Unverified50% confidenceTradeoffExact time
AST-based chunking using tree-sitter improves retrieval accuracy compared to splitting by line count or character count, because it returns complete logical code units
1
Sources
50%
Confidence
Long-term
Relevance
8/8/2026
First Seen
Sources
Reference：为AI Agent打造的本地语义搜索引擎
Related Claims
UnverifiedReference uses tree-sitter for code-aware chunking, dividing content by semantic units such as functions and classes rather than mechanically splitting by character count.77% similarUnverifiedProviding specific input/output examples (few-shot prompting) can significantly improve the accuracy of code generation, especially when handling edge cases and exception logic.69% similarUnverifiedSymbol uses Tree-sitter syntax parsing technology to refine retrieval granularity to semantic units such as functions and classes.69% similarUnverifiedIn code scenarios, slicing by function boundaries preserves more complete semantic units than slicing by fixed character counts.68% similarUnverifiedAST deobfuscation is a technique that parses obfuscated JavaScript code into a tree structure and restores code readability by traversing and transforming nodes.67% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/712598API
curl https://kongchang.com/api/v1/knowledge/claims/712598MCP
get_claim(id=712598)