Closes SPEC §3.4. The matrix indexer is now a learning meta-index per
feedback_meta_index_vision.md — every successful (query → answer)
pair recorded via /matrix/playbooks/record boosts that answer for
future similar queries.
This is the architectural piece that lifts vectord from "static
hybrid search" to the meta-index J originally framed in Phase 19 of
the Rust system.
What's new:
- internal/matrix/playbook.go — PlaybookEntry, PlaybookHit,
ApplyPlaybookBoost. Pure-function boost math:
distance' = distance * (1 - 0.5 * score)
Score 0 = no boost (factor 1.0); score 1 = halve distance
(factor 0.5). Capped at 0.5 deliberately so a single high-
confidence playbook can't dominate the base ranking forever
(runaway-feedback-loop guard).
- Retriever.Record(entry, corpus) — embeds query_text, ensures
playbook corpus exists (idempotent), upserts via deterministic
sha256-derived ID (last score wins on re-record of same triple).
- Retriever.Search extended with UsePlaybook + PlaybookCorpus +
PlaybookTopK + PlaybookMaxDistance. Reuses the query vector —
no extra embed call. Missing-corpus 404 = no-op (cold-start
state before any Record call), not an error.
- POST /v1/matrix/playbooks/record (matrixd) — caller submits
{query_text, answer_id, answer_corpus, score, tags?}; gets
{playbook_id} back.
Storage: a vectord index named "playbook_memory" (configurable per
request) with embed(query_text) as the vector and the
PlaybookEntry JSON as metadata. Just another corpus — observable
from /vectors/index, persistable through G1P, etc.
Match key for boost: (AnswerID, AnswerCorpus). Cross-corpus ID
collisions don't false-match — verified by
TestApplyPlaybookBoost_CorpusAttributionRespected.
End-to-end smoke (scripts/playbook_smoke.sh, all assertions PASS):
- Baseline search: widget-c at distance 0.6566 (rank 3)
- Record playbook: query → widget-c, score=1.0
- Re-search with use_playbook=true:
widget-c distance: 0.3283 (rank 2)
ratio: 0.5 EXACTLY (matches boost math precisely)
playbook_boosted: 1
- widget-c jumped from #3 to #2 — learning loop visible
Tests:
- 8 unit tests in internal/matrix/playbook_test.go covering
Validate, BoostFactor (5 cases), the no-boost identity, the
boost-moves-result-up scenario, highest-score wins on duplicate
matches, cross-corpus attribution, JSON round-trip, and
rejection of empty metadata
- scripts/playbook_smoke.sh integration test (3 assertions PASS)
15-smoke regression sweep all green (D1-D6, G1, G1P, G2,
storaged_cap, pathway, matrix, relevance, downgrade, playbook).
SPEC §3.4 NOW COMPLETE: 5 of 5 components shipped. The matrix
indexer's port is done as a substrate; remaining work is operational
(rating signal sources, telemetry, eventual structured filtering for
staffing data — none in §3.4).
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
181 lines
5.4 KiB
Go
181 lines
5.4 KiB
Go
package matrix
|
|
|
|
import (
|
|
"encoding/json"
|
|
"testing"
|
|
)
|
|
|
|
func TestPlaybookEntry_Validate(t *testing.T) {
|
|
good := PlaybookEntry{
|
|
QueryText: "x", AnswerID: "y", AnswerCorpus: "z", Score: 0.5,
|
|
}
|
|
if err := good.Validate(); err != nil {
|
|
t.Errorf("good entry should validate: %v", err)
|
|
}
|
|
|
|
cases := []struct {
|
|
name string
|
|
entry PlaybookEntry
|
|
}{
|
|
{"empty query", PlaybookEntry{AnswerID: "y", AnswerCorpus: "z", Score: 0.5}},
|
|
{"empty answer id", PlaybookEntry{QueryText: "x", AnswerCorpus: "z", Score: 0.5}},
|
|
{"empty corpus", PlaybookEntry{QueryText: "x", AnswerID: "y", Score: 0.5}},
|
|
{"score too high", PlaybookEntry{QueryText: "x", AnswerID: "y", AnswerCorpus: "z", Score: 1.5}},
|
|
{"score negative", PlaybookEntry{QueryText: "x", AnswerID: "y", AnswerCorpus: "z", Score: -0.1}},
|
|
}
|
|
for _, c := range cases {
|
|
if err := c.entry.Validate(); err == nil {
|
|
t.Errorf("%s: expected validation error, got nil", c.name)
|
|
}
|
|
}
|
|
}
|
|
|
|
func TestPlaybookEntry_BoostFactor(t *testing.T) {
|
|
cases := []struct {
|
|
score float64
|
|
want float64
|
|
}{
|
|
{0.0, 1.0},
|
|
{0.5, 0.75},
|
|
{1.0, 0.5},
|
|
{-0.1, 1.0}, // clamped
|
|
{1.5, 0.5}, // clamped
|
|
}
|
|
for _, c := range cases {
|
|
got := PlaybookEntry{Score: c.score}.BoostFactor()
|
|
if abs(got-c.want) > 1e-9 {
|
|
t.Errorf("BoostFactor(score=%.2f): want %.4f, got %.4f", c.score, c.want, got)
|
|
}
|
|
}
|
|
}
|
|
|
|
func TestApplyPlaybookBoost_NoHitsLeaveResultsAlone(t *testing.T) {
|
|
results := []Result{
|
|
{ID: "a", Distance: 0.1, Corpus: "x"},
|
|
{ID: "b", Distance: 0.2, Corpus: "x"},
|
|
}
|
|
n := ApplyPlaybookBoost(results, nil)
|
|
if n != 0 {
|
|
t.Errorf("expected 0 boosted, got %d", n)
|
|
}
|
|
if results[0].ID != "a" || results[1].ID != "b" {
|
|
t.Errorf("results reordered without hits: %v", results)
|
|
}
|
|
}
|
|
|
|
func TestApplyPlaybookBoost_BoostMovesResultUp(t *testing.T) {
|
|
// Initial: a (0.10) beats b (0.20) beats c (0.30).
|
|
// Playbook says (answer=c, score=1.0) should be boosted → c's
|
|
// distance becomes 0.30 * 0.5 = 0.15. New ordering: a, c, b.
|
|
results := []Result{
|
|
{ID: "a", Distance: 0.10, Corpus: "x"},
|
|
{ID: "b", Distance: 0.20, Corpus: "x"},
|
|
{ID: "c", Distance: 0.30, Corpus: "x"},
|
|
}
|
|
hits := []PlaybookHit{
|
|
{PlaybookID: "p1", Distance: 0.05, Entry: PlaybookEntry{
|
|
AnswerID: "c", AnswerCorpus: "x", Score: 1.0,
|
|
}},
|
|
}
|
|
n := ApplyPlaybookBoost(results, hits)
|
|
if n != 1 {
|
|
t.Errorf("expected 1 boosted, got %d", n)
|
|
}
|
|
if results[0].ID != "a" || results[1].ID != "c" || results[2].ID != "b" {
|
|
t.Errorf("expected order a,c,b after boost; got %v", idsOf(results))
|
|
}
|
|
if abs(float64(results[1].Distance)-0.15) > 1e-6 {
|
|
t.Errorf("expected c distance 0.15 after boost; got %.4f", results[1].Distance)
|
|
}
|
|
}
|
|
|
|
func TestApplyPlaybookBoost_HighestScoreWinsForSameAnswer(t *testing.T) {
|
|
results := []Result{
|
|
{ID: "a", Distance: 0.30, Corpus: "x"},
|
|
}
|
|
// Two playbook hits both pointing at "a". Score=0.4 (weak boost)
|
|
// + Score=0.9 (strong boost). Strong should win — distance gets
|
|
// multiplied by 1-0.5*0.9 = 0.55, not by 1-0.5*0.4 = 0.80.
|
|
hits := []PlaybookHit{
|
|
{PlaybookID: "p_weak", Distance: 0.05, Entry: PlaybookEntry{
|
|
AnswerID: "a", AnswerCorpus: "x", Score: 0.4,
|
|
}},
|
|
{PlaybookID: "p_strong", Distance: 0.05, Entry: PlaybookEntry{
|
|
AnswerID: "a", AnswerCorpus: "x", Score: 0.9,
|
|
}},
|
|
}
|
|
ApplyPlaybookBoost(results, hits)
|
|
wantDist := 0.30 * 0.55
|
|
if abs(float64(results[0].Distance)-wantDist) > 1e-6 {
|
|
t.Errorf("strong-score boost should win: want %.4f, got %.4f", wantDist, results[0].Distance)
|
|
}
|
|
}
|
|
|
|
func TestApplyPlaybookBoost_CorpusAttributionRespected(t *testing.T) {
|
|
// Playbook references answer_id="a" in corpus="x".
|
|
// Results have answer_id="a" in corpus="y" — DIFFERENT corpus.
|
|
// Boost should NOT apply; the (id, corpus) tuple is the join key,
|
|
// not just id (otherwise different-corpus collisions would create
|
|
// false positives).
|
|
results := []Result{
|
|
{ID: "a", Distance: 0.30, Corpus: "y"},
|
|
}
|
|
hits := []PlaybookHit{
|
|
{PlaybookID: "p1", Distance: 0.05, Entry: PlaybookEntry{
|
|
AnswerID: "a", AnswerCorpus: "x", Score: 1.0,
|
|
}},
|
|
}
|
|
n := ApplyPlaybookBoost(results, hits)
|
|
if n != 0 {
|
|
t.Errorf("cross-corpus collision should not boost: got %d", n)
|
|
}
|
|
if abs(float64(results[0].Distance)-0.30) > 1e-6 {
|
|
// 1e-6 tolerance accounts for float32→float64 conversion;
|
|
// the assertion that matters is "unchanged from input."
|
|
t.Errorf("distance should be unchanged: got %.6f", results[0].Distance)
|
|
}
|
|
}
|
|
|
|
func TestPlaybookEntry_RoundTripJSON(t *testing.T) {
|
|
e := NewPlaybookEntry("forklift query", "w-12345", "workers", 0.85, []string{"chicago", "verified"})
|
|
raw, err := e.MarshalMetadata()
|
|
if err != nil {
|
|
t.Fatalf("marshal: %v", err)
|
|
}
|
|
got, err := UnmarshalPlaybookMetadata(raw)
|
|
if err != nil {
|
|
t.Fatalf("unmarshal: %v", err)
|
|
}
|
|
if got.QueryText != e.QueryText || got.AnswerID != e.AnswerID ||
|
|
got.AnswerCorpus != e.AnswerCorpus || got.Score != e.Score {
|
|
t.Errorf("round-trip mismatch: want %+v, got %+v", e, got)
|
|
}
|
|
if len(got.Tags) != 2 || got.Tags[0] != "chicago" {
|
|
t.Errorf("tags lost in round-trip: %v", got.Tags)
|
|
}
|
|
if got.RecordedAtNs == 0 {
|
|
t.Error("RecordedAtNs not set by NewPlaybookEntry")
|
|
}
|
|
}
|
|
|
|
func TestUnmarshalPlaybookMetadata_RejectsEmpty(t *testing.T) {
|
|
if _, err := UnmarshalPlaybookMetadata(json.RawMessage{}); err == nil {
|
|
t.Error("empty metadata should error")
|
|
}
|
|
}
|
|
|
|
func abs(f float64) float64 {
|
|
if f < 0 {
|
|
return -f
|
|
}
|
|
return f
|
|
}
|
|
|
|
func idsOf(rs []Result) []string {
|
|
out := make([]string, len(rs))
|
|
for i, r := range rs {
|
|
out[i] = r.ID
|
|
}
|
|
return out
|
|
}
|