KnowledgeForge: mining gold from the ITSM ticket graveyard
aws ·
KnowledgeForge is an AWS solution that mines resolved IT service management (ITSM) incident tickets to generate knowledge base articles, addressing the problem of valuable resolution knowledge being locked in ticket histories. It also curates existing knowledge bases by classifying articles, removing duplicates, scoring quality, and rewriting weak content, with a knowledge manager approving the final output. The system uses Amazon Bedrock with Anthropic Claude Sonnet 4.5 for content generation, Amazon S3 Vectors for embedding-based duplicate detection, and AWS Step Functions for orchestrating the curation workflow. Generation runs on Amazon ECS with AWS Fargate because the work is long-running and bursty, and it uses retrieval augmented generation to ground new articles in existing content. Curation is orchestrated as a two-phase distributed map, with sequential deduplication per batch and parallel quality scoring and improvement. The closed-loop design stores vectors for every article so future generation can reuse them, and the whole pipeline integrates with ServiceNow for human review. The article is a technical how-to that explains the architecture, prerequisites, and code repository for deploying the solution.
aws