Preserving application structure at deploy time for self-reasoning services

Short summary

The author argues that bolting LLMs onto observability stacks is expensive, centralized, and prone to hallucinating architecture that was already explicit in source code. They propose a Cognitive Autonomic Framework that extracts a Runtime Semantic Topology from the AST at build time, producing a compact graph of dependencies, failure domains, retry policies, and permitted repairs that travels inside the container. At runtime, an agent reads this self-model instead of guessing system structure from logs, enabling targeted, safe remediation actions.

• •Critiques LLM-based observability as costly, centralized, and prone to architectural hallucination
• •Introduces Runtime Semantic Topology extracted at build time from AST, riding inside the container
• •Agent uses a closed list of permitted repairs per node rather than open-ended LLM guessing

Generated with AI, which can make mistakes.