From 2a9fee99d8575793a7c7781f27c2e5600da59c14 Mon Sep 17 00:00:00 2001
From: Peter Steinberger <steipete@gmail.com>
Date: Fri, 16 Oct 2026 14:22:18 -0700
Subject: [PATCH] fix: avoid SQLite variable limit in thread lookups

---
 CHANGELOG.md                      |  2 ++
 docs/clustering.md                |  2 ++
 internal/store/thread_ids_test.go | 40 +++++++++++++++++++++++++++++++
 internal/store/vectors.go         | 13 ++++------
 4 files changed, 48 insertions(+), 9 deletions(-)
 create mode 100644 internal/store/thread_ids_test.go

diff --git a/CHANGELOG.md b/CHANGELOG.md
index 8bbf9c02..502adb49 100644
--- a/CHANGELOG.md
+++ b/CHANGELOG.md
@@ -2,6 +2,8 @@
 
 ## Unreleased
 
+- Fix large neighbor and search result lookups failing with SQLite's SQL variable limit. Thanks @masatohoshino for reporting.
+
 ## 0.15.2 - 2026-10-16
 
 **Highlights:** Reliable local database initialization inside Git worktrees and refreshed dependencies.
diff --git a/docs/clustering.md b/docs/clustering.md
index 74d9bd1e..294c3eb8 100644
--- a/docs/clustering.md
+++ b/docs/clustering.md
@@ -130,6 +130,8 @@ gitcrawl neighbors owner/repo --number 123 --limit 10
 
 Useful for "what else looks like this?" without committing to a cluster. If the source row is present but lacks a current embedding, the CLI command backfills that one row when an OpenAI key is configured; otherwise it prints the exact `gitcrawl embed --number ... --limit 1` command to run. The TUI's `n` shortcut and "Enter on a member" also load vector neighbors for the selected row.
 
+Large `--limit` values are supported without hitting SQLite's SQL variable limit when loading result metadata. Neighbor search still scores all eligible stored vectors, so larger archives can take longer even with a small result limit.
+
 ## Tuning recipes
 
 ### My clusters are too greedy
diff --git a/internal/store/thread_ids_test.go b/internal/store/thread_ids_test.go
new file mode 100644
index 00000000..1433ec7a
--- /dev/null
+++ b/internal/store/thread_ids_test.go
@@ -0,0 +1,40 @@
+package store
+
+import (
+	"context"
+	"path/filepath"
+	"testing"
+)
+
+func TestThreadsByIDsBeyondSQLiteParameterLimit(t *testing.T) {
+	ctx := context.Background()
+	st, err := Open(ctx, filepath.Join(t.TempDir(), "gitcrawl.db"))
+	if err != nil {
+		t.Fatal(err)
+	}
+	defer st.Close()
+	repoID, threadIDs := seedVectorThreads(t, ctx, st)
+	otherRepoID, err := st.UpsertRepository(ctx, Repository{Owner: "fixture", Name: "other", FullName: "fixture/other"})
+	if err != nil {
+		t.Fatal(err)
+	}
+	otherID, err := st.UpsertThread(ctx, Thread{RepoID: otherRepoID, GitHubID: "other", Number: 1, Kind: "issue", State: "open", Title: "other repository", LabelsJSON: "[]", AssigneesJSON: "[]", RawJSON: "{}"})
+	if err != nil {
+		t.Fatal(err)
+	}
+	ids := make([]int64, 32768)
+	for i := range ids {
+		ids[i] = int64(i + 100000)
+	}
+	ids[0] = threadIDs[0]
+	ids[len(ids)/2] = otherID
+	ids[len(ids)-2] = threadIDs[0]
+	ids[len(ids)-1] = threadIDs[1]
+	threads, err := st.ThreadsByIDs(ctx, repoID, ids)
+	if err != nil {
+		t.Fatalf("large thread lookup: %v", err)
+	}
+	if len(threads) != 2 || threads[threadIDs[0]].Number != 301 || threads[threadIDs[1]].Number != 302 {
+		t.Fatalf("expected only the two requested repository threads, got %+v", threads)
+	}
+}
diff --git a/internal/store/vectors.go b/internal/store/vectors.go
index 1d20095f..85057f10 100644
--- a/internal/store/vectors.go
+++ b/internal/store/vectors.go
@@ -8,7 +8,6 @@ import (
 	"errors"
 	"fmt"
 	"math"
-	"strings"
 	"unicode"
 )
 
@@ -138,17 +137,13 @@ func (s *Store) ThreadsByIDs(ctx context.Context, repoID int64, ids []int64) (ma
 	if len(ids) == 0 {
 		return map[int64]Thread{}, nil
 	}
-	placeholders := make([]string, 0, len(ids))
-	args := []any{repoID}
-	for _, id := range ids {
-		placeholders = append(placeholders, "?")
-		args = append(args, id)
-	}
+	// Integer slices always marshal successfully; one JSON parameter avoids SQLite's bind limit.
+	idJSON, _ := json.Marshal(ids)
 	rows, err := s.db.QueryContext(ctx, `
 		select `+s.threadSelectColumns(ctx, "")+`
 		from threads
-		where repo_id = ? and id in (`+strings.Join(placeholders, ",")+`)
-	`, args...)
+		where repo_id = ? and id in (select value from json_each(?))
+	`, repoID, string(idJSON))
 	if err != nil {
 		return nil, fmt.Errorf("select threads by id: %w", err)
 	}