split-brain-harness 1.1.0

Soul-injected two-stage LLM telemetry pipeline. Wraps any LLM with affective/intent/cognitive analysis, deterministic verification, and a Stage 0 deobfuscation normalizer. Drop-in OpenAI-compatible proxy.
Documentation 
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29

# split-brain-harness configuration
# Copy to config.toml and edit. Environment variables override any value here.
# Path to a custom config file: SBH_CONFIG=/path/to/config.toml

# Backend: "ollama-native" | "openai-compat" | "anthropic" | "local-embedded"
backend = "ollama-native"

# API endpoint (backend-specific default used if omitted)
# ollama-native:  http://localhost:11434
# openai-compat:  http://localhost:8080
# anthropic:      https://api.anthropic.com
endpoint = "http://localhost:11434"

# Model name (backend-specific default used if omitted)
# ollama-native / openai-compat: llama3.2:3b
# anthropic:                     claude-sonnet-4-6
model_name = "llama3.2:3b"

# API key — required for the "anthropic" backend
# api_key = "sk-ant-..."

# Path to a custom soul.md (empty = embedded default)
# soul_path = "/path/to/soul.md"

# Verification mode: "deterministic" | "llm" | "none"
# deterministic — fast consistency checks only, no extra API call (default)
# llm           — deterministic checks + second LLM call using VERIFIER_SYSTEM_PROMPT
# none          — skip verification entirely
verify_mode = "deterministic"