Aletheia — memory & knowledge platform for AI agents
PostgreSQL-first memory and knowledge platform for AI coding agents and human engineers. A 24/7 .NET operator process observes and reacts to events from connected systems (git, Jira, Slack, Discord, logs) and distills them into observations and verified claims in a knowledge graph.
Hybrid retrieval fuses four lanes — exact-id lookup, full-text search over observations and claims, and pgvector cosine kNN — with weighted reciprocal-rank fusion. Every retrieval writes a served-result manifest, so "what did this agent actually see" is replayable from the database. Agents consume it through MCP tools (search, save, workspace context packs); humans through an ops console.
Project information
- Technology used: .NET (C#), PostgreSQL, pgvector, Model Context Protocol (MCP), OpenRouter embeddings, GitLab CI/CD
- Project goal: Give AI agents persistent, auditable memory
- Date: 2026