Graph-Based SDK Schema Retrieval for Network Troubleshooting Agents
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Implementing a network troubleshooting agent using large language models (LLMs) is challenging due to the complexity of processing large quantities of data from network controllers, requiring multiple API calls and aggregation, and the need for relevant software development kits (SDKs) to provide accurate answers, which current static tools cannot handle effectively.
Innovation Solution
A device maintains a graph-based schema representing entities and relations in a computer network, using a retrieval augmented generation module to generate a schema in a programming language for an LLM-based troubleshooting agent, enabling efficient processing and answering queries.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If multiple API calls and data aggregation are performed to answer network troubleshooting questions, then the accuracy and completeness of the answer is improved, but the response time and processing complexity increases
Solution Approach 1:
The patent pre-generates and stores natural language queries along with their corresponding API call sequences and processing logic in a knowledge base before actual troubleshooting operations. When a user asks a question, the system performs a graph search to retrieve pre-prepared schemas and API sequences that directly answer the query, eliminating the need for real-time multi-step API orchestration and data aggregation. This preliminary preparation resolves the contradiction by providing accurate answers through pre-computed knowledge while maintaining fast response times.
2Device complexity
If static tools are used for processing network data, then the system complexity is reduced, but the capability to handle complex analytical and troubleshooting questions deteriorates
Solution Approach 1:
The patent introduces a graph-based retrieval-augmented generation system as an intermediary between the user's natural language query and the network controller's APIs. The graph search component retrieves relevant pre-defined schemas and API call sequences based on the query, acting as a mediator that translates complex troubleshooting requirements into structured API operations. This intermediary layer enables the system to handle complex analytical questions with high adaptability while maintaining manageable system complexity through modular architecture and pre-computed knowledge structures.
3Measurement precision
If relevant SDK information is provided to the LLM to enable correct API calls, then the troubleshooting accuracy is improved, but the information processing load and schema generation complexity increases
Solution Approach 1:
The patent extracts and isolates only the specific SDK schemas, API call sequences, and processing logic that are directly relevant to answering a given troubleshooting query. Instead of providing the LLM with comprehensive SDK documentation, the graph search retrieves a focused subset of pre-defined schemas matching the query intent. This extraction approach maintains high troubleshooting accuracy by providing necessary API information while reducing schema generation complexity through selective, query-specific information delivery.
Data Source
AI summary
In one implementation, a device maintains a graph in which nodes of the graph represent entities in a computer network and edges of the graph represent relations between those entities. The device performs a search of the graph based on a query for input to a language model-based troubleshooting agent for the computer network. The device generates, based on the search, a schema in a particular programming language to answer the query. The device provides the schema to the language model-based troubleshooting agent to generate an answer to the query.