Description of the Problem:

When the Thinking models are used in Lightspeed, the chat conversation name/title fails to display correctly or is entirely missing in the sidebar. This issue occurs when a user sends a message to a Thinking model, resulting in the conversation title not being properly generated or displayed.

However, when the Thinking feature is disabled, the chat conversation displays a valid, auto-generated title in the sidebar as expected.

Prerequisites:

• rhdh-plugins Lightspeed workspace

• Thinking feature enabled in Lightspeed settings

Steps to Reproduce:

1. Open the Lightspeed plugin.

1. Start a new chat conversation.

1. Send a message and wait for the AI response that includes Thinking output.

1. Observe the conversation title in the sidebar.

Actual Results:

• When the Thinking is enabled, the chat conversation title in the sidebar does not display a valid name/title. It may appear Incorrectly formatted

Expected Results:

• The chat conversation should always display a meaningful, auto-generated title in the sidebar, regardless of whether the Thinking feature is enabled or disabled. Title generation should be consistent in both scenarios.

Reproducibility:

• Always reproducible when the Thinking model is used.

Build Details:

• Dev mode environment

Additional Information:

• Title generation works correctly when the Thinking feature is disabled.

• The issue likely stems from how response content is parsed when the Thinking output is included.

• The discrepancy in sidebar behavior is apparend when comparing the Thinking model vs Non Thinking Model.

Thinking Models:

• deepseek-r1:8b

• deepseek-r1:7b

• qwen3:1.7b

Non-Thinking Models:

• mistral:7b

• llama3.2:3b