From 4f8f94a9fcc307111bc6baa1eaa407ee079892c0 Mon Sep 17 00:00:00 2001
From: Harsh Raj Singhania <harshrajsinghania@proton.me>
Date: Tue, 22 Sep 2026 14:06:22 +0530
Subject: [PATCH] fix: reject batch_size < 1 and handle empty-query benchmarks

Fixes AgentPostmortem/Tenantq#22 and #23.
---
 src/tenantq/benchmark.py  |  4 ++--
 src/tenantq/embeddings.py |  2 ++
 src/tenantq/ingest.py     |  2 ++
 tests/test_benchmark.py   | 23 +++++++++++++++++++++++
 tests/test_ingest.py      | 28 ++++++++++++++++++++++++++++
 5 files changed, 57 insertions(+), 2 deletions(-)
 create mode 100644 tests/test_ingest.py

diff --git a/src/tenantq/benchmark.py b/src/tenantq/benchmark.py
index e919335..37cb9c7 100644
--- a/src/tenantq/benchmark.py
+++ b/src/tenantq/benchmark.py
@@ -152,8 +152,8 @@ def run_benchmark(
                 idx_recall.append(recall_at_k(retrieved, ref, 10))
 
         total_time = sum(latencies) / 1000.0
-        avg_r5 = sum(r5) / len(r5)
-        avg_r10 = sum(r10) / len(r10)
+        avg_r5 = (sum(r5) / len(r5)) if r5 else 0.0
+        avg_r10 = (sum(r10) / len(r10)) if r10 else 0.0
         RECALL_GAUGE.labels(mode=mode, k="5").set(avg_r5)
         RECALL_GAUGE.labels(mode=mode, k="10").set(avg_r10)
         result.modes.append(
diff --git a/src/tenantq/embeddings.py b/src/tenantq/embeddings.py
index d614f3a..6399c60 100644
--- a/src/tenantq/embeddings.py
+++ b/src/tenantq/embeddings.py
@@ -123,5 +123,7 @@ def build_embedder(kind: str, dense_model: str, sparse_model: str, dense_dim: in
 
 
 def batched(seq: Sequence, size: int) -> Iterable[Sequence]:
+    if size < 1:
+        raise ValueError("batch_size must be >= 1")
     for i in range(0, len(seq), size):
         yield seq[i : i + size]
diff --git a/src/tenantq/ingest.py b/src/tenantq/ingest.py
index c20ce8b..747ce5f 100644
--- a/src/tenantq/ingest.py
+++ b/src/tenantq/ingest.py
@@ -83,6 +83,8 @@ def ingest_documents(
     parallelism: int = 4,
 ) -> IngestReport:
     """Embed and upsert ``documents`` in parallel batches."""
+    if batch_size < 1:
+        raise ValueError("batch_size must be >= 1")
     start = time.perf_counter()
     batches = list(batched(list(documents), batch_size))
 
diff --git a/tests/test_benchmark.py b/tests/test_benchmark.py
index ff3d9db..08f254d 100644
--- a/tests/test_benchmark.py
+++ b/tests/test_benchmark.py
@@ -31,3 +31,26 @@ def test_benchmark_produces_numbers(ingested, settings, embedder, dataset):
     assert not math.isnan(modes["dense"].index_recall_at_10)
     # hybrid should be at least as good as the weaker single mode on recall@10
     assert modes["hybrid"].recall_at_10 >= min(modes["dense"].recall_at_10, modes["sparse"].recall_at_10)
+
+
+def test_run_benchmark_empty_queries_returns_empty_modes(settings):
+    from unittest.mock import MagicMock
+
+    from tenantq.data import Dataset
+
+    embedder = MagicMock()
+    embedder.embed_dense.return_value = []
+    result = run_benchmark(
+        MagicMock(),
+        settings,
+        embedder,
+        Dataset(documents=[], queries=[]),
+        modes=("dense",),
+    )
+    assert result.settings_summary["n_queries"] == 0
+    assert len(result.modes) == 1
+    mode = result.modes[0]
+    assert mode.n_queries == 0
+    assert mode.recall_at_5 == 0.0
+    assert mode.recall_at_10 == 0.0
+    assert mode.qps == 0.0
diff --git a/tests/test_ingest.py b/tests/test_ingest.py
new file mode 100644
index 0000000..1652512
--- /dev/null
+++ b/tests/test_ingest.py
@@ -0,0 +1,28 @@
+"""Ingestion input validation."""
+
+from __future__ import annotations
+
+from unittest.mock import MagicMock
+
+import pytest
+
+from tenantq.embeddings import batched
+from tenantq.ingest import ingest_documents
+
+
+@pytest.mark.parametrize("batch_size", [0, -1])
+def test_ingest_documents_rejects_batch_size_below_one(settings, batch_size):
+    with pytest.raises(ValueError, match="batch_size must be >= 1"):
+        ingest_documents(
+            MagicMock(),
+            settings,
+            MagicMock(),
+            documents=[],
+            batch_size=batch_size,
+        )
+
+
+@pytest.mark.parametrize("size", [0, -3])
+def test_batched_rejects_non_positive_size(size):
+    with pytest.raises(ValueError, match="batch_size must be >= 1"):
+        list(batched([1, 2, 3], size))