820
total weekly downloads
10 packages
melete-ai
Melete — the Self-Driving Discovery Brain. A closed-loop active-experiment engine with a pluggable oracle and a signed, offline-verifiable discovery trace. Mneme remembers; Melete discovers.
mneme-ai
Mneme — the memory layer for your codebase. Knows the WHY, the WHAT, the WHERE-IT-BREAKS.
@mneme-ai/core
Core indexing, retrieval, and graph engine for Mneme
@mneme-ai/correlator
Phase 3 — error/incident correlation engine. The moat: connect git + code + errors + time.
@mneme-ai/embeddings
Embedding providers (OpenAI, Ollama, bundled-WASM, hash-fallback) for Mneme
@mneme-ai/gephyra
GEPHYRA — the living bridge / Toll Booth of Truth for AI agents. Real-time semantic truth-customs (verify · quarantine injection · honesty toll · conscience nudge · signed crossing) on top of any protocol. The deployable surface of Mneme.
@mneme-ai/matrix
THE MATRIX RAIL — the local-first gRPC wire server for Mneme. Any AI agent (any language) connects over HTTP/2 + Protobuf to reach every Mneme function through one typed, signed, streaming pipe. Invoke (unary, proof-carrying) + Pipe (bidi, chunked → any-s
@mneme-ai/mcp
MCP server that exposes Mneme to Claude Code, Cursor, Continue, and other AI clients
@mneme-ai/sdk
World-class premium SDK for Mneme — in-process native API for AI vendor embedding. 30-80× faster than CLI subprocess.
@mneme-ai/xray
Mneme Repo X-Ray — a signed, raw-free, deterministic X-Ray of any repo. Every number is reproducible from git/AST/metadata and sealed with an offline-verifiable NOTARY receipt. No source code ever leaves the machine; no LLM guesses anything.