TazLab Knowledge Layers
Scope
This page defines the roles and boundaries of the main TazLab knowledge layers so agents know where to look and where to write.
Current Synthesis
TazLab currently uses three distinct but complementary knowledge layers:
AGENTS.ctx/memory/as the active present-state and recent-chronology baselinemnemosyneas the semantic retrieval layer for stored memorieswiki.tazlab.netas the durable, browsable, interlinked documentation and synthesis layer
These layers should cooperate, not collapse into one another.
Layer Roles
AGENTS.ctx/memory/
- best for current verified state, recent chronology, and active technical debt
- should answer what is true now and what changed recently
- remains the baseline when present-state truth matters
mnemosyne
- best for semantic lookup across past memories
- useful for recovering rationale, prior discoveries, and related history quickly
- retrieved memories are informative, but they are not automatically the current source of truth
wiki.tazlab.net
- best for durable documentation, cross-project explanations, topic synthesis, and organized navigation
- intended to be readable by both humans and agents without relying on chat history
- should accumulate stable pages such as system maps, repository maps, runbooks, concept pages, and source summaries
Practical Boundary Rule
If the question is about current operational truth, check memory and live repositories first.
If the question is about semantic recall, prior rationale, or related past work, use mnemosyne as a retrieval aid.
If the task benefits from stable documentation, cross-linking, and durable synthesis, update or consult wiki.tazlab.net.
Supporting Pages
Open Questions
- Which classes of memory updates should later be mirrored into the wiki automatically?
- Which wiki pages should eventually be sourced from repeatable generation or lint workflows?