Teknium teknium1

Hermes background-review cost-control benchmark — data & harnesses

Reproducibility bundle for NousResearch/hermes-agent PR #49252 (background-review: aux-model routing + context digest + adaptive cadence).

All numbers in the PR come from these files. Benchmarks ran live against the Anthropic API; token usage is the real wire usage read off each messages.stream().get_final_message().usage, split into the four non-overlapping Anthropic buckets (fresh input_tokens, cache_read_input_tokens, cache_creation_input_tokens, output_tokens) and

Hermes Agent Security Policy

This document outlines the security protocols, trust model, and deployment hardening guidelines for the Hermes Agent project.

1. Vulnerability Reporting

Hermes Agent does not operate a bug bounty program. Security issues should be reported via GitHub Security Advisories (GHSA) or by emailing security@nousresearch.com. Do not open public issues for security vulnerabilities.

Required Submission Details

Title & Severity: Concise description and CVSS score/rating.

	unwanted_words = [
	"text-based AI language model",
	"domestic violence",
	"please refrain",
	"derogatory",
	"inappropriate",
	"offensive",
	"racism",
	"racist",
	"racial",

	import torch
	import peft
	import json
	import shutil
	from peft.utils import _get_submodules
	import os
	import bitsandbytes as bnb
	from bitsandbytes.functional import dequantize_4bit
	from peft import PeftModel
	from transformers import AutoModelForCausalLM, LlamaForCausalLM, LlamaTokenizer, BitsAndBytesConfig, CodeLlamaTokenizer