feat(ingest): add STORE_MIN_SEVERITY second-tier severity gate

CLAUDE.md

@@ -217,6 +217,7 @@ Key settings in `internal/config/config.go`:
 - `LOG_FTS_ENABLED` (false) — when truthy (`true`/`yes`/`on`/`1`), provisions the SQLite FTS5 `logs_fts` virtual table + sync triggers at startup; when false, log-search uses vectordb (semantic) plus a 24h-clamped LIKE fallback. Toggle off and reclaim disk via `POST /api/admin/drop_fts` (refused while the flag is on).
 - `DLQ_MAX_FILES` (1000), `DLQ_MAX_DISK_MB` (500), `DLQ_MAX_RETRIES` (10)
 - `GRAPHRAG_WORKER_COUNT` (16), `GRAPHRAG_EVENT_QUEUE_SIZE` (100000) — sized for 100–200 services; raise further if `otelcontext_graphrag_events_dropped_total` climbs
+- `INGEST_MIN_SEVERITY` (`INFO`), `STORE_MIN_SEVERITY` (`""` = same as ingest) — two-tier log severity gate. The ingest gate runs at the OTLP receiver and **drops the log entirely** below the threshold (no in-memory enrichment either). The store gate runs at the persist boundary inside the async pipeline (`internal/ingest/pipeline.go:process`) and **only skips the DB row write** — the log still flows through `LogCallback` so vectordb indexing, GraphRAG Drain template mining, and span/trace correlation see it. Use case: `INGEST_MIN_SEVERITY=DEBUG STORE_MIN_SEVERITY=WARN` keeps SQLite small while letting in-memory anomaly detection benefit from the verbose stream. Setting `STORE_MIN_SEVERITY` ≤ `INGEST_MIN_SEVERITY` is a no-op (logged as a warning at startup). Drops surface via `Pipeline.Stats().StoreFiltered`.
 - `INGEST_ASYNC_ENABLED` (true), `INGEST_PIPELINE_QUEUE_SIZE` (50000), `INGEST_PIPELINE_WORKERS` (8) — async ingest pipeline (`internal/ingest/pipeline.go`). Hybrid backpressure: <90% accept all, 90–100% drop healthy batches (errors/slow always pass), 100% return gRPC `RESOURCE_EXHAUSTED`. Set `INGEST_ASYNC_ENABLED=false` to revert to synchronous DB writes inside `Export()`. Drops surface as `otelcontext_ingest_pipeline_dropped_total{signal,reason}`.
 - `GRPC_MAX_RECV_MB` (16), `GRPC_MAX_CONCURRENT_STREAMS` (1000) — OTLP gRPC server caps, validated to 1..256 and 1..1_000_000
 - `RETENTION_BATCH_SIZE` (50000), `RETENTION_BATCH_SLEEP_MS` (1) — purge pacing; raise the sleep on busy production DBs

internal/config/config.go

@@ -25,6 +25,14 @@ type Config struct {
 	IngestAllowedServices  string
 	IngestExcludedServices string
 
+	// Storage Filtering. Logs that pass IngestMinSeverity (so they reach the
+	// receiver and feed in-memory consumers like vectordb / GraphRAG) but
+	// fall below StoreMinSeverity are skipped during the DB persist pass —
+	// only the row-write is dropped, not the in-memory enrichment. Empty
+	// (default) means StoreMinSeverity == IngestMinSeverity, i.e. no
+	// behavior change vs. the single-threshold semantics.
+	StoreMinSeverity string
+
 	// DB Connection Pool
 	DBMaxOpenConns    int
 	DBMaxIdleConns    int
@@ -244,6 +252,7 @@ func Load(customPath string) (*Config, error) {
 		DLQReplayInterval: getEnv("DLQ_REPLAY_INTERVAL", "5m"),
 
 		IngestMinSeverity:      getEnv("INGEST_MIN_SEVERITY", "INFO"),
+		StoreMinSeverity:       getEnv("STORE_MIN_SEVERITY", ""),
 		IngestAllowedServices:  getEnv("INGEST_ALLOWED_SERVICES", ""),
 		IngestExcludedServices: getEnv("INGEST_EXCLUDED_SERVICES", ""),
 

internal/ingest/otlp.go

@@ -682,6 +682,11 @@ func getServiceName(attrs []*commonpb.KeyValue) string {
 	return "unknown-service"
 }
 
+// ParseSeverity is the exported wrapper for parseSeverity. Used by main.go
+// to translate the STORE_MIN_SEVERITY env value into the integer rank the
+// pipeline's second-tier filter expects.
+func ParseSeverity(level string) int { return parseSeverity(level) }
+
 // Filtering Helpers
 func parseSeverity(level string) int {
 	switch strings.ToUpper(level) {

internal/ingest/pipeline.go

@@ -152,6 +152,15 @@ type Pipeline struct {
 	tenantMu       sync.Mutex
 	tenantInFlight map[string]int
 
+	// storeMinSeverity is the second-tier severity gate applied at persist
+	// time inside process(). Logs in a Batch with severity below this
+	// threshold are dropped from the BatchCreateAll write but still feed
+	// the LogCallback (so vectordb / GraphRAG / Drain mining still see
+	// them). 0 disables the second tier — every log that survived
+	// IngestMinSeverity at the receiver is also persisted.
+	storeMinSeverity int
+	storeFiltered    atomic.Int64
+
 	stopCh chan struct{}
 	once   sync.Once
 	wg     sync.WaitGroup
@@ -225,6 +234,22 @@ func (p *Pipeline) SetPerTenantCap(n int) {
 	p.perTenantCap = n
 }
 
+// SetStoreMinSeverity configures the second-tier severity gate applied at
+// persist time. Logs below `level` are dropped from the BatchCreateAll write
+// but still flow through the LogCallback so in-memory consumers (vectordb,
+// GraphRAG Drain mining, anomaly correlation) keep working. 0 disables the
+// second tier — every log surviving IngestMinSeverity at the receiver is
+// also persisted (legacy behavior).
+//
+// `level` is the integer rank from parseSeverity ("DEBUG"=10 .. "FATAL"=50).
+// Startup-only — call before Start().
+func (p *Pipeline) SetStoreMinSeverity(level int) {
+	if level < 0 {
+		level = 0
+	}
+	p.storeMinSeverity = level
+}
+
 // TenantDropped reports the cumulative number of healthy submissions
 // rejected because the submitting tenant was at the per-tenant cap.
 // Distinct from RejectedFull (queue at hard capacity) and
@@ -353,6 +378,7 @@ func (p *Pipeline) Stats() PipelineStats {
 		DroppedHealthy:  p.droppedHealthy.Load(),
 		RejectedFull:    p.rejectedFull.Load(),
 		ProcessFailures: p.processFailures.Load(),
+		StoreFiltered:   p.storeFiltered.Load(),
 		QueueDepth:      len(p.queue),
 		Capacity:        p.cfg.Capacity,
 	}
@@ -365,6 +391,7 @@ type PipelineStats struct {
 	DroppedHealthy  int64
 	RejectedFull    int64
 	ProcessFailures int64
+	StoreFiltered   int64 // logs dropped by STORE_MIN_SEVERITY at persist time
 	QueueDepth      int
 	Capacity        int
 }
@@ -434,15 +461,33 @@ func (p *Pipeline) process(b *Batch) {
 		return
 	}
 
-	if err := p.writer.BatchCreateAll(b.Traces, b.Spans, b.Logs); err != nil {
+	// Apply the second-tier store-severity gate. Logs below the threshold
+	// are dropped from the persist set but still flow through the callback
+	// so in-memory enrichers (vectordb, GraphRAG Drain) keep seeing them.
+	logsToPersist := b.Logs
+	if p.storeMinSeverity > 0 && len(b.Logs) > 0 {
+		kept := make([]storage.Log, 0, len(b.Logs))
+		for _, l := range b.Logs {
+			if shouldIngestSeverity(l.Severity, p.storeMinSeverity) {
+				kept = append(kept, l)
+			} else {
+				p.storeFiltered.Add(1)
+			}
+		}
+		logsToPersist = kept
+	}
+
+	if err := p.writer.BatchCreateAll(b.Traces, b.Spans, logsToPersist); err != nil {
 		slog.Error("ingest pipeline: BatchCreateAll failed", "error", err)
 		p.processFailures.Add(1)
 		return
 	}
 
 	// Callbacks fire only after the transaction commits successfully — a
 	// rolled-back batch must not feed downstream consumers (GraphRAG etc.)
-	// data that no longer exists in the DB.
+	// data that no longer exists in the DB. The LogCallback intentionally
+	// iterates over the FULL b.Logs slice, not logsToPersist — even logs
+	// dropped by the store-severity gate must reach in-memory enrichers.
 	if b.SpanCallback != nil {
 		for _, s := range b.Spans {
 			b.SpanCallback(s)

internal/ingest/pipeline_test.go

@@ -547,3 +547,127 @@ func TestPipeline_PanicInCallbackRecovered(t *testing.T) {
 		t.Errorf("expected ProcessFailures > 0 after callback panic")
 	}
 }
+
+// TestPipeline_StoreMinSeverity_DropsBelowThresholdFromPersist verifies that
+// when SetStoreMinSeverity is configured, logs below the threshold are
+// dropped from BatchCreateAll — but the LogCallback still fires for them
+// so in-memory enrichers (vectordb, GraphRAG) keep seeing every log that
+// passed IngestMinSeverity at the receiver.
+func TestPipeline_StoreMinSeverity_DropsBelowThresholdFromPersist(t *testing.T) {
+	t.Parallel()
+	w := &fakeWriter{}
+	p := NewPipeline(w, nil, PipelineConfig{Capacity: 10, Workers: 1, SoftThreshold: 0.9})
+	// Threshold = WARN (rank 30); INFO (20) is below, ERROR (40) is above.
+	p.SetStoreMinSeverity(ParseSeverity("WARN"))
+	p.Start(context.Background())
+	defer p.Stop()
+
+	var callbackSeen []string
+	var cbMu sync.Mutex
+	cb := func(l storage.Log) {
+		cbMu.Lock()
+		defer cbMu.Unlock()
+		callbackSeen = append(callbackSeen, l.Severity)
+	}
+
+	b := &Batch{
+		Type:   SignalLogs,
+		Tenant: "t1",
+		Logs: []storage.Log{
+			{Body: "info-row", Severity: "INFO"},
+			{Body: "warn-row", Severity: "WARN"},
+			{Body: "err-row", Severity: "ERROR"},
+		},
+		LogCallback: cb,
+	}
+	if err := p.Submit(b); err != nil {
+		t.Fatalf("submit: %v", err)
+	}
+
+	if !waitFor(t, 5*time.Second, func() bool { return p.Stats().Processed >= 1 }) {
+		t.Fatalf("batch never processed")
+	}
+
+	// Persist: only WARN + ERROR should reach the writer.
+	w.mu.Lock()
+	persistedCount := 0
+	persistedSeverities := []string{}
+	for _, call := range w.logsCalls {
+		for _, l := range call {
+			persistedCount++
+			persistedSeverities = append(persistedSeverities, l.Severity)
+		}
+	}
+	w.mu.Unlock()
+	if persistedCount != 2 {
+		t.Fatalf("expected 2 logs persisted (WARN+ERROR), got %d: %v", persistedCount, persistedSeverities)
+	}
+	for _, sev := range persistedSeverities {
+		if sev == "INFO" {
+			t.Errorf("INFO log was persisted but should have been gated by store-min-severity")
+		}
+	}
+
+	// Callback: must fire for ALL THREE logs (INFO included), since the
+	// in-memory enrichment path is independent of the persist gate.
+	cbMu.Lock()
+	defer cbMu.Unlock()
+	if len(callbackSeen) != 3 {
+		t.Fatalf("expected LogCallback to fire 3 times (incl. gated INFO), got %d: %v", len(callbackSeen), callbackSeen)
+	}
+	infoCb := false
+	for _, sev := range callbackSeen {
+		if sev == "INFO" {
+			infoCb = true
+		}
+	}
+	if !infoCb {
+		t.Errorf("INFO log did not reach LogCallback — in-memory enrichment path broken: %v", callbackSeen)
+	}
+
+	// Stats: storeFiltered should report exactly 1 (the INFO drop).
+	if got := p.Stats().StoreFiltered; got != 1 {
+		t.Errorf("Stats().StoreFiltered = %d, want 1", got)
+	}
+}
+
+// TestPipeline_StoreMinSeverity_Disabled_PersistsAllLogs verifies the legacy
+// path: when SetStoreMinSeverity is NOT called (or set to 0), every log in a
+// Batch is persisted regardless of severity.
+func TestPipeline_StoreMinSeverity_Disabled_PersistsAllLogs(t *testing.T) {
+	t.Parallel()
+	w := &fakeWriter{}
+	p := NewPipeline(w, nil, PipelineConfig{Capacity: 10, Workers: 1, SoftThreshold: 0.9})
+	// No SetStoreMinSeverity call → gate disabled.
+	p.Start(context.Background())
+	defer p.Stop()
+
+	b := &Batch{
+		Type:   SignalLogs,
+		Tenant: "t1",
+		Logs: []storage.Log{
+			{Body: "info-row", Severity: "INFO"},
+			{Body: "debug-row", Severity: "DEBUG"},
+			{Body: "err-row", Severity: "ERROR"},
+		},
+	}
+	if err := p.Submit(b); err != nil {
+		t.Fatalf("submit: %v", err)
+	}
+	if !waitFor(t, 5*time.Second, func() bool { return p.Stats().Processed >= 1 }) {
+		t.Fatalf("batch never processed")
+	}
+
+	w.mu.Lock()
+	defer w.mu.Unlock()
+	total := 0
+	for _, call := range w.logsCalls {
+		total += len(call)
+	}
+	if total != 3 {
+		t.Fatalf("expected all 3 logs persisted with gate disabled, got %d", total)
+	}
+	if got := p.Stats().StoreFiltered; got != 0 {
+		t.Errorf("Stats().StoreFiltered = %d, want 0 with gate disabled", got)
+	}
+}

main.go

@@ -505,6 +505,32 @@ func main() {
 			Workers:  cfg.IngestPipelineWorkers,
 		})
 		ingestPipeline.SetPerTenantCap(cfg.IngestPipelinePerTenantCap)
+
+		// Second-tier severity gate. Empty STORE_MIN_SEVERITY means "use the
+		// same threshold as INGEST_MIN_SEVERITY" — i.e. behavior is identical
+		// to the legacy single-threshold path. Only enable the gate when the
+		// store threshold is strictly higher than the ingest threshold; equal
+		// or lower is wasted work since the receiver has already dropped the
+		// affected logs.
+		ingestRank := ingest.ParseSeverity(cfg.IngestMinSeverity)
+		storeRank := ingestRank
+		if cfg.StoreMinSeverity != "" {
+			storeRank = ingest.ParseSeverity(cfg.StoreMinSeverity)
+		}
+		if storeRank > ingestRank {
+			ingestPipeline.SetStoreMinSeverity(storeRank)
+			slog.Info("🪛 Store-severity gate enabled",
+				"ingest_min", cfg.IngestMinSeverity,
+				"store_min", cfg.StoreMinSeverity,
+				"note", "logs below store_min reach in-memory consumers but are not persisted",
+			)
+		} else if cfg.StoreMinSeverity != "" && storeRank < ingestRank {
+			slog.Warn("STORE_MIN_SEVERITY is lower than INGEST_MIN_SEVERITY — has no effect; receiver already filters",
+				"ingest_min", cfg.IngestMinSeverity,
+				"store_min", cfg.StoreMinSeverity,
+			)
+		}
+
 		ingestPipeline.Start(context.Background())
 		traceServer.SetPipeline(ingestPipeline)
 		logsServer.SetPipeline(ingestPipeline)