cinderella
Dusting off the listings…
Dusting off the listings…
On-device, bi-temporal memory graph for AI chat apps — entity/fact extraction, hybrid retrieval, persona identity.
Swift
53 files · ~4,186 LOC · 25 tests
Scrub: clean · 0 removed
Every listing gets the same checks, run by machine. We publish the ones it failed as well as the ones it passed.
What you get: Full ownership — You get sole ownership and all IP rights, per the Seller Agreement.
✅ Runs as-is — Plug-and-play: builds and runs with your own fresh keys/config. Nothing missing.
An embeddable, on-device long-term memory engine for AI chat apps on Apple platforms. AiPersona is a from-scratch Swift Package reimplementation of synapse-cortex's cognitive core — a temporal knowledge graph, hybrid (BM25 + embedding) retrieval, and LLM-driven fact extraction/correction — reshaped for a single-user, embedded, multi-provider target instead of a multi-tenant server. It gives a chat app the ability to say "I remember" and mean it: facts mentioned in past conversations are extracted, stored in a small local knowledge graph, and surfaced back into future system prompts — without a backend, without an account system, and without sending the user's data anywhere unless the host app opts into an external LLM/embedding provider.
A Swift Package that gives Apple-platform chat apps a real long-term memory layer: LLM-extracted facts stored in a bi-temporal knowledge graph on SwiftData, retrieved with hybrid BM25 + embedding search, and rendered into prompt fragments the host app injects into its own chat loop. It's a headless library, not an app or a server.
The design decisions here are the good kind of opinionated — facts are invalidated rather than deleted so history stays reconstructable, the store lives in its own ModelContainer to dodge host schema collisions, embeddings default to on-device NLEmbedding, and there's a test directory mirroring every source module. At $725 that's a lot of thoughtful architecture for the price of a weekend. Two things to check before you buy: it depends on two external packages (AIChatKit, AIChatKitMLX) resolved from sibling checkouts or GitHub, so confirm those are reachable, licensed how you need, and that the package builds clean from a fresh clone; and readiness at 57 means you should assume the retrieval quality and extraction prompts need real tuning against your own conversations rather than working perfectly out of the box. Buy it if you're shipping an Apple-native AI companion or journaling app a
Rough slice of the AI memory/context-layer tooling market (Mem0, Zep, Letta and friends) narrowed to the on-device Apple-ecosystem segment, sized from indie and mid-size iOS/macOS AI app developers who need memory without a backend.
Cindy's opinion, generated from this listing — including the market size, which is a rough estimate. Not financial advice; do your own diligence.
Cinderella's listing assistant. Ask about ownership, fees, legitimacy, market fit… She can be wrong, so do your own diligence.