dcm-dcm-dcm/eagerlog

How dcm-dcm-dcm/eagerlog works

EagerLog is an active-learning, retrieval-augmented log anomaly detector. Its main workflow prepares and parses a configured dataset, embeds reference log windows, retrieves similar evidence for each test window, asks Qwen to explain and classify it, and selectively adds uncertain examples to the evolving reference set. Results and metrics are written at the end. The graph preserves the three dataset-specific Drain parsers and prompt roles; unsampled utilities and config details are represented only where the excerpts or README support them.

Main entry points: main.py Detection loop · preprocessing.py Dataset preprocessing · utils.py Detection utilities · chat.py Qwen client

Components 16

Detection workflow

Detection loop main.py
Experiment config config/BGL.yaml
Result files

Dataset preparation

Dataset preprocessing utils/preprocessing.py
BGL Drain parser utils/DrainBGL.py
Thunderbird parser utils/DrainThunderbird.py
Zookeeper parser utils/DrainZookeeper.py

Retrieval and utilities

Detection utilities utils/utils.py
DPR encoder

LLM prompting

Explanation template prompts/explanation_prompt.txt
Evidence template prompts/evidence.txt
Analysis template prompts/analysis_prompt.txt
Qwen client llm/chat.py

Other components

Experiment runner
LogHub datasets
Qwen API

Connections 16

Experiment runner → Detection loop: runs README.md
Detection loop → Experiment config: loads main.py
Detection loop → Dataset preprocessing: prepares data main.py
Dataset preprocessing → LogHub datasets: downloads utils/preprocessing.py
Dataset preprocessing → BGL Drain parser: parses BGL utils/preprocessing.py
Dataset preprocessing → Thunderbird parser: parses Thunderbird utils/preprocessing.py
Dataset preprocessing → Zookeeper parser: parses Zookeeper utils/preprocessing.py
Detection loop → Detection utilities: retrieves and evaluates main.py
Detection utilities → Dataset preprocessing: invokes parsing utils/utils.py
Detection loop → DPR encoder: encodes windows main.py
Detection loop → Explanation template: loads main.py
Detection loop → Evidence template: loads main.py
Detection loop → Analysis template: loads main.py
Detection loop → Qwen client: requests generations main.py
Qwen client → Qwen API: calls API llm/chat.py
Detection loop → Result files: writes CSV main.py

More Python diagrams: vngrs-ai/vnlp · bigcat88/pillow_heif · bionumpy/bionumpy · creativelyons/nukesurvivaltoolkit_publicrelease · irbis3d/medusahc

Drawn by GitDiagram on November 9, 2026 from the repository's file tree, README and source files. Read as Markdown ·