Obsy Copilot¶

Obsy Copilot is provided with the workspace and does not require self-deployment; the owner must first enable the workspace AI feature. Users can ask questions directly in natural language, or let Obsy Copilot perform analysis or content generation on specific pages based on the current object, accessible workspace data, and enabled capabilities.

Obsy Copilot is designed for immediate tasks initiated by users, helping them complete data queries, analysis, and content creation within the workspace more efficiently and accurately.

Before use, the workspace owner must enable the Workspace AI feature under Manage > Workspace Settings > AI Analysis. The free features also require this switch to be enabled first.

Getting Started¶

Click the Obsy AI icon in the top navigation bar of the platform, and the Obsy Copilot drawer expands on the right side of the page. You can open Obsy Copilot on any page and switch pages or continue operating the platform during the conversation. Click the icon again to collapse the drawer.

The top of the drawer provides an entry point to Obsy Agent Teams; click it to enter the corresponding workbench.

On supported sites, you can also directly select a workspace Agent within the drawer, start a task with the current page context, and view progress and results. To continue asking follow-up questions, open the original task conversation. For instructions, see Using Agents in the Workspace.

The Obsy Copilot drawer supports the following general operations:

New conversation: Click + to create a new conversation that does not inherit the current conversation context;
Conversation history: View, continue, or manage previous conversations;
Maximize or restore: Expand to full screen to view complex content; click again to restore the default width;
Close: Collapse the drawer. The current conversation content is retained, and running tasks continue in the background.

Asking Questions with Current Page Context¶

When you open Obsy Copilot on a supported page, it can understand your question based on the current object, time range, and filter conditions, reducing the need to repeatedly describe page information. For example, on an error trace details page, you can ask "Why did this trace fail?", and on the Host list page, you can ask "Which hosts require priority attention?".

It is recommended to use it as follows:

First navigate to the details page or list page you want to analyze, and set an appropriate time range and filter conditions;
Open Obsy Copilot and describe the phenomenon, risk, or result you want to understand;
If you have switched objects or adjusted filter conditions, confirm that the page already displays the new analysis scope before asking.

Page context only takes effect on supported pages and does not expand the current user's data or feature permissions. When the information provided by the page is insufficient, Obsy Copilot will explain the missing evidence or ask you to provide more conditions. For the supported scope and usage, see Observability Analysis.

Selecting Model Intelligence¶

Before sending a question, you can select the model intelligence level in the input area based on task complexity:

Level	Use Cases
Auto	Automatically balances response speed and analysis effectiveness based on the task; suitable for most everyday questions.
Fast	Prioritizes response speed; suitable for simple queries, content organization, and quick confirmation.
Standard	Balances accuracy and efficiency; suitable for routine analysis and content generation.
Pro	Uses deeper reasoning; suitable for complex troubleshooting, planning, and multi-step analysis.

If you are unsure which to choose, Auto is recommended. Model intelligence only affects how subsequent questions in the current conversation are handled; it does not expand the data, features, or operational permissions accessible to the current user. When a higher level is used for complex tasks, analysis may take longer and consume more credits.

Selecting the Answer Format¶

Click Answer Format at the top of the conversation window to choose how answers are presented according to your reading preference:

Rich cards: Present answers as structured cards, highlighting information hierarchy and key points;
Markdown: Present answers in Markdown format, preserving the original text structure.

Viewing the Run Process¶

After sending a question, you can view the current execution steps and progress in the answer area, and expand or collapse the run process. Steps with process output can be expanded to view specific details. After execution completes, you can still expand to view the run process of this session.

Background Task Execution¶

After starting a task, you do not need to keep the sidebar open. You can collapse the drawer and switch pages in the current workspace or view other conversations; the task continues running.

Click the Task Progress icon to the right of Obsy Copilot in the top navigation bar to view task status, and open the corresponding conversation to view progress, results, or the reason for incompletion. After a task completes or fails, the entry updates its status. After you view the results, the corresponding unread reminder is cleared. To stop a task, open the corresponding conversation and stop it explicitly; collapsing the drawer does not stop the task.

During background execution, keep the current browser tab open. After refreshing the page, closing the tab, logging out, or switching workspaces, the background execution state of this page cannot be restored.

Waiting and Automatic Retry¶

When the model service is busy or requests are rate-limited, Obsy Copilot automatically waits and retries, and the page displays the waiting time and retry progress. There is no need to resubmit your question while waiting. If it still has not completed after retries, please try again later as prompted on the page.

Conversation History¶

Obsy Copilot saves conversation history so you can continue previous analysis. Click the conversation history icon at the top of the drawer to view the conversation list sorted by most recent update time.

In the conversation history list, you can:

Click a conversation to restore the existing context and continue asking questions;
Rename conversations for easier identification later;
Delete conversations that are no longer needed.

Note

Conversation history is isolated by workspace and user, and is visible only to the current user;
Conversation history is retained for up to 30 days and is automatically cleaned up after the retention period.

Quick Questions¶

When you first open Obsy Copilot or start a new conversation, the window displays quick questions, for example:

What can you do?
Tell me about the platform's core features.
Help me summarize the latest alert description.

Quick questions help users get started with a conversation. When performing a specific task, it is recommended to include the analysis object, time range, and expected result. For example, instead of just entering "Help me take a look", you can enter "Analyze the current error trace, explain the most likely failure node, impact scope, and next troubleshooting recommendations".

Capability Overview¶

Obsy Copilot capabilities are divided into the following four categories based on usage goals:

Capability Category	Included Capabilities	Key Value
Observability Analysis	Page Analysis and Root Cause Identification, Suspicious App Blank Screen Detection, Infrastructure Page Analysis, List and Asset Analysis, Database Query Optimization, Alert Aggregation and Compression, Initial Incident Analysis.	Combines the current object, page scope, and observability data to extract anomaly clues, shortening the time for incident assessment and root cause investigation.
Data Retrieval	DQL / PromQL Assistant, Natural Language Search in Explorer.	Converts natural language into query statements or filter conditions, lowering the barrier to using DQL, PromQL, and Explorer syntax and improving data retrieval efficiency.
Resource Generation	Create and Modify Dashboards, Generate Pipeline Parsing Rules, Parse Fields from Log Details and Create a Pipeline.	Generates dashboard drafts and Pipeline parsing rules based on business goals and workspace metadata, reducing repetitive configuration and the effort of writing from scratch.
Knowledge Services	Create Notes and Runbooks, Product Documentation Q&A.	Organizes analysis results into reusable notes or operational procedures and provides usage guidance based on product documentation, reducing the cost of knowledge organization, lookup, and sharing.

The completeness and accuracy of Obsy Copilot's analysis results and generated content may be affected by factors such as site version, deployment configuration, user permissions, current page context, and the completeness of observability data collection. When using the results, it is recommended to verify key conclusions against raw data and the actual environment, and complete necessary validation before performing subsequent operations.

Recommendations¶

When you need to analyze an object, first navigate to the corresponding details page or list page, then open Obsy Copilot. This reduces repeated description and keeps results within the current filter scope.
State the goal, time range, and expected output in your question. When you need to generate queries or configurations, also provide the data domain, samples, and key constraints.
Obsy Copilot only uses data and features that the current user is authorized to access. When data, permissions, or necessary context is missing, the results will state the limitations; do not assume the issue has been fully verified based on that.
Confirm the queries, configurations, and remediation recommendations generated by Obsy Copilot before using them in production environments. Processes involving resource creation remain subject to the permissions and save operations of the corresponding modules.