AI Insights
Overview
The Insights Assistant in DeepTraQ turns complex network-perimeter vulnerability data into concise, actionable intelligence using state-of-the-art AI. From a scan result (for example, the Azure DevScan), you can launch the Insights Assistant to generate a structured threat analysis, perform root-cause analysis (RCA), and produce remediation artifacts such as patch scripts — all without leaving the platform. The assistant is model-agnostic and can enrich answers with real-time web search results while providing direct links to the sources used.
Supported Platforms
- Network perimeter and cloud-hosted scans (example shown: Azure DevScan).
- Model-agnostic AI engines (selectable at runtime): ChatGPT, Claude, Gemini, and other configured providers.
- Web search integration for grounding AI responses with live internet sources (security blogs, exploit databases, CVE listings).
Prerequisites
- Access to the Insights Assistant feature in your DeepTraQ instance.
- At least one AI model configured in the model selector (e.g., ChatGPT, Claude, Gemini).
- Web search tool enabled if you require the assistant to perform real-time internet lookups.
- User permissions to view scan results and launch the Insights Assistant.
Steps
- Open the network scan you want to investigate (for example, Azure DevScan) and locate the relevant network-perimeter findings.
- Click the DeepTraq AI Insights icon next to the finding to launch the Insights Assistant.
- (Optional) Choose a pre-built prompt template to speed triage and standardize the investigation.
- Open the model selector and pick the AI engine you want to use for the analysis (examples: ChatGPT, Claude, Gemini).
- Ask the assistant to perform a root-cause analysis (RCA) for the selected vulnerability — for example: “Perform RCA for this network-perimeter vulnerability and list affected components.”
- DeepTraQ injects CVE data from the internal database and (if enabled) performs additional web search to enrich context while producing the RCA.
- If you want additional perspectives, change the underlying AI model mid-conversation and request the same analysis again to compare outputs.
- Type follow-up prompts directly in the chat to refine the analysis or ask for specific artifacts (for example, “Generate a patch script for this vulnerability”).
- Review the generated patch script in the assistant output and use the provided actions to download or apply the script according to your workflow.
- Inspect the References or source links included in the response to validate the agent’s findings — click through to external security blogs, exploit databases, or CVE pages for deeper validation.
- Use the chat history panel to review previous investigation sessions or to pick up where you left off; all assistant interactions are automatically logged.
Field Reference
| Field | Description | Example |
|---|---|---|
| DeepTraq AI Insights icon | Launches the Insights Assistant for a specific finding directly from the scan results. | Click the AI icon next to an Azure DevScan finding. |
| Pre-built prompt templates | Ready-made prompts to accelerate triage and ensure consistent investigations. | "Root-cause analysis and remediation steps." |
| Model selector | UI control to choose the AI engine used for the analysis. | Select "Gemini" or "ChatGPT" from the dropdown. |
| Chat input | Area to type prompts and follow-up questions to the assistant. | "Generate a patch script for CVE-2024-XXXX." |
| CVE data injection | Automated enrichment where DeepTraQ injects internal CVE metadata into the model context. | CVE details, affected components, observed indicators. |
| Web search tool | Optional capability that allows the assistant to query the internet and cite sources for grounding. | Search results from exploit-db or security vendor advisories. |
| Patch script generator | Assistant output that produces a candidate remediation script for the vulnerability. | Bash or PowerShell patch script to update a vulnerable service. |
| References / Source links | Clickable external links that back the assistant's conclusions to prevent hallucinations. | Link to a vendor advisory or CVE entry used by the assistant. |
| Chat history (Investigation log) | Automatically saved conversation history for each Insights session to allow review and continuation. | Previous RCA session with timestamp and model used. |