Engram FAQs

Question 1

What is Engram and how does it help developers?

Accepted Answer

Engram is a GPU-accelerated local semantic code index designed to provide precise context to AI coding assistants like Claude Code. It intelligently indexes your codebase, allowing AI to retrieve only the most relevant code snippets, significantly improving context window efficiency and AI accuracy.

Question 2

What makes Engram's code search and indexing so fast?

Accepted Answer

Engram leverages NVIDIA GPUs with CUDA and ONNX Runtime for high-speed semantic embeddings and an HNSW vector index for approximate nearest-neighbor search. This combination enables sub-3ms query times for rapid, context-aware code retrieval.

Question 3

Does Engram support large or multiple codebases?

Accepted Answer

Yes, Engram offers robust multi-project support, allowing a single process to index and search across multiple codebases simultaneously. It also features incremental updates via file watching and content hashing for efficient, live index maintenance on large projects.

Question 4

How does Engram ensure precise code context for AI?

Accepted Answer

Engram uses AST-aware code chunking with tree-sitter to parse code into precise semantic units (like functions or classes). This ensures that the AI receives focused, relevant snippets rather than broad file reads, enhancing the quality of context.

Question 5

What are the primary technical requirements for running Engram?

Accepted Answer

Engram currently runs on Windows 10/11 and highly benefits from an NVIDIA GPU with CUDA 12.x+ and cuDNN 9.x for its GPU-accelerated features. Core indexing and symbol search functionalities can operate without a GPU.

Engram

Engram

主な機能

ユースケース

主な機能

ユースケース