AstrMap – Unix Philosophy for the AI Era (Ditch the RAG)

Name: AstrMap – Unix Philosophy for the AI Era (Ditch the RAG)
Availability: InStock
Author: hubby247

by hubby247·Feb 20, 2026·1 point·0 comments

Visit Project View on HN

AI Analysis

●●●BangerZero to OneBig BrainSolve My Problem

Replaces RAG with deterministic AST maps; costs near-zero tokens and works offline.

Strengths

•Targets real pain point in agentic workflows: RAG's unreliability and context loss on code.
•Genuinely novel approach: treats codebase structure as indexable text instead of vectors.
•Practical MVP: single Go binary, multi-language support, runs locally with zero setup.

Weaknesses

•Early stage with limited track record; dependency tracking (PRO feature) not yet proven.
•AST-based indexing assumes well-structured code; handles edge cases and metaprogramming unclear.

Post Description

Hey HN, I got incredibly tired of RAG for code. Vectorizing source code into arbitrary chunks and hoping a cosine-similarity search finds the right structural context is a black box that just doesn't reliably work for agentic coding. It's anti-KISS.

So I wrote AstrMap in Go.

It rips through your repo in milliseconds and generates a highly compressed, AI-readable AST "Map" (like a giant table of contents with line numbers and functions). You feed the Map to your LLM (costs almost 0 tokens), and the LLM uses it as a deterministic radar to pinpoint exactly which files it needs to modify. It treats your architecture as pure, parseable text.

It's completely free, a single binary, local, and supports Go, JS/TS, Python, HTML/CSS. Give it a try and let me know if it speeds up your agentic coding workflows.

Link: https://github.com/hubby247/astrmap