Skip to content

🚀 MeridianDB v1.0.0 — Release Notes

Overview

MeridianDB is now feature-complete and ready to ship. This release finalizes the architecture for an AI-first database that redefines retrieval for agents — going beyond semantic search with temporal, contextual, and behavioral dimensions.

Our mission is to solve catastrophic forgetting in AI systems, helping agents strike the right balance between stability and plasticity.

✨ What’s New in v1.0.0

  • Integrated Consistency Model

  • Queue-based architecture ensures eventual consistency without developer orchestration.

  • Redundant storage (Vector + D1) preserves multidimensional context and reliability.

  • Multi-Dimensional Retrieval Engine

  • Semantic: Over-fetched, refined retrieval beyond standard RAG pipelines.

  • Temporal: Data decays over time, supports factual/irrelevant tagging, and frequency weighting.
  • Contextual: Filters results by task, environment, and developer-defined context.

  • Behavioral: Tracks retrieval impact on tasks for continuous feedback and improvement.

  • Cloudflare-Native Architecture

  • Global low-latency access.

  • Event-driven processing for scalable and cost-efficient operations.

  • Automatic retries, failover, and eventual consistency built-in.

  • Operational Simplicity

  • No need to glue multiple databases together — SDK + Operator UI included.

  • Simple developer API: store, retrieve, log.

🏗️ Architecture Highlights

  • Worker API gateway handles human and SDK requests.
  • Eventual consistency via write queues and workers for vectorization.
  • SQL-based feature scoring (no graph traversal) for performance and scalability.
  • Built-in behavioral logging creates feedback loops for task success tracking.

🛠️ Benefits

  1. Integrated Consistency & Reliability – No ghost embeddings, fewer race conditions.
  2. Simple Developer Experience – One API, no orchestration headaches.
  3. Native AI Feedback Loop – Behavioral insights for better retrieval over time.
  4. Cloudflare-Native Scalability – Low-latency global access by default.

⚠️ Known Limitations

  • Eventual Consistency: Reads may lag slightly after writes.
  • Feature Engineering: Developers must supply contextual features (future auto-context generation planned).
  • Storage Cost: Temporal decay requires periodic cleanup jobs.
  • Learning Curve: New retrieval model may require time to adopt.
  • Cloudflare Coupling: Architecture is optimized for Cloudflare ecosystem.

🗓️ Launch Date

Target shipping date: October 7, 2025.