| layout | title | permalink | parent | has_children | nav_order |
|---|---|---|---|---|---|
page |
How GUAC components work together |
/guac-components/ |
How GUAC works |
true |
2 |
{: .no_toc }
Table of contents
{: .text-delta } - TOC {:toc}The full GUAC component deployment is a set of asynchronous services that combine to form a robust and scalable pipeline. In some of our [demos]({{ site.baseurl }}{%link guac-use-cases.md %}), you may have seen these components work in concert. Read on to learn more of what goes on behind the hood!
The GraphQL server serves the GUAC defined nodes through GraphQL queries. It is an abstraction layer for GUAC integrations and other GUAC components. Our setup docs use the built-in, in-memory backend for the server. Currently the server also supports a Neo4j backend. Any future backend database support added to GUAC will not affect the GraphQL interface that the server provides.
| Name | Short Description |
|---|---|
| Collector | Reads or watches locations for new documents and collects them when found |
| Ingestor | Takes documents (ex: SBOMs) and parses them into the GUAC data model/ontology |
| Assembler | Takes GUAC objects and puts them in a datastore, queryable by the GraphQL server |
| CollectSub | Takes identifiers of interest and creates a subscription for collectors to follow |
Collectors gather data from various sources. These sources can be internal to the organization, public (like open source), or third-party vendors. There are different collectors for the different types of locations that GUAC can watch (files, storage buckets, Git repositories, etc.).
Collectors can be configured to use a CollectSub service to know which data sources are of interest. For example, a Git collector may subscribe to the CollectSub service to know which Git repositories it should get its data from. This is a way to get an idea of what data sources the instance of GUAC is looking at, at a glance.
Collectors and certifiers then take the documents and pass them onto the ingestor via Nats.
Another class of GUAC collector components are certifiers. Certifiers run outside the server, but are not part of the ingestion pipeline. They watch the server for new nodes in the server and add additional information attached to those nodes.
For example, the OSV certifier will watch GUAC for new packages, then try to discover OSV vulnerabilities for those packages. If any vulnerabilities are found, the certifier will attach a "CertifyVuln" node to the package to signify that the package is connected to the OSV vulnerability.
Ingestors take in documents and parse them into the [GUAC data model/ontology]({{ site.baseurl }}{%link guac-ontology.md %}). This process extracts meaning from documents and translates them into a common reasoning model (GUAC ontology). In the process, it also finds identifiers of interest that it passes to the CollectSub service to request additional information for.
Today, GUAC can understand multiple data formats like SPDX, CycloneDX, and SLSA. The ingestor listens for documents to parse via NATS, and talks to the assembler via a GraphQL API.
The assembler takes the parsed GUAC ontology objects from the ingestor and creates entries within a database. This database is used as a source of truth for GUAC queries.
The assembler exposes a set of GraphQL mutate interfaces. The assembler is physically part of the GraphQL server, but logically part of ingestion.
The collect subscriber service provides a way to ask for a datasource to be used or indicate that more data about a software identifier is desired. For example, in parsing an SBOM, if it sees the use of a package with a pURL, the ingestor creates an entry to the CollectSub service to indicate more information about the pURL is desired (via gRPC).
The collectors all subscribe to this service and will automatically retrieve more information about the pURL (or other identifier/datasource) if they know how to handle it. For example, the deps.dev collector knows how to handle pURLs and retrieves more information about the pURL entries created from the ingestor parsing the SBOM.