feat(brain): hybrid BM25 + pgvector retrieval (opt-in)
All checks were successful
CI / Lint / Test / Vet (push) Successful in 15s
CI / Mirror to GitHub (push) Successful in 3s

Wires nomic-embed-text (iguana ollama) + pgvector on the shared
postgres18 into brain_query / brain_answer via Reciprocal Rank Fusion.
Pure BM25 stays the default; setting BRAIN_PG_DSN and BRAIN_EMBED_URL
together opts in. Setting one without the other is misconfiguration →
exit 1.

New packages:

- internal/embed
  Client.Embed(ctx, text) → []float32 via POST {URL}/api/embed.
  Defaults to nomic-embed-text:latest (768 dim). nil-on-empty-URL so
  callers gate on a single nil check.

- internal/vectorstore
  PGStore wraps a pgxpool against postgres18. Init creates
  brain_embeddings(path PK, vector(768), updated_at) + HNSW cosine
  index idempotently. Upsert / Delete / Search / KnownPaths.
  Sync(brainDir, store, embedder) diffs brain/wiki/ against the store
  and upserts new files / deletes removed ones; StartSync runs it on
  a ticker (default 300s). Integration tests gated by BRAIN_PG_TEST_DSN.

- scripts/brain-embeddings-init.sql
  One-time DBA setup: brain DB, brain_app role, vector extension,
  GRANTs. Idempotent.

Search layer:

- search.QueryOptions gains Vector + Embedder fields.
- QueryContext is the cancellable variant; Query stays for callers.
- When both are set, BM25 (top-N) and pgvector (top-4N) candidates
  merge via Reciprocal Rank Fusion (k=60, Cormack et al. 2009 — no
  tuning knob, robust to scale differences between rankers).
- Vector-only hits are hydrated from disk so callers see uniform
  Result records (path, title, excerpt, wing, hall, score).
- Wing/hall filters still apply to vector candidates via path-prefix.
- On embedder/vector errors the search falls back to BM25 — embedding
  outage degrades quality but doesn't take the brain offline.

MCP wiring:

- mcp.Server.WithHybridRetrieval(v, e) opt-in setter, same shape as
  WithReranker.
- brainQuery and brainAnswer pass the wired vector/embedder through
  to search.QueryContext.

REST:

- POST /backfill-embeddings drives Sync synchronously. Returns
  {added, deleted, errors[]}. 503 when feature is unconfigured.

cmd/server/main.go:

- BRAIN_PG_DSN + BRAIN_EMBED_URL together enable hybrid; one alone
  → exit 1.
- vectorAdapter bridges *PGStore (returns []Hit) to
  search.VectorSearcher (which takes []VectorHit) without either
  package importing the other.
- BRAIN_EMBED_SYNC_INTERVAL (default 300s) controls the background
  Sync ticker.

Backend pivot from Qdrant to pgvector recorded in DECISIONS.md
2026-05-18 (supersedes 2026-04-08): postgres18 already runs in
databases/ ns, Qdrant was never deployed, one engine beats two.

Dependency: github.com/jackc/pgx/v5 — modern, native pgvector via
parametric vector literals.

Tests:
- embed.Client: empty-URL nil, request shape, dimension, upstream
  error propagation, empty-text rejection.
- vectorstore.PGStore: dimension validation (unit); upsert/search/
  KnownPaths (integration, BRAIN_PG_TEST_DSN-gated).
- vectorstore.Sync: adds new files, skips known, deletes
  disappeared, skips _index.md, no-op when nil, collects embedder
  errors.
- search.Query: hybrid promotes vector-only hits via RRF; falls
  back to BM25 on embedder error.

Closes hyperguild#8.
This commit is contained in:
Mathias
2026-05-18 23:11:25 +02:00
parent a56a4db963
commit 57462b52ff
16 changed files with 1068 additions and 14 deletions

View File

@@ -0,0 +1,137 @@
package vectorstore_test
import (
"context"
"errors"
"os"
"path/filepath"
"testing"
"github.com/mathiasbq/hyperguild/ingestion/internal/vectorstore"
"github.com/stretchr/testify/assert"
"github.com/stretchr/testify/require"
)
type stubStore struct {
known map[string]struct{}
upserts map[string][]float32
deletes []string
failNext error
}
func (s *stubStore) KnownPaths(_ context.Context) (map[string]struct{}, error) {
out := make(map[string]struct{}, len(s.known))
for k := range s.known {
out[k] = struct{}{}
}
return out, nil
}
func (s *stubStore) Upsert(_ context.Context, path string, v []float32) error {
if s.failNext != nil {
err := s.failNext
s.failNext = nil
return err
}
if s.upserts == nil {
s.upserts = make(map[string][]float32)
}
s.upserts[path] = v
return nil
}
func (s *stubStore) Delete(_ context.Context, path string) error {
s.deletes = append(s.deletes, path)
return nil
}
type stubEmbedder struct {
vec []float32
err error
}
func (e stubEmbedder) Embed(_ context.Context, _ string) ([]float32, error) {
return e.vec, e.err
}
func writeNote(t *testing.T, dir, rel, body string) {
t.Helper()
full := filepath.Join(dir, rel)
require.NoError(t, os.MkdirAll(filepath.Dir(full), 0o755))
require.NoError(t, os.WriteFile(full, []byte(body), 0o644))
}
func TestSync_AddsNewFiles(t *testing.T) {
dir := t.TempDir()
writeNote(t, dir, "wiki/jepa-fx/facts/x.md", "body of x")
writeNote(t, dir, "wiki/jepa-fx/facts/y.md", "body of y")
store := &stubStore{known: map[string]struct{}{}}
emb := stubEmbedder{vec: make([]float32, 768)}
res, err := vectorstore.Sync(context.Background(), dir, store, emb)
require.NoError(t, err)
assert.Equal(t, 2, res.Added)
assert.Empty(t, res.Deleted)
assert.Contains(t, store.upserts, "wiki/jepa-fx/facts/x.md")
assert.Contains(t, store.upserts, "wiki/jepa-fx/facts/y.md")
}
func TestSync_SkipsAlreadyKnown(t *testing.T) {
dir := t.TempDir()
writeNote(t, dir, "wiki/a/facts/x.md", "x")
store := &stubStore{known: map[string]struct{}{"wiki/a/facts/x.md": {}}}
emb := stubEmbedder{vec: make([]float32, 768)}
res, err := vectorstore.Sync(context.Background(), dir, store, emb)
require.NoError(t, err)
assert.Equal(t, 0, res.Added)
assert.Empty(t, store.upserts)
}
func TestSync_DeletesDisappearedFiles(t *testing.T) {
dir := t.TempDir()
require.NoError(t, os.MkdirAll(filepath.Join(dir, "wiki"), 0o755))
// store has a path that doesn't exist on disk anymore
store := &stubStore{known: map[string]struct{}{"wiki/old/facts/ghost.md": {}}}
res, err := vectorstore.Sync(context.Background(), dir, &stubStoreWithDelete{stubStore: store}, stubEmbedder{vec: make([]float32, 768)})
require.NoError(t, err)
assert.Equal(t, 1, res.Deleted)
}
// stubStoreWithDelete is a thin wrapper to capture Delete calls;
// stubStore already implements Delete but we need the wrapper to mix
// store interfaces with sync-specific expectations.
type stubStoreWithDelete struct {
*stubStore
}
func TestSync_SkipsIndexFiles(t *testing.T) {
dir := t.TempDir()
writeNote(t, dir, "wiki/a/_index.md", "moc")
writeNote(t, dir, "wiki/a/facts/real.md", "body")
store := &stubStore{known: map[string]struct{}{}}
res, err := vectorstore.Sync(context.Background(), dir, store, stubEmbedder{vec: make([]float32, 768)})
require.NoError(t, err)
assert.Equal(t, 1, res.Added)
assert.NotContains(t, store.upserts, "wiki/a/_index.md")
}
func TestSync_NoOpWhenComponentsNil(t *testing.T) {
dir := t.TempDir()
writeNote(t, dir, "wiki/a/facts/x.md", "x")
res, err := vectorstore.Sync(context.Background(), dir, nil, nil)
require.NoError(t, err)
assert.Equal(t, 0, res.Added)
}
func TestSync_CollectsEmbedderErrors(t *testing.T) {
dir := t.TempDir()
writeNote(t, dir, "wiki/a/facts/x.md", "x")
store := &stubStore{known: map[string]struct{}{}}
emb := stubEmbedder{err: errors.New("upstream down")}
res, err := vectorstore.Sync(context.Background(), dir, store, emb)
require.NoError(t, err)
assert.Equal(t, 0, res.Added)
assert.Len(t, res.Errors, 1)
}