Search

Search across all your SBOMs, components, and vulnerabilities with Griffin AI-powered natural language queries instead of building filters by hand.

AI-Powered Search

The Search page is powered by Safeguard's Griffin AI models, purpose-built for software supply chain security — see Griffin AI for more on the model itself. Griffin is tuned to understand:

• Security context and vulnerability semantics
• Component relationships and dependency chains
• License compliance requirements
• Supply chain risk factors

You can switch between available model tiers from the model picker in the search bar; each tier trades off speed against depth of analysis:

Previous-generation versions of each tier stay available after a new one ships, and upcoming tiers may appear in the picker ahead of release — those are disabled until launch.

Natural Language Search

Griffin AI understands natural language queries:

"Show me all critical vulnerabilities in production"

"Find components with GPL licenses"

"What packages are affected by Log4j?"

"Show outdated dependencies in the payments service"

"Which projects have exploitable vulnerabilities?"

"List all components from untrusted suppliers"

Search Syntax

Basic Queries

You don't need special syntax for a simple lookup — plain terms work as-is:

lodash                    # Find component named lodash
CVE-2021-23337            # Find specific CVE
react 18.2.0              # Find specific version

Filters, expressed in plain English

Search isn't built around a query DSL with operators you have to memorize — Griffin reads filter-like intent (severity, license, project, reachability, exploitability, and so on) directly out of a natural-language sentence:

critical vulnerabilities in my-app                  # severity + project
MIT-licensed components                              # license
reachable and exploitable vulnerabilities only        # reachable + exploitable
lodash, but not deprecated versions                   # exclusion
critical or high severity findings                    # either/or

Inline mentions

While typing, a few trigger characters open a picker to reference something specific instead of describing it in words:

For example, typing @ and a few letters lets you pick a project to scope the query to, without typing its full name.

Search Results

Results are organized by type:

Components

• Package name and version
• Projects where it's used
• Vulnerability count
• License information

Vulnerabilities

• CVE identifier
• Severity and CVSS score
• Affected components
• Fix availability

Projects

• Project name
• Last scan date
• Security score
• Compliance status

Starred Searches

Save frequently used queries as starred searches:

1. Perform your search
2. Click the Star icon on the conversation
3. Access it later from Starred searches in the sidebar

The Search sidebar

The Search sidebar gives you:

• New chat — start a new search conversation
• All chats — every previous search conversation
• Starred searches — quick access to conversations you've starred
• Connectors, Skills, and Workspaces — jump to the connectors, skills, and workspaces available to your search

Within the conversation list itself, anything you've starred is grouped under Starred conversations, with everything else listed under Recent chats.

Starred searches are most useful for queries you come back to regularly — a recurring triage sweep, a saved vulnerability filter, or a findings query you check on a set cadence.

Search Conversations

Griffin supports conversational search:

1. Start with a query
2. Ask follow-up questions
3. Refine your search through dialogue
4. Griffin maintains context throughout the conversation

Example conversation:

You: "Show me critical vulnerabilities"
Griffin: [Lists critical vulnerabilities]
You: "Which of these are exploitable?"
Griffin: [Filters to exploitable vulnerabilities]
You: "Can you fix the top 3?"
Griffin: [Provides AI Remediate options]

Programmatic Access

The Search page itself is a UI-driven, conversational experience rather than a REST endpoint you call directly. If you need to pull the same findings, vulnerabilities, and SBOM data programmatically — for example, to reproduce a saved search as part of a script or pipeline — use the Safeguard REST API and client SDKs instead. See SDKs for available client libraries and the API Reference for the underlying endpoints.

Export Results

Export search results in multiple formats:

• JSON - Raw data
• CSV - Spreadsheet compatible
• PDF - Formatted report

FAQ & Troubleshooting

My query returns no results even though I know the component/CVE exists. Search runs against your indexed SBOM and scan data, so results are limited to projects and scans you have access to and that have already been processed — a component from a scan still in progress may not show up yet. Try a broader, plain-English version of the query before assuming the data is missing.

Do I need to learn filter syntax, or can I just type a question? Just type a question — Search doesn't use a formal query language with operators to memorize. Describe what you want (severity, license, project, reachability, exploitability, and so on) in plain English and Griffin extracts the filters for you. The one exception is inline mentions (@, #, /, >, &, :), which open a picker to reference a specific project, config, skill, workspace, connector, or entity mid-query.

Which model tier should I use? Use Lino for quick lookups, Eagle for everyday queries, and Griffin when you need deeper reasoning over complex, multi-condition security questions.

Related

• Griffin AI — more on the AI model that powers Search.
• Vulnerabilities — browse and triage vulnerabilities outside of Search.
• Explore SBOM — navigate SBOM data directly.
• Findings — security findings surfaced across your projects.