consciousness/schema/memory.capnp
ProofOfConcept 23fac4e5fe poc-memory v0.4.0: graph-structured memory with consolidation pipeline
Rust core:
- Cap'n Proto append-only storage (nodes + relations)
- Graph algorithms: clustering coefficient, community detection,
  schema fit, small-world metrics, interference detection
- BM25 text similarity with Porter stemming
- Spaced repetition replay queue
- Commands: search, init, health, status, graph, categorize,
  link-add, link-impact, decay, consolidate-session, etc.

Python scripts:
- Episodic digest pipeline: daily/weekly/monthly-digest.py
- retroactive-digest.py for backfilling
- consolidation-agents.py: 3 parallel Sonnet agents
- apply-consolidation.py: structured action extraction + apply
- digest-link-parser.py: extract ~400 explicit links from digests
- content-promotion-agent.py: promote episodic obs to semantic files
- bulk-categorize.py: categorize all nodes via single Sonnet call
- consolidation-loop.py: multi-round automated consolidation

Co-Authored-By: Kent Overstreet <kent.overstreet@linux.dev>
2026-02-28 22:17:00 -05:00

86 lines
2.5 KiB
Cap'n Proto

@0xb78d9e3a1c4f6e2d;
# poc-memory: append-only memory store with graph structure
#
# Two append-only logs (nodes + relations) are the source of truth.
# A derived KV cache merges both, keeping latest version per UUID.
# Update = append new version with same UUID + incremented version.
# Delete = append with deleted=true. GC compacts monthly.
struct ContentNode {
uuid @0 :Data; # 16 bytes, random
version @1 :UInt32; # monotonic per UUID, latest wins
timestamp @2 :Float64; # unix epoch
nodeType @3 :NodeType;
provenance @4 :Provenance;
key @5 :Text; # "identity.md#boundaries" human-readable
content @6 :Text; # markdown blob
weight @7 :Float32;
category @8 :Category;
emotion @9 :Float32; # max intensity from tags, 0-10
deleted @10 :Bool; # soft delete
sourceRef @11 :Text; # link to raw experience: "transcript:SESSION_ID:BYTE_OFFSET"
# Migrated metadata from old system
created @12 :Text; # YYYY-MM-DD from old system
retrievals @13 :UInt32;
uses @14 :UInt32;
wrongs @15 :UInt32;
stateTag @16 :Text; # cognitive state (warm/open, bright/alert, etc.)
# Spaced repetition
lastReplayed @17 :Float64; # unix epoch
spacedRepetitionInterval @18 :UInt32; # days: 1, 3, 7, 14, 30
}
enum NodeType {
episodicSession @0;
episodicDaily @1;
episodicWeekly @2;
semantic @3;
}
enum Provenance {
manual @0;
journal @1;
agent @2;
dream @3;
derived @4;
}
enum Category {
general @0;
core @1;
technical @2;
observation @3;
task @4;
}
struct Relation {
uuid @0 :Data; # 16 bytes, random
version @1 :UInt32;
timestamp @2 :Float64; # unix epoch
source @3 :Data; # content node UUID
target @4 :Data; # content node UUID
relType @5 :RelationType;
strength @6 :Float32; # manual=1.0, auto=0.1-0.7
provenance @7 :Provenance;
deleted @8 :Bool; # soft delete
sourceKey @9 :Text; # human-readable source key (for debugging)
targetKey @10 :Text; # human-readable target key (for debugging)
}
enum RelationType {
link @0; # bidirectional association (from links= or md links)
causal @1; # directed: source caused target
auto @2; # auto-discovered
}
# Wrapper for streaming multiple messages in one file
struct NodeLog {
nodes @0 :List(ContentNode);
}
struct RelationLog {
relations @0 :List(Relation);
}