nauro-core 0.13.4

Shared pure-Python logic for Nauro: parsing, validation, context assembly, constants.

nauro-core

Shared pure-Python library for the Nauro ecosystem. Provides parsing, validation, context assembly, and constants used by both the Nauro CLI and the remote MCP server.

Installation

pip install nauro-core

What's inside

• format — compiled regexes and parse/format functions for the Nauro markdown protocol (decision titles, metadata fields, section headers)
• constants — limits, thresholds, valid values, file paths shared across all Nauro surfaces
• parsing — pure str → dict functions: parse_decision, parse_questions, extract_stack_summary, decisions_summary_lines
• context — build_l0/build_l1/build_l2 context assembly from pre-loaded files and parsed decisions
• validation — structural screening, hash dedup, Jaccard similarity for decision conflict detection

Design principles

• No I/O — no filesystem, no network; callers inject pre-loaded data. Runtime dependencies are compute-only (BM25, parsing, validation); embeddings are an optional extra (nauro-core[embeddings])
• Function injection — callers pass pre-loaded data (files: dict[str, str], decisions: list[dict]); nauro-core never reads files or calls APIs
• Import isolation — enforced via import-linter: nauro-core cannot import from nauro or mcp_server

Usage

from nauro_core import parse_decision, build_l0, compute_hash

# Parse a decision markdown file
decision = parse_decision(markdown_content, "042-use-s3.md")

# Build L0 context from pre-loaded files
context = build_l0(files={"state.md": state, "stack.md": stack}, decisions=decisions)

# Check for duplicate decisions
hash_val = compute_hash(title="Use S3 for storage", rationale="...")

License

Apache 2.0