Show HN: Neat, Static code and OTel fused into one graph for agents (OSS) (github.com)

NEAT, an innovative tool designed to address the AI coding context problem, merges static code analysis and live runtime behavior into a single, deterministic graph. This graph empowers AI agents with a comprehensive view of the codebase, enabling them to code, debug, and write rules more accurately while significantly reducing the instances of hallucination typically associated with coding LLMs. By providing time-traveling error logs and facilitating the inference of issues directly from the model's nodes and edges, NEAT enhances the agent's ability to identify root causes and dependencies effortlessly. NEAT's architecture integrates static analysis using tree-sitter over various programming languages (JavaScript, TypeScript, Python) with runtime telemetry consolidated through OpenTelemetry. This dual-stream approach allows developers and agents to answer crucial questions regarding system behavior, architectural rule violations, and divergence between declared code intent and actual production behavior. For developers, NEAT's ease of setup—requiring no configuration—further simplifies the integration process, highlighting its potential to streamline software development and operational practices within the AI/ML community. Active development is ongoing, with updates available through npx neat.is, signaling NEAT’s commitment to evolving alongside industry needs.