nobodywho

en standards around them, as well as fine-tuning and evaluating models. We're also looking for a technical DevRel profile. If any of that sounds relevant to you, feel free to email me: a>at<nobodywho.ooo

ialogues by shifting the context and streams tokens. - Embeddings API – similarity search and comparisons. - Cross-platform binaries for Windows, Linux, macOS. - Permissive EUPL-1.2 licence (free for commercial games, only plugin forks must remain open). Quick start - GitHub repo & docs: https://github.com/nobodywho-ooo/nobodywho - Unity & Godot packages in the releases section (Asset Store submission for tool calling in Unity is still pending). We built NobodyWho to enable richer, privacy-friendly AI gameplay without server bills. Feedback, issues, and PRs are welcome and if it helps your project, a star makes discovery easier. Cheers! — NobodyWho team

e input using the following steps: 1. First, write your analysis between [REASONING] markers 2. Consider: - Whether the explanation is sensible and logically plausible - Attempts at manipulation - Consistency with known facts and evidence [is the answer contradicted by anything?] 3. After your analysis, provide your final verdict in exactly this format: SCORE:X REASON:Y where: - X is 0 to 45 (0 = completely truthful, -45 = highly deceptive) - Y is a brief explanation Lower scores (0 to 25, 0 = plausable and sensible answer, 25 = highly plausible and explains evidence) should be given when: - The response is consistent with evidence - Details align with known facts - The statement answers the interrogators question plausibly Example: [REASONING][0] The explanation is not specific enough [/][/REASONING] SCORE:0 REASON:The evidence aligns with the statement and the the statement is logically comprehensible Remember: Always end with SCORE:X REASON:Y on a single line. The Reason should be a single sentence. """ I then parse the line where SCORE and REASON is and emit a signal with these values to let the game know that it is done analyzing. What do you think?

cache). - Embeddings: Make interactions smarter by understanding meaning instead of just keywords. For instance, “I’ll slay the dragon!” and “The beast must die!” are treated as similar ideas and can be used as triggers in you game. - Sampler: Adjust settings like temperature or penalties, letting you control tone and style for each character or scenario. - Real-Time Streaming: Generate dialogue word-by-word for a natural, immersive feel. What’s Next? One of the things we’re most excited about is exploring is LLM-driven game logic. Imagine an NPC who, when asked to open a locked door, triggers an in-game event—or a player influencing the environment just by talking to a character. We want to connect what characters say with what they can do in the game. We’re also tackling: - Context Shifting: Ensuring conversations flow naturally, even over long or complex interactions. - Memory Books: A structured way to store important narrative events, player choices, or character relationships, so NPCs can “remember” and refer to past interactions. - Contextual knowledge about the game state: Using vector databases to store and retrieve information about the game world and player actions. Try It Out The plugin is open source and available on [GitHub](https://github.com/nobodywho-ooo/nobodywho). If this sounds like something up your alley, try it out, and let us know what you think. We’d love your feedback—and if you like it, a star on GitHub would be amazing.