Files

T

Igor Lins e Silva f20a1a30fe docs(website): align mempalaceofficial.com with honest benchmarks

Part of #875. Bring the VitePress site into line with the new README
and the reproducibility scorecard: drop category-error comparisons,
drop retracted claims, retain only metrics and caveats that survive
audit.

website/index.md
 - New tagline matches README (local-first, verbatim, pluggable backend,
   96.6% R@5 raw, zero API calls).
 - Replace the "MemPalace hybrid 100% / Supermemory ~99% / Mastra
   94.87% / Mem0 ~85%" comparison table with a single honest table
   showing MemPalace's own retrieval-recall numbers (raw 96.6%,
   hybrid v4 held-out 98.4%). Add an explicit sentence explaining why
   we no longer publish a cross-system table on the landing page
   (retrieval recall vs QA accuracy are different metrics).
 - Soften the "ChromaDB-powered vector search" feature blurb to be
   backend-agnostic, since the retrieval layer is pluggable.

website/reference/benchmarks.md
 - Full rewrite of the retrieval-recall tables. No more "100%"
   headline; honest held-out 98.4% R@5 replaces it. Added the
   model-agnostic rerank result (99.2% R@5 / 100% R@10 with
   minimax-m2.7 via Ollama) to show the pipeline is not Haiku-specific.
 - Drop the LoCoMo "Hybrid v5 + Sonnet rerank (top-50) 100%" row.
   With per-conversation session counts of 19-32 and top_k=50, the
   retrieval stage returns every session by construction — the number
   measures an LLM's reading comprehension, not retrieval.
 - Drop the cross-system comparison tables. Link out to each project's
   own research page (Mastra, Mem0, Supermemory) for their published
   numbers and metric definitions.
 - Rewrite reproduction commands to use the correct repository and
   demonstrate the new --llm-backend ollama flag.

website/concepts/the-palace.md
 - Remove the "+34%" row / paragraph. Wing/room filtering is standard
   metadata filtering in the vector store, not a novel retrieval
   mechanism — the April-7 note already retracted that framing; this
   finishes the retraction on the website where it had remained.

website/guide/searching.md
 - Same treatment for "34% retrieval improvement". Reframe as
   operational scoping, not a novel boost.

website/reference/contributing.md
 - Update the "palace structure matters" bullet to reflect the same
   framing: scoping-not-magic.

website/concepts/knowledge-graph.md
 - Replace the MemPalace-vs-Zep feature matrix with a short "related
   work" note that links to Zep's own documentation for authoritative
   details on their deployment model. Avoids claims we cannot verify
   at source.

2026-04-14 21:37:45 -03:00

4.1 KiB

Raw Blame History

The Palace

Ancient Greek orators memorized entire speeches by placing ideas in rooms of an imaginary building. Walk through the building, find the idea. MemPalace applies the same principle to AI memory.

Structure

Your conversations are organized into a navigable hierarchy:

graph LR
    classDef wingPerson fill:#1e1b4b,stroke:#4f46e5,color:#e0e7ff,stroke-width:2px,rx:8px,ry:8px;
    classDef wingProject fill:#164e63,stroke:#06b6d4,color:#cffafe,stroke-width:2px,rx:8px,ry:8px;
    classDef room fill:#312e81,stroke:#6366f1,color:#e0e7ff,stroke-width:1px,rx:4px,ry:4px;
    classDef closet fill:#3b0764,stroke:#8b5cf6,color:#f3e8ff,stroke-width:1px,rx:4px,ry:4px;
    classDef drawer fill:#0f766e,stroke:#14b8a6,color:#ccfbf1,stroke-width:1px,rx:4px,ry:4px;
    classDef tunnel_link stroke:#8b5cf6,stroke-width:2px,stroke-dasharray: 5 5;

    subgraph W1 [WING: Person]
        direction TB
        RA["Room A"]
        RB["Room B"]
        CA["Closet"]
        DA["Drawer (verbatim)"]
        
        RA -- "hall" --> RB
        RA --> CA --> DA
    end

    subgraph W2 [WING: Project]
        direction TB
        RA2["Room A"]
        RC["Room C"]
        CA2["Closet"]
        DA2["Drawer (verbatim)"]
        
        RA2 -- "hall" --> RC
        RA2 --> CA2 --> DA2
    end

    RA <==> |tunnel bridge| RA2

    class W1 wingPerson;
    class W2 wingProject;
    class RA,RB,RA2,RC room;
    class CA,CA2 closet;
    class DA,DA2 drawer;

Components

Wings

A person or project. As many as you need.

Every project, person, or topic gets its own wing in the palace. Wings are the top-level organizational unit.

Rooms

Specific topics within a wing. Examples: auth-migration, graphql-switch, ci-pipeline.

Rooms are named ideas. They're auto-detected from your folder structure during mempalace init, and you can create additional rooms manually.

Halls

Halls are the conceptual categories that describe how related memories connect within a wing:

hall_facts — decisions made, choices locked in
hall_events — sessions, milestones, debugging
hall_discoveries — breakthroughs, new insights
hall_preferences — habits, likes, opinions
hall_advice — recommendations and solutions

Tunnels

Connections between wings. When the same room appears in different wings, the graph layer can treat that as a cross-wing connection.

wing_kai       / hall_events / auth-migration  → "Kai debugged the OAuth token refresh"
wing_driftwood / hall_facts  / auth-migration  → "team decided to migrate auth to Clerk"
wing_priya     / hall_advice / auth-migration  → "Priya approved Clerk over Auth0"

Same room. Three wings. The graph can use that shared room name as a bridge.

Closets

Closets are the summary layer in the broader MemPalace vocabulary: compact notes that point back to the original content. In the current implementation, the main persisted storage path is still the underlying drawer text plus metadata.

Drawers

The original stored text chunks. This is the primary retrieval layer used by the current search and benchmark flows.

Why Structure Matters

Wing and room identifiers become metadata filters at query time. Narrowing a search to a specific wing (or wing + room) means the vector store only scores candidates inside that scope, which is useful when you have many unrelated projects or people filed in the same palace.

This is standard metadata filtering in the underlying vector store, not a novel retrieval mechanism. The useful property here is operational — clear scoping rules that a human or an agent can apply predictably — not a magic retrieval boost.

The palace supports graph traversal across wings:

MCP tool: mempalace_traverse
  arguments: { "start_room": "auth-migration" }
  → discovers rooms in wing_kai, wing_driftwood, wing_priya

MCP tool: mempalace_find_tunnels
  arguments: { "wing_a": "wing_code", "wing_b": "wing_team" }
  → auth-migration, deploy-process, ci-pipeline

This is the navigation story: shared room structure gives the model more than one way to reach relevant context.

4.1 KiB Raw Blame History