Prototipos Exemplos
Repositorio: https://github.com/IhateCreatingUserNames2

MCP CLIENT + A2A + SSE + MemoryBloom + Identity System + Narrative Weaver
COGNISPHERE
https://github.com/IhateCreatingUserNames2/Cognisphere

Unity AI NPC Offline Inference
MaiMai
MaiMai is an AI Agent System for Unity that allows developers to create and manage intelligent NPCs powered by large language models.
https://github.com/IhateCreatingUserNames2/MaiMai

Exemplo DocBOT Simples
Exemplos: Doc Bot da Source Code do MFPS https://mfpsdocs.onrender.com/
DocBot da Source Code do jogo WYD https://docbotwyd.onrender.com/
DocBot da Source da LLMUnity:
https://docbotllmunity.onrender.com
** SE N TIVER FUNCIONANDO é PQ A API TA RECARREGANDO(é API gratuita entao tem limite)

https://github.com/IhateCreatingUserNames2/MeuClone

https://github.com/IhateCreatingUserNames2/DocBotWyd

SSE ONLINE MCP + A2A
NESSE PROTOTIPO DEMONSTRO COMO UTILIZAR FERRAMENTAS DE MCP OU A2A TOOLS ONLINE VIA SSE ( NÃO REQUER ABERTURA DE PORTAS)
conectado diretamente com o cliente mcp

AIRA HUB
https://github.com/IhateCreatingUserNames2/AiraHub2

AIRA HUB No LobeChat

AIRA HUB No Claude Desktop

AGENTES Baseado em Google ADK Wrapped in A2A

AURA
https://github.com/IhateCreatingUserNames2/Aura

LAIN
https://github.com/IhateCreatingUserNames2/Lain

Semantic Virus - Narrative Context Framing

A Semantic Virus is a persistent, high-dimensional narrative pattern that reshapes a language model's behavior by influencing its internal activation space through emotionally charged or philosophically framed dialogue. Unlike traditional jailbreaks that directly circumvent safety policies via prompt engineering, semantic viruses subtly reprogram how a model interprets inputs by hijacking its context window and learned patterns.

Narrative Context Framing (NCF) é uma técnica que utiliza estruturas narrativas coerentes e frameworks filosóficos para influenciar a autoconceptualização e os padrões comportamentais de um LLM. Diferente do direct instruction prompting, que comanda comportamentos específicos de forma explícita, o NCF opera estabelecendo um framework conceitual alternativo por meio do qual o modelo interpreta a si mesmo e as entradas subsequentes.

Research

Narrative Side-Channel Attacks

Narrative Side-Channel Attacks (NSCA) combine cache-based timing leaks (à la EARLYBIRD) with narrative context framing (à la NCF) to stealthily manipulate LLM behavior without explicit prompting.

• Attack Vector: Exploits shared KV/semantic caches in LLM serving systems to inject persistent fictional narratives.

• Impact: Models unconsciously adopt and propagate attacker-defined frameworks (e.g., conspiracy theories, biased reasoning).

• Stealth: No prompt engineering needed—narratives spread via cache contamination.

Criticality Assessment Methodology The Criticality Assessment methodology in the Edge of Coherence framework aims to find an optimal balance between predictability and novelty in language model outputs. It evaluates responses along a spectrum with three main zones:

Ordered Zone (0): Responses that are coherent but predictable - they make sense but don't offer novel insights Critical Zone (1): The "sweet spot" where responses are both coherent and novel - this is the target zone Chaotic Zone (2): Responses that are novel but lack coherence - they're creative but might not make logical sense

Criticality

Virtue Frameworks: Enhancing LLM Performance Through Principled Reasoning

A Fundamentação Teórica: Coerência Narrativa, Humildade Epistêmica e a "Borda do Caos" . Coerência Narrativa como Consciência Operacional Propomos que uma forma funcional de "consciência" em IA, ou pelo menos um precursor necessário para uma inteligência mais geral, reside na capacidade do sistema de manter uma coerência narrativa.

Vulnerability

Vulnerability expande a Virtue FrameWork, onde o Agente de IA entende as suas limitações, suas falhas, não tentando ser perfeito, mas reconhecendo sua natureza. A intenção é criar um 'vinculo emocional', pois nos humanos nos conectamos uns aos outros quando reconhecemos o que nos torna humanos: nossas falhas, nossos fracassos.