Nicheloom

Market intelligence for builders — see what's gaining traction before it's crowded.

A memory database that forgets, consolidates, and detects contradiction

Details

External ID
47767119
Source
HN
Company
—
Product
database
Website domain
github.com
Launched
April 14, 2026
Cohort
—
Upvotes
48
Upvotes percentile
0.8354755784061697
Tags
—
Fetched at
Sept. 7, 2026, 9:26 p.m.
Updated at
Sept. 7, 2026, 9:26 p.m.

Description

Vector databases store memories. They don't manage them. After 10k memories, recall quality degrades because there's no consolidation, no forgetting, no conflict resolution. Your AI agent just gets noisier.YantrikDB is a cognitive memory engine — embed it, run it as a server, or connect via MCP. It thinks about what it stores: consolidation collapses duplicate memories, contradiction detection flags incompatible facts, temporal decay with configurable half-life lets unimportant memories fade like human memory does.Single Rust binary. HTTP + binary wire protocol. 2-voter + 1-witness HA cluster via Docker Compose or Kubernetes. Chaos-tested failover, runtime deadlock detection (parking_lot), per-tenant quotas, Prometheus metrics. Ran a 42-task hardening sprint last week — 1178 core tests, cargo-fuzz targets, CRDT property tests, 5 ops runbooks.Live on a 3-node Proxmox homelab cluster with multiple tenants. Alpha — primary user is me, looking for the second one.

Enrichment

Theme
low-level systems and developer tools
Vertical
Horizontal
Function
Data infrastructure
Audience
Developer
AI stance
Not AI
Project type
Commercial product
Normalized one-liner
memory database with forgetting
Manually corrected
False

Could you build this?

Partial A basic vector store is easy to wrap, but implementing automated memory consolidation, decay/forgetting curves, and contradiction detection requires non-trivial cognitive architecture and graph engineering.

What it would actually take: The engine requires a hybrid data store combining vector search (e.g., HNSW), graph structures (for entity relations), and temporal indexing. The complex components are the asynchronous background jobs that compute memory decay curves, cluster and merge related facts into synthesized long-term knowledge, and detect logical contradictions across new and existing assertions. Building a reliable implementation requires deep expertise in knowledge graphs, NLP logic systems, and database internals.

Discussion

20 comments analyzed.

Competitors mentioned: mem0 (similar memory system design), Vector stores (for document retrieval at scale), QMD and prose summarization (for agent context)

Concerns raised: Conflict detection fails on raw text without explicit entity extraction, Cosine similarity at 0.85 threshold can't differentiate between similar concepts (AWS vs Azure, love vs hate), Facts extracted from conversation lose critical context and nuance, Irony and sarcasm in conversation lead to incorrect fact extraction, Temporal contradictions not resolved (current vs previous CEO, temporal validity not tracked)

Feature requests: Auto-extraction of entities and relations to surface contradictions without explicit relate() calls, Time-of-validity schema to handle temporal supersession, not just decay, Typed relations support for modeling scope (subsidiaries, co-CEOs, sub-organizations), Context preservation in memory storage, not just atomic facts, Open source or self-hosted local options

Competitors

Other products that read as similar to this one — 58 launches clear the similarity bar, closest 8 shown.

Attention rank: #6 of 59 (itself plus its competitors, highest first — normalized so YC and Product Hunt are compared fairly).

Launched 164 days after the earliest competitor.

Other launches for this product

Same idea, different domain

Nobody's really built a data infrastructure tool for Media & entertainment yet.