# v0.43.0 — Embedding API & Advanced RAG Patterns

> **Full technical details:** [v0.43.0.md-full.md](v0.43.0.md-full.md)

**Status: Planned** | **Scope: Large**

> Ergonomic `embedding_stream_table()` API for one-statement RAG corpus setup,
> materialised k-NN graph research for fixed-pivot retrieval, per-tenant ANN
> indexing patterns, outbox-emitted embedding events, and ecosystem positioning
> for the post-hardening AI stack.

---

## What is this?

v0.41.0-v0.42.0 build the production embedding primitives and hybrid-search
coverage. v0.43.0 consolidates that work into a higher-level ergonomic surface:
an API that sets up common RAG corpus patterns in one call, documentation for
multi-tenant ANN deployments, downstream embedding events, and a research spike
for materialised k-NN graphs.

---

## `embedding_stream_table()` ergonomic API

Today, users still write the full denormalization query by hand. v0.43.0 adds a
higher-level helper that can generate the query, create the stream table,
provision indexes, configure post-refresh actions, and return a dry-run preview
for expert users who want to inspect the generated SQL first.

---

## Advanced RAG patterns

This release also documents multi-tenant ANN patterns, explores materialised
k-NN graphs as a parallel research spike, and extends the outbox so embedding
changes can feed downstream agents, rerankers, or external services.

---

## Also in v0.43.0

- Public starter repository and examples for the post-hardening AI stack
- Best-effort ecosystem positioning with pgvector / pgai partners, but never as
  a release gate

---

## Scope

v0.43.0 is a large release because the ergonomic API hides real query-shape and
indexing complexity. The k-NN graph work remains research unless the trade-off
looks decisively favorable; the release should not block on speculative work.

---

*Previous: [v0.42.0 — Hybrid Search & Sparse Vector Aggregates](v0.42.0.md)*
*Next: [v1.0.0 — Stable Release](v1.0.0.md-full.md)*