Repository navigation
LogAsis v0.10.0 — AI Analyst Accuracy & Investigation Improvements
Release date: October 2026
Version: v0.10.0
LogAsis v0.10.0 introduces significant improvements to log analysis, evidence-grounded question answering, investigation workflows, and evaluation tooling. This release focuses on making analysis more consistent, traceable, and reliable by strengthening deterministic query processing, claim verification, and evidence-based responses.
🚀 What's New
Evidence-Grounded Question Answering
- Added a deterministic QA engine for supported log questions.
- Improved question understanding and extraction of relevant query filters.
- Strengthened claim grounding and verification to help keep answers tied to available evidence.
- Improved handling of exact-count questions over indexed log events.
Investigation & Log Analysis Improvements
- Enhanced AI investigation and analyst response workflows.
- Improved event indexing and dataset handling.
- Refined query processing and timeline-related analysis.
- Added incident profiling capabilities to support structured investigation context.
- Improved dynamic question suggestions for exploring log datasets.
Benchmarking & Evaluation
- Added training and evaluation utilities for preparing datasets, generating questions, and benchmarking analysis.
- Added benchmark artifacts, including prepared datasets, generated questions, predictions, and evaluation results.
- Expanded automated tests covering answer quality, claim grounding, deterministic QA, investigation behavior, incident profiling, parsing, and query processing.
Documentation
- Updated the README and changelog for v0.10.0.
- Expanded AI investigation and evaluation documentation.
- Added accuracy-improvement documentation and an audit report dated October 17, 2026.
🧪 Validation
- Automated tests: 121 passed.
- Deterministic benchmark: 18 questions evaluated.
- QA accuracy on the reported benchmark: 100%.
- Exact-count accuracy on the reported benchmark: 100%.
- Citation validity: 100%.
- Evidence recall: 100%.
- Unsupported-claim rate: 0% in the reported benchmark.
Benchmark results apply to the evaluated test set and deterministic mode. They should not be interpreted as a guarantee of equivalent accuracy across all log sources or AI providers.
🔧 Technical Changes
- Added
core/deterministic_qa.py. - Added
core/incident_profile.py. - Added
core/question_suggestions.py. - Added
core/question_understanding.py. - Added the
training/utilities and expanded thetests/suite. - Updated AI agent, provider, claim-grounding, event-indexing, dataset, query-engine, parser, GUI, and investigation-related components.
⚠️ Notes
- AI-provider results may differ from deterministic QA results.
- Analysis remains dependent on the quality, completeness, and format of the available logs.
- Verify provider-specific behavior and real-world log scenarios before relying on results for operational security decisions.
📦 Installation & Upgrade
Download or clone the repository, check out the v0.10.0 tag, install the required dependencies, and follow the setup instructions in the README.
Repository: https://github.com/TechParthOps/LogAsis
Release tag: v0.10.0
🙏 Acknowledgment
Thank you to everyone testing and improving LogAsis. This release advances the project toward more evidence-driven, transparent, and dependable log analysis.