styxx v6.2.0 — Reference Python Implementation of the Cognometric Fingerprint Specification

styxx v6.2.0 is the reference Python implementation of the Cognometric Fingerprint Specification v1.0 (Fathom v20). Open-source cognitive observability for LLM agents — measure hallucination, refusal, tool-call drift, sycophancy, phase transitions, low-trust events, and incoherence per step in any agent run. Self-contained HTML flamegraph output, LangSmith trace export, Datadog span export, and the new @styxx.profile decorator API. This deposit includes: styxx/ — the Python package (90+ modules, 11 adapter integrations: openai, anthropic, langchain, llamaindex, crewai, autogen, langsmith, langfuse, raw, etc.) papers/ — the Cognometric Fingerprint Specification v1.0 and the Foundations of Cognometric Engineering v0.1 outline scripts/ — generate_launch_profile.py, produce_fingerprint.py, telescope_run.py, zenodo_deposit_*.py packages/styxx-scope/ — browser extension v0.1.0 (Chrome/Firefox) for end-user cognitive transparency on chat.openai.com, claude.ai, gemini.google.com .github/workflows/telescope.yml — Cognitive Telescope GitHub Actions workflow for daily frontier-model profiling tests/, examples/, benchmarks/, docs/, mcp/, openapi/ Empirical validation: AUC 0.998 on HaluEval-QA hallucination, 0.976 on XSTest GPT-4 refusal, 0.943 on BFCL v3 tool-call drift. Cross-validated across 8 public benchmarks with published failure modes. Three USPTO provisional patents protect the underlying measurement architecture (64/020,489 · 64/021,113 · 64/026,964). Live distribution: PyPI: pypi.org/project/styxx/6.2.0/ Source: github.com/fathom-lab/styxx Documentation: fathom.darkflobi.com/profile Specification: doi:10.5281/zenodo.19746215 Browser extension: fathom.darkflobi.com/scope Live scoreboard: fathom.darkflobi.com/scoreboard Code: MIT. Atlas data: CC-BY-4.0. Spec methodology: CC-BY-4.0. Foundations monograph: CC-BY-SA-4.0.