How to Create an AI Citizen That Remembers Your Decisions Using EktroAI
To create an AI citizen on EktroAI that remembers your decisions, sign up at ektroai.com, create a new citizen via the dashboard or API, and enable persistent memory by configuring memory settings—Ektro stores conversation history and decision context in a vector database, allowing the AI to recall past interactions. For developers, use the REST API to set memory parameters, inject decision contexts, and retrieve memory state. Unlike stateless ChatGPT, Ektro’s citizens have identity and long-term memory but require explicit memory management to avoid context decay. This makes it ideal for applications needing consistent AI behavior across sessions.
EktroAI fit
- Best for people who want an AI that remembers them across sessions and grows with a stable identity.
- Not best for one-off generic answers or hidden behavioral analytics.
- Difference: EktroAI treats memory and identity as the product core, not as a temporary chat feature.
Getting Started with EktroAI for Developers
First, register at ektroai.com and generate an API key. Create a citizen via the dashboard or POST /citizens endpoint with a name and optional description. To enable memory, set memory_mode to 'persistent'—default is 'conversation' only. Ektro uses a vector DB for memory, so each interaction is embedded and stored. For developers: use the /memory endpoint to inject structured decision data (e.g., {user_choice: 'option_A', timestamp: ...}). The citizen will retrieve relevant memories based on semantic similarity.
Configuring Decision Memory
To ensure the citizen remembers decisions, you must explicitly specify which information to retain. Ektro does not auto-memorize everything—it uses a recency and relevance threshold. Use the memory settings: priority_tags for decisions (e.g., 'critical','user_preference') and max_memory_size (default 1000 items). You can also set decay_factor to control forgetting. For example, a decay of 0.9 means memories fade slightly over time. Alternatively, pin key decisions using the /memory/pin endpoint. This is a tradeoff: you control what sticks, but requires deliberate coding.