# Using Deep Reasoning
#### **Where to Find It**
The **Deep Reasoning Mode** toggle is located at the **top-right corner** of the chat interface.\
You can enable or disable it on a **per-question basis** during a session, allowing you to choose when deeper analysis is necessary.
***
#### **What It Does**
Example of Intent Graph when Deep Reasoning mode is enabled
When enabled, Deep Reasoning Mode activates **multi-intent decomposition** for complex problem solving. The system:
* Breaks down the main question into a **graph of interdependent sub-intents**.
* Generates autonomous **semantic graph** for each sub-intent, allowing for rich contextual exploration.
* Uses these semantic graph to build a structured understanding of the problem, improving interpretability and precision.
This is especially useful for queries involving **compound logic**, **ambiguous phrasing**, or **multi-step analytical chains**.
***
#### **Impact & Things to Consider**
* **Increased Analytical Depth:** Produces more thoughtful, nuanced answers for complex or layered questions.
* **Higher Latency:** This mode introduces **additional compute overhead**, so responses may take longer to generate.
* **Flexible Control:** You can toggle it **on or off for each question**, depending on the level of reasoning required.
***
#### **Inspector Behavior in Sub-Intent Nodes**
The '**Inspector'** - which reveals detailed metadata for entities in the semantic graph on double clicking any node - is currently available **only for entities in the semantic graph of the main intent i.e. the root node, instead of the sub-intents**.
* For **sub-intent nodes**, the Inspector is **disabled by design** to keep these graph streamlined and easier to navigate.
* Clicking on entities in sub-intents will not open the Inspector.
* If deeper details about an entity are needed, navigate back to the **root node’s semantic graph** where inspection is fully supported.
