Tag
JetBrains developers share field notes on building a production-grade RAG pipeline for semantic code search (Air Context), covering parsing, chunking, and vectorization as Part 1 of a series aimed at giving LLM coding agents precise, citable evidence from large codebases.
Benzi is a compiler-backed AI coding agent that uses deterministic static analysis and code indexing instead of reading raw source code, reporting 78.2% on SWE-bench Verified for under 10 cents per fix while reading far fewer tokens than Claude Code, DeepSeek's harness, and OpenCode.
A new open-source tool called codebase-memory-mcp indexes entire codebases like the Linux kernel in minutes using AST knowledge graphs, achieving massive efficiency gains for AI agents with 99% token reduction and 83% answer quality.
A new MCP server reduces Claude Code's tool calls by 94% by indexing the codebase into a local knowledge graph, allowing agents to query the graph instead of scanning files.