BGTS Context Engine

•
180
•

Generates deterministic, graph-based code context for AI coding agents, integrating PostgreSQL, Apache AGE, and pgvector for accurate retrieval.

The BGTS Context Engine revolutionizes how AI coding agents interact with unfamiliar codebases by providing deterministic, graph-based context rather than imprecise embedding searches. Unlike traditional retrieval methods that return text similar to a query, this engine indexes repositories into a rich graph of symbols, calls, references, and type hierarchies to pinpoint the exact code that participates in a specific behavior. This approach ensures that for the same task and commit, agents receive identical, reproducible context packs, allowing developers to debug agent failures effectively and improve their reasoning by understanding precisely what information was provided.

主要功能

01Offers incremental re-indexing based on git diffs to maintain index currency efficiently
02Builds a comprehensive code graph instead of relying on text chunks
03Ensures deterministic and reproducible context retrieval for consistent agent behavior
04Supports six programming languages with automatic cross-language call edge detection
05Runs offline with a deterministic embedding provider or connects to local OpenAI-compatible servers
0657 GitHub stars

使用案例

01Provide precise and relevant code context to AI coding agents for complex tasks
02Understand the blast radius and impact of code changes across repositories and languages
03Debug and improve AI agent reasoning by replaying the exact context provided during a task