Overview
Condrox AI produces two classes of publications: static documents authored alongside the system (white papers, FAQs, technical references), and automated reports generated by the Evidence Engine on every run. The automated reports are reproducible. Running python scripts/generate_evidence.py regenerates every artifact using the same scenarios and measurement rules.
Static Publications
| Title | Type | Audience | Link |
|---|---|---|---|
| Condrox AI: A Non-LLM Cognitive Architecture for Language Understanding | White Paper | Academic, Researchers | Read → |
| Condrox AI Technical Deep Dive | Technical Reference | Engineers | Read → |
| System Architecture | Architecture Reference | Engineers, Architects | Read → |
| Component Status Report | Status Report | Engineers | Read → |
| Knowledge Store Reference | Reference | Engineers | Read → |
| Technical FAQ | FAQ | Engineers, Reviewers | Read → |
| Development Phases | Engineering Log | Engineers, Historians | Read → |
| Comparison with LLM-based Systems | Comparative Analysis | Press, Researchers | Read → |
| Market Value Analysis | Valuation Analysis | Investors, Business | Read → |
| Media Kit | Press Kit | Press, Media | Read → |
Automated Evidence Reports
Every evidence run produces the following artifacts in Reports/evidence/. They are regenerated automatically and reflect the live state of the system.
| Artifact | Edition | Purpose |
|---|---|---|
*.json | Machine-readable | Full evidence payload for programmatic analysis |
*.html | Dashboard | Human-readable HTML report with metrics and charts |
*_lab_edition.academic.html | Lab Edition | Lab-grade report with methodology, references, and signature |
*_academic_edition.academic.html | Academic Edition | Academic edition with a reasoning focus |
*_peer_review_edition.academic.html | Peer Review Edition | Peer-review edition with conversation evidence |
index.html | Evidence Portal | Static portal with report library, score-history chart, and version history |
Evidence Dimensions
The Evidence Engine evaluates seven dimensions, each with its own evaluator and metric set.
| Dimension | Phase | Metrics |
|---|---|---|
| System Verification | Phase 1 | Boot, API health, scenario runs, intent and mode accuracy |
| Conversational Evidence | Phase 2 | Empathy, relevance, consistency, multi-turn stability, context preservation, no-dictionary-dump |
| Reasoning Evidence | Phase 3 | Step-by-step structure, logical coherence, explanation quality, planning, code analysis, correctness signal |
| Agent & Pipeline Evidence | Phase 4 | Agent usage, metadata, RAG, memory, personality, stage timing |
| Full-System Integration | Phase 5 | End-to-end conversation, reasoning, agent, RAG, and API health |
| Academic Reports | Phase 6 | Lab, Academic, and Peer Review editions with methodology and signature |
| Evidence Portal | Phase 7 | Static portal with score history, report library, and version history |
Latest Report Snapshot
WARN status is driven by the local API health check (API not running) and several conversation and reasoning metrics that accurately flag current response limitations. These are real measurements, not failures to hide.
Reproducibility
All automated reports are fully reproducible. Running the following command in the project root regenerates every artifact:
python scripts/generate_evidence.pyTo open the portal directly from the Condrox runtime manager, run:
condrox evidence-portalFor more details on the Evidence Engine, see the Evidence Engine page.