28 lines
1015 B
Markdown
28 lines
1015 B
Markdown
# Phase 1 – Generic Persistence Backbone
|
||
|
||
Phase 1 introduces the additive `DOC` schema and the first concrete persistence layer for the
|
||
new generalized platform model.
|
||
|
||
## What is implemented
|
||
- `DOC.doc_tenant`
|
||
- `DOC.doc_document`
|
||
- `DOC.doc_source`
|
||
- `DOC.doc_content`
|
||
- `DOC.doc_text_representation`
|
||
- `DOC.doc_embedding_model`
|
||
- `DOC.doc_embedding`
|
||
- `DOC.doc_relation`
|
||
|
||
## Intent
|
||
The generic model now exists as real JPA entities, repositories, Flyway migration, and thin
|
||
transactional services. Existing TED runtime behavior is intentionally unchanged.
|
||
|
||
## Important limitation
|
||
The actual TED processing pipeline still writes only to the legacy TED-specific model. Dual-write
|
||
and migration come in later phases.
|
||
|
||
## Vector storage note
|
||
`doc_embedding` already separates vectorization lifecycle from the document root. The transient
|
||
`embeddingVector` field is intentionally not wired into Hibernate yet. Writing native pgvector data
|
||
and moving the vectorization pipeline to the new table is part of Phase 2.
|