Arbitrary import pipeline support with full metadata preservation
LOGOS must accept imports from external sources with full
metadata preservation. The import pipeline is a first-class
architectural concern, not a utility feature bolted on after
the fact.
Why It's Needed
A significant portion of the knowledge and reasoning that
feeds LOGOS originates outside LOGOS — in AI conversations,
in external tools, in other systems. If that content cannot
be imported with full fidelity, the knowledge it represents
is either lost or lives outside LOGOS where it cannot be
cross-referenced, searched, or fed to CORTEX.
The Claude chat import pipeline is the first concrete
instance of this need. Conversations happen in Claude,
reasoning happens there, decisions get made there. That
content needs to live in LOGOS where it is referenceable
and traceable — not just archived as a flat file.
Beyond Claude, the pattern applies to any external source:
other AI systems, issue trackers, email threads, documents,
or any future tool that produces knowledge relevant to
LOGOS.
The Requirement
LOGOS supports importing content from external sources
through a defined pipeline architecture. An import:
- Preserves provenance — source system, source ID,
original timestamps, and author information are
retained as metadata, not discarded - Preserves structure — threaded conversations
remain threaded, document sections remain sections,
the original organization is not flattened - Preserves metadata — all available metadata from
the source is carried through — model version,
temperature, tool use, thinking blocks, project
membership, or any other source-specific fields - Is idempotent — importing the same content twice
does not create duplicates; re-importing an updated
version updates the existing content as a new event - Is auditable — every import is itself an event
in the event log, with the import source, timestamp,
and the identity of who triggered it
Extensibility
The pipeline architecture supports new source types
without requiring platform changes. Adding a new import
source means implementing a defined interface, not
modifying core LOGOS. The Claude importer is the first
implementation of that interface, not a special case
hardcoded into the platform.
Acceptance
A Claude conversation export is imported into LOGOS.
The resulting topic preserves the original conversation
structure, timestamps, and all available metadata.
Re-importing the same export produces no duplicates.
Importing an updated export updates the existing content
as a new revision event. The import itself appears in
the event log. A second import source — a different
format entirely — can be added by implementing the
import interface without modifying core LOGOS.
Sources
- LOGOS → Chat Import Pipeline → Claude export does not include model version (concrete metadata preservation gap that import pipeline must handle)
- LOGOS → Chat Import Pipeline → Claude export does not include project membership (concrete structure preservation gap that import pipeline must handle)
- LOGOS → Requirements → Arbitrary post metadata (imported content needs somewhere to store source metadata)
- LOGOS → Requirements → Event-sourced storage as the foundation (imports are events — idempotency and auditability depend on this)
- LOGOS → Requirements → Full data export — no lock-in (import and export are two ends of the same pipeline architecture)