains log messages and generates search patterns to find similar issues.
  • Alerting — provides possible causes and remediation suggestions for log rate changes.

For example, in the log details, you’ll see prompts for What’s this message? and How do I find similar log messages?:

obs ai logs prompts

Clicking a prompt generates a message specific to that log entry:

Observability AI assistant example

You can continue a conversation from a contextual prompt by clicking Start chat to open the AI Assistant chat.

Add the AI Assistant connector to alerting workflows

edit

You can use the Observability AI Assistant connector to add AI-generated insights and custom actions to your alerting workflows. To do this:

  1. Create (or edit) an alerting rule and specify the conditions that must be met for the alert to fire.
  2. Under Actions, select the Observability AI Assistant connector type.
  3. In the Connector list, select the AI connector you created when you set up the assistant.
  4. In the Message field, specify the message to send to the assistant:

    Add an Observability AI assistant action while creating a rule in the Observability UI

You can ask the assistant to generate a report of the alert that fired, recall any information or potential resolutions of past occurrences stored in the knowledge base, provide troubleshooting guidance and resolution steps, and also include other active alerts that may be related. As a last step, you can ask the assistant to trigger an action, such as sending the report (or any other message) to a Slack webhook.

Currently you can only send messages to Slack, email, Jira, PagerDuty, or a webhook. Additional actions will be added in the future.

When the alert fires, contextual details about the event—such as when the alert fired, the service or host impacted, and the threshold breached—are sent to the AI Assistant, along with the message provided during configuration. The AI Assistant runs the tasks requested in the message and creates a conversation you can use to chat with the assistant:

AI Assistant conversation created in response to an alert

Conversations created by the AI Assistant are public and accessible to every user with permissions to use the assistant.

It might take a minute or two for the AI Assistant to process the message and create the conversation.

Note that overly broad prompts may result in the request exceeding token limits. For more information, refer to Token limits. Also, attempting to analyze several alerts in a single connector execution may cause you to exceed the function call limit. If this happens, modify the message specified in the connector configuration to avoid exceeding limits.

When asked to send a message to another connector, such as Slack, the AI Assistant attempts to include a link to the generated conversation.

The server.publicBaseUrl setting must be correctly specified under Kibana settings, or the AI Assistant is unable to generate this link.

Message sent by Slack by the AI Assistant includes a link to the conversation

The Observability AI Assistant connector is called when the alert fires and when it recovers.

To learn more about alerting, actions, and connectors, refer to Alerting.

Known issues

edit

Token limits

edit

Most LLMs have a set number of tokens they can manage in single a conversation. When you reach the token limit, the LLM will throw an error, and Elastic will display a "Token limit reached" error in Kibana. The exact number of tokens that the LLM can support depends on the LLM provider and model you’re using. If you are using an OpenAI connector, you can monitor token usage in OpenAI Token Usage dashboard. For more information, refer to the OpenAI Connector documentation.