llmcc 0.2.47

LLM Context Compiler - Universal context builder for any language and document type

llmcc

"Prompts are the modern assembly language, models are the modern CPU."

llmcc is a universal context builder for any language, any document.

abstract

llmcc explores automated context generation through symbolic graph analysis. bridging the semantic gap between human-written code/documents and AI model understanding, using modern compiler design principles.

design

examples

llmcc --dir codex/codex-rs --lang rust --design-graph --pagerank --top-k 100

codex: graph

llmcc --dir kimi-cli --lang python --design-graph --pagerank --top-k 80

kimi-cli: graph

run

llmcc [OPTIONS] < --file <FILE>...|--dir <DIR>... >

Input (required, one of):

• -f, --file <FILE>... — Individual files to compile (repeatable)
• -d, --dir <DIR>... — Directories to scan recursively (repeatable)

Language (optional):

• --lang <LANG> — Language: 'rust' or 'python' [default: rust]

Analysis (optional):

• --design-graph — Generate high-level design graph
• --pagerank --top-k <K> — Rank by importance (PageRank) and limit to top K
• --query <NAME> — Symbol/function to analyze
• --depends — Show what the symbol depends on
• --dependents — Show what depends on the symbol
• --recursive — Include transitive dependencies (vs. direct only)

Output format (optional):

• --summary — Show file paths and line ranges (vs. full code texts)
• --print-ir — Internal: print intermediate representation
• --print-block — Internal: print basic block graph

Examples:

# Design graph with PageRank ranking
llmcc --dir crates --lang rust --design-graph --pagerank --top-k 100

# Dependencies and dependents of a symbol
llmcc --dir crates --lang rust --query CompileCtxt --depends
llmcc --dir crates --lang rust --query CompileCtxt --dependents --recursive

# Cross-directory analysis
llmcc --dir crates/llmcc-core/src --dir crates/llmcc-rust/src --lang rust --design-graph --pagerank --top-k 25

# Multiple files
llmcc --file crates/llmcc/src/main.rs --file crates/llmcc/src/lib.rs --lang rust --query run_main

python

Install the published package from PyPI:

pip install llmcc

With the package available, invoke the API directly:

import llmcc

help(llmcc.run)

graph = llmcc.run(
	dirs=["crates/llmcc-core/src"],
	lang="rust",
	query="CompileCtxt",
	depends=True,
	summary=True,
)
print(graph)