p2.3: pin chunk 0, hybrid score, keep question on retry

Question-mode selection always keeps the lead chunk, then fills
the budget by 0.7 cosine + 0.3 term overlap. prompt_version 2.
too-large embed retry no longer halves the question (index 0).

internal/digest/digest_test.go

@@ -334,3 +334,49 @@ func TestEmbedTooLargeRetriesAndCompletes(t *testing.T) {
 		t.Fatalf("text %q", r.Text)
 	}
 }
+
+func TestHalveInputsKeepsQuestion(t *testing.T) {
+	in := []string{strings.Repeat("q", 100), strings.Repeat("c", 100)}
+	out := halveInputs(in)
+	if out[0] != in[0] {
+		t.Fatal("question halved")
+	}
+	if utf8.RuneCountInString(out[1]) != 50 {
+		t.Fatalf("chunk %d", utf8.RuneCountInString(out[1]))
+	}
+}
+
+func TestPinChunk0Infobox(t *testing.T) {
+	mux := http.NewServeMux()
+	mux.HandleFunc("/embed", func(w http.ResponseWriter, r *http.Request) {
+		var body struct {
+			Input []string `json:"input"`
+		}
+		_ = json.NewDecoder(r.Body).Decode(&body)
+		var data []map[string]any
+		for i := range body.Input {
+			vec := []float64{1, 0}
+			if i == 1 {
+				vec = []float64{0, 1}
+			}
+			data = append(data, map[string]any{"index": i, "embedding": vec})
+		}
+		_ = json.NewEncoder(w).Encode(map[string]any{"data": data})
+	})
+	ts := httptest.NewServer(mux)
+	defer ts.Close()
+	e := &Engine{EmbedURL: ts.URL + "/embed", Client: ts.Client()}
+	lead := "Initial release: 2003\nDeveloper: Example\nLicense: MIT"
+	prose := strings.Repeat("The scheduler and memory manager dominate this chapter. ", 400)
+	cleaned := lead + "\n" + strings.Repeat(prose+"\n", 40)
+	if EstTokens(cleaned) <= SmallDoc {
+		t.Fatalf("fixture too small: %d", EstTokens(cleaned))
+	}
+	body, nUsed, nTotal, err := e.selectText(context.Background(), cleaned, "when was the initial release?")
+	if err != nil {
+		t.Fatal(err)
+	}
+	if !strings.Contains(body, "Initial release: 2003") {
+		t.Fatalf("chunk 0 dropped used=%d total=%d body=%q", nUsed, nTotal, body[:min(200, len(body))])
+	}
+}