mem1.wiki
Opinionated knowledge base about memory and RAG systems for AI agents.
Tools, methods, patterns, anti-patterns, and the topic graph that ties them together.
Smart search results
No relevant entries — try fewer keywords or turn Smart search off.
- Anytype Object-oriented, local-first, end-to-end encrypted PKM with P2P sync via the Any-Sync protocol. Architecturally the most ambitious tool in this list; the UX still trails Notion and Obsidian.
tools.pkm - Capacities Object-based PKM that splits the difference between Notion's databases and Roam's daily-notes. Cleanest object model in the consumer PKM space; cloud-only is the main reason to skip it.
tools.pkm - LightRAG Simple and fast GraphRAG implementation that combines entity-relationship graphs with vector retrieval. Lighter than Microsoft GraphRAG, pluggable vector backend.
tools.memory.graph_rag - Logseq Open-source, outliner-first, local-first PKM with daily-notes as the default capture surface. Closest free alternative to Roam Research; the DB rewrite shifts the bet to a SQLite future.
tools.pkm - Mem0 Memory layer for AI agents. Stores user/session facts, decays unused memory, and surfaces relevant context per turn. Drop-in for OpenAI-style chat loops.
tools.memory.agent_memory - Notion Block-and-database hybrid that became the default team wiki. Strong for structured project bases, weak as a personal thinking tool — too many features compete for attention.
tools.pkm - Obsidian Local-first markdown editor with bidirectional links and a deep plugin ecosystem. Best-in-class as a personal thinking surface; weakest for real-time team collaboration.
tools.pkm - Qdrant Open-source vector database written in Rust. Strong payload filtering, hybrid (sparse + dense) search, and a managed cloud option. Default-OK choice for most RAG stacks.
tools.memory.vector_dbs - Roam Research The tool that mainstreamed bidirectional links and daily notes. Historically transformative, currently coasting — most of its ideas now ship in Logseq, Obsidian, and Capacities for less money and more transparency.
tools.pkm
No matches in this section.
- GraphRAG Retrieval method that builds an entity-relation graph from the corpus, summarises graph communities, and uses those summaries plus graph traversal alongside vector search.
methods.rag - HyDE — Hypothetical Document Embeddings Query expansion technique. Instead of embedding the query directly, ask the LLM to draft a hypothetical answer, then embed that. Often improves recall on short, abstract queries.
methods.retrieval - PARA Tiago Forte's organising scheme for personal knowledge — Projects, Areas, Resources, Archive. Pragmatic file-cabinet shape for execution-oriented PKM. The right scaffolding for Notion-style team workspaces; the wrong shape for thinking.
methods.pkm - Zettelkasten Luhmann's slip-box method — atomic notes, dense linking, no rigid hierarchy. Foundational influence on every modern bidirectional-link PKM tool. Easy to romanticise, hard to actually run.
methods.pkm
No matches in this section.
- Atomic notes One idea per note, each note independently understandable, linked explicitly. The structural primitive behind Zettelkasten, Evergreen notes, and most modern PKM tools.
patterns.memory_management - Hierarchical Navigable Small World (HNSW) Multi-layer graph index for approximate nearest neighbour search. De-facto default for vector databases when latency matters more than perfect recall.
patterns.indexing - Cross-encoder reranker Two-stage retrieval. First, ANN search returns top-K candidates. Then a cross-encoder scores each (query, candidate) pair jointly and re-orders by relevance.
patterns.ranking
No matches in this section.
- Char-count chunking on prose Splitting documents into fixed N-character windows for retrieval. Cheap, ubiquitous, and quietly destroys retrieval quality on real prose.
anti_patterns.chunking - The read-only wiki nobody updates The Confluence / Notion / wiki space that everyone references and nobody edits. Drifts from reality, nobody trusts it, but it stays up because removing it feels worse than ignoring it. The default end state of most team knowledge bases.
anti_patterns.governance - Single-retriever pipeline Trusting one retriever (usually dense vector) to handle every query shape. Works on demos, fails on long-tail queries — keyword-heavy, code, IDs, exact phrases.
anti_patterns.retrieval
No matches in this section.
- Retrieval Methods Named retrieval methods — query reformulation, hypothetical answer expansion,
- GraphRAG Tools Graph-augmented retrieval implementations. These tools combine an entity-relation graph