{
  "id": 12117171,
  "title": "When You Need Me 💛 — A Private AI Companion Built for a Friend",
  "url": "https://urgent.news/2026/10/05/when-you-need-me-a-private-ai-companion-built-for-a-friend",
  "topic": "ai",
  "section": "AI",
  "published": "2026-10-05T09:28:07.000Z",
  "source": {
    "name": "Dev.to",
    "slug": "dev-to",
    "url": "https://dev.to/monisha14206/when-you-need-me-a-private-ai-companion-built-for-a-friend-4k27"
  },
  "original_language": "en",
  "account": "What I Built When You Need Me is a private AI companion designed for a real friend. The AI companion fills the gap when someone we care about needs to talk, share their feelings, or simply have someone listen, but we may be busy, unavailable, or unable to respond right away. The project aims to address moments where direct communication might be challenging.\n\nInstead of creating yet another generic chatbot, the AI companion is built to understand the user's communication style and incorporate it with the user's knowledge about their friend. This personalized approach makes the AI response feel more natural and relatable.\n\nTo achieve this, the system performs several key tasks:\n\n1. Analyze approved conversation samples to create a communication-style profile. This includes extracting tone, vocabulary, sentence style, emoji usage, humor, language mixing, and support style.\n\n2. Build a friend profile based on the information the user chooses to provide. This includes personality traits, interests, what makes them happy or upset, preferred support style, humour preferences, topics to avoid, important boundaries, etc.\n\n3. Generate responses using the communication style, friend profile, current conversation, and safety rules. The AI combines these elements using a local open-weight model (Ollama with llama3.2:3b) to create personalized replies.\n\n4. Explainability feature: The system can also explain why a response mimics the user's communication style. For instance, it might mention using a casual conversational tone, familiar expressions, short supportive sentences, and matching the user's emoji style.\n\n5. Safety handling: The AI companion includes a safety layer that assesses the seriousness of conversations. In potentially distressing situations, instead of providing unrestricted responses, the safety layer can offer predefined supportive statements encouraging the user to seek help from trusted individuals or professionals.\n\nA distinguishing feature of this AI companion is that it does not claim to be the user itself. It clearly identifies itself as an AI companion inspired by the user's communication style. This transparency helps maintain the user's understanding of the AI's role and limitations.\n\nThe project prioritizes privacy and safety, running the AI inference locally on the user's computer using open-weight models. This approach ensures personal conversations remain private, offers control over the system's functionality, allows for customization based on the user's communication style, and doesn't require paying for proprietary AI services.\n\nBy leveraging open-source AI technology, the project demonstrates how students and independent developers can experiment with AI applications without needing expensive infrastructure or proprietary services. Open innovation is crucial in this context, as it provides greater control over data handling, AI behavior, and system modifications while emphasizing the importance of privacy and responsible AI usage.",
  "summary": "What I Built When You Need Me 💛 When You Need Me is a private AI companion built for a real friend. Sometimes someone we care about wants to talk, share how they feel, or simply have someone listen — but we may be busy, unavailable, or unable to respond immediately. I built this project to create a small bridge during those moments. Instead of building another generic chatbot, I wanted the AI to…",
  "key_points": [
    "What I Built When You Need Me is a private AI companion for a real friend",
    "AI companion analyzes conversation samples to create communication-style profile",
    "Safety layer offers predefined supportive statements in distressing situations"
  ],
  "editors_take": null,
  "illustration": null,
  "coverage": {
    "outlets": 1,
    "also_reported_by": []
  },
  "ai_generated": true,
  "disclaimer": "Summaries, key points and the editor’s take are written by software from other outlets’ reporting and may contain errors — always check the linked original."
}