| Tartalmi leírás: |
Working with 5G core signalling data is very challenging, considering the complexity of the protocols, and the exponential growth of control plane signaling. Service providers must process vast volumes of heterogeneous data generated by protocols such as NGAP, PFCP, and SBI (HTTP/2) to ensure network reliability and optimize services. Traditional analytical tools, which rely on static log parsing, fail to adapt to dynamic protocol changes. This paper introduces the CDR Intelligence Server, a novel architecture that integrates Large Language Models (LLMs) with a deterministic, multi-agent workflow engine. The graph-based orchestration framework that translates natural language intents into executable data processing pipelines. Furthermore, we detail the design of a specialized "Tool Ecosystem" categorized into Filtering, Reduction, Statistics, and Analysis domains. This approach allows the reasoning of LLMs to be safely grounded in precise data manipulation. Extensive evaluation using real-world telecom traces demonstrates that this system significantly improves the ergonomics of network troubleshooting. Operators are able to identify root causes of failures through intuitive dialogues rather than manual query construction.
|
