Cisco Opensourced Tool to Build AI Bill of Materials (github.com)

Cisco has announced the open-sourcing of a new tool called AI Bill of Materials (AI BOM), designed to streamline the inventory and analysis of AI framework components within codebases and container images. This innovative tool performs static analysis on Python source code, leveraging libcst for comprehensive parsing, and matches the extracted symbols against a DuckDB catalog to create an AI BOM. It also offers optional enrichment through large language models (LLMs) to identify and extract model names, enhancing the overall utility of the analysis. The significance of this tool for the AI/ML community lies in its ability to promote transparency and accountability in AI component usage, which is critical as the industry grapples with issues of model provenance and ethical AI deployment. The AI BOM tool not only aids developers in understanding the relationships between components and workflows through a lightweight call graph but also supports various output formats, including JSON and a web UI powered by FastAPI. By enabling easy auditing and reporting of AI components, Cisco's AI BOM tool empowers engineers to effectively manage their AI assets, paving the way for more robust and responsible AI systems.