PyPI page
Home page
Author:
None
License:
MIT
Summary:
Cognitive observability for LLM agents. @styxx.profile returns a per-step cognometric readout: drift, confabulation, refusal, sycophancy, phase transition, low trust, incoherence — localized to the step that produced them. Calibrated AUC: 0.998 hallucination (HaluEval-QA), 0.976 refusal (XSTest-GPT-4), 0.943 tool-call drift (BFCL v3) — text-only register instruments with documented construct ceilings. Reference implementation of the Cognometric Fingerprint Specification v1.0. Pure Python.
Latest version:
7.24.3
Required dependencies:
mcp
|
numpy
|
scikit-learn
Optional dependencies:
anthropic
|
circuit-tracer
|
crewai
|
cryptography
|
guardrails-ai
|
jsonschema
|
langchain-core
|
langfuse
|
langsmith
|
llama-index-core
|
matplotlib
|
openai
|
pillow
|
pyautogen
|
pytest
|
ruff
|
scipy
|
sentence-transformers
|
torch
|
transformer-lens
|
transformers
Downloads last day:
8
Downloads last week:
462
Downloads last month:
2,223