Feature Request: Add Reasoning Level Control Setting

[Karl Livesey]

I would like the ability to configure the reasoning level (e.g., low, medium, high, or extended reasoning) for the AI agent in Rovo Dev CLI.

This could be useful for:

• Controlling the depth of analysis for complex problems

• Balancing response speed vs. thoroughness

• Optimizing token usage for different types of tasks

• Getting more detailed step-by-step reasoning when debugging complex issues

Suggested implementation:

• Add a /reasoning command to set the level interactively

• Include a reasoningLevel setting in ~/.rovodev/config.yml

• Allow per-session or per-model reasoning level configuration

This would give users more control over how the AI approaches different tasks, similar to reasoning controls available in other AI tools.

[Dr Valeri Colon _Connect Centric_]

To quote @Trudy Claspill 

As this is a user community where users help users, it is not the best path to get a feature request to Atlassian.

As a Product Admin you should have access to open a Technical Support ticket to ask Atlassian to open the change request or let you know the link if one already exists.

Alternately under the help button within Jira there is a Feedback option where any user can send their feedback directly to Atlassian. That does not generate s trackable support case.

[Jorge Cammarota]

Based on the official documentation, as of today, the Rovo Dev CLI does not have:

A /reasoning command.

A reasoningLevel key in ~/.rovodev/config.yml.

Explicit configuration for "reasoning level" per session or per model.

What currently exists is:

In Rovo Chat (web interface), there is a selection for "Reasoning mode" (Let Rovo decide / Think deeper / Deep research):

Configure Rovo's output

In the Rovo Dev CLI, you primarily control the agent's behavior via:

agent.temperature

agent.modelId

agent.enableDeepPlanTool

agent.experimental.enableDelegationTool

...along with tool permissions, YOLO, etc.

Manage Rovo Dev CLI settings

Rovo Dev CLI commands

Current Workarounds for "Reasoning Level"

While a dedicated reasoningLevel does not exist, you can approximate the concept using available features:
1. Configure Planning Depth

In ~/.rovodev/config.yml, you can enable deeper planning for complex tasks:
yaml

agent:
enableDeepPlanTool: true # for more structured reasoning in complex tasks

2. Create "Reasoning Profiles" via Different Config Files

For example, create three configuration files:

~/.rovodev/config-low.yml:
yaml

agent:
temperature: 0.1
enableDeepPlanTool: false

~/.rovodev/config-medium.yml:
yaml

agent:
temperature: 0.3
enableDeepPlanTool: true

~/.rovodev/config-high.yml:
yaml

agent:
temperature: 0.4
enableDeepPlanTool: true

Then run the CLI with the desired profile:
bash

acli rovodev run --config-file ~/.rovodev/config-low.yml
acli rovodev run --config-file ~/.rovodev/config-medium.yml
acli rovodev run --config-file ~/.rovodev/config-high.yml

3. Use Prompts or Subagents as "Reasoning Levels"

Since the CLI supports subagents, you can create a dedicated agent for deep reasoning.

Use subagents in Rovo Dev CLI

For example, you could create a deep-reasoner subagent with a system prompt that instructs it to always explain its reasoning step-by-step.

Links:

https://support.atlassian.com/rovo/docs/configure-rovo-chats-output/

https://support.atlassian.com/rovo/docs/manage-rovo-dev-cli-settings/

https://support.atlassian.com/rovo/docs/rovo-dev-cli-commands/