nsml-parser 0.1.0

NSML++: Neuro-Symbolic Memory Lattice for Intelligent Log Parsing

NSMLParser (Neuro-Symbolic Memory Lattice Parser)

NSMLParser is a lightweight, intelligent log parser designed to extract structured key-value fields from unstructured logs using a hybrid of fuzzy matching, positional reasoning, and heuristic inference. It includes:

• Fuzzy key correction
• Backfilled value inference
• Canonical semantic mapping
• Confidence scoring
• Lightweight tokenizer (no spaCy or heavy dependencies)

Features

• ✅ Intelligent tokenization and parsing
• ✅ Fuzzy matching for variant key names (case_number, casenum → caseNumber)
• ✅ Heuristic inference for misplaced or missing key-value pairs
• ✅ Positional backtracking for unstructured logs
• ✅ Confidence scoring per field
• ✅ Designed to work without external APIs or heavyweight NLP libraries

Example

from nsml_parser import NSMLParser

parser = NSMLParser()

log = """Running HTTPS_PROXY='http://proxy.net:8080' /opt/radar/bin/apprise -consumers "FUNNYTEAM_AUDIT" -cluster "POD109" -product "blah-base" -sp "KP" -dc "MM" -caseNumber -release "rrrrr" -event "ERROR_RUNNING_PLSQL_TARGET" -eventTime "2023-06-09_04:31:04" -stage "some issue here" -status "FAILED" -type "ERROR" -source "ops0-release2-2-RR" "123456" --message "something" """

result = parser.parse(log)

print(result)