diff --git a/README.md b/README.md
index 799536c3..c9fa4d08 100644
--- a/README.md
+++ b/README.md
@@ -63,6 +63,7 @@ git clone https://github.com/The-OpenROAD-Project/ORAssistant.git
 ```
 
 **Step 2**: Copy the `.env.example` file, and update your `.env` file with the appropriate API keys.
+> **Note:** When configuring `EMBEDDINGS_TYPE`, valid options are `HF`, `GOOGLE_GENAI`, or `GOOGLE_VERTEXAI`. If using `HF`, no API key is needed but startup is slower. If using `GOOGLE_GENAI`, ensure `GOOGLE_API_KEY` is set.
 
 Modify the Docker `HEALTHCHECK_` variables based on the hardware requirements.
 If you have a resource-constrained PC, try increasing `HEALTHCHECK_START_PERIOD` to a value large
@@ -99,6 +100,7 @@ uv sync
 ```
 
 **Step 2**: Copy the `.env.example` file, and update your `.env` file with the appropriate API keys.
+> **Note:** When configuring `EMBEDDINGS_TYPE`, valid options are `HF`, `GOOGLE_GENAI`, or `GOOGLE_VERTEXAI`. If using `HF`, no API key is needed but startup is slower. If using `GOOGLE_GENAI`, ensure `GOOGLE_API_KEY` is set.
 
 ```bash
 cd backend
diff --git a/backend/.env.example b/backend/.env.example
index 5ce3d60e..3a0848ce 100644
--- a/backend/.env.example
+++ b/backend/.env.example
@@ -22,6 +22,7 @@ GOOGLE_GEMINI=2.0_flash
 
 LLM_TEMP=1
 
+# Choose embeddings type (options: HF, GOOGLE_GENAI, GOOGLE_VERTEXAI)
 EMBEDDINGS_TYPE=GOOGLE_GENAI
 GOOGLE_EMBEDDINGS=gemini-embedding-001
 HF_EMBEDDINGS=thenlper/gte-large
diff --git a/evaluation/auto_evaluation/tests/conftest.py b/evaluation/auto_evaluation/tests/conftest.py
new file mode 100644
index 00000000..442ec292
--- /dev/null
+++ b/evaluation/auto_evaluation/tests/conftest.py
@@ -0,0 +1,4 @@
+import sys
+import os
+
+sys.path.insert(0, os.path.join(os.path.dirname(__file__), ".."))
diff --git a/evaluation/auto_evaluation/tests/test_metrics.py b/evaluation/auto_evaluation/tests/test_metrics.py
new file mode 100644
index 00000000..d2b853be
--- /dev/null
+++ b/evaluation/auto_evaluation/tests/test_metrics.py
@@ -0,0 +1,177 @@
+"""
+Unit tests for evaluation metrics module.
+Tests verify metric factory functions return correct types,
+use correct thresholds, and handle NotImplementedError cases.
+"""
+import pytest
+from unittest.mock import MagicMock
+from deepeval.metrics import (
+    ContextualPrecisionMetric,
+    ContextualRecallMetric,
+    HallucinationMetric,
+    AnswerRelevancyMetric,
+    BiasMetric,
+    ToxicityMetric,
+    GEval,
+)
+from deepeval.models.base_model import DeepEvalBaseLLM
+
+from src.metrics.retrieval import (
+    make_contextual_precision_metric,
+    make_contextual_recall_metric,
+    make_contextual_relevancy_metric,
+    make_faithfulness_metric,
+    make_hallucination_metric,
+    PRECISION_THRESHOLD,
+    RECALL_THRESHOLD,
+    HALLUCINATION_THRESHOLD,
+)
+from src.metrics.content import (
+    make_answer_relevancy_metric,
+    make_bias_metric,
+    make_toxicity_metric,
+    ANSRELEVANCY_THRESHOLD,
+    BIAS_THRESHOLD,
+    TOXICITY_THRESHOLD,
+)
+from src.metrics.geval import make_correctness_metric
+
+
+@pytest.fixture
+def mock_model():
+    """Mock DeepEvalBaseLLM model for testing."""
+    return MagicMock(spec=DeepEvalBaseLLM)
+
+
+class TestRetrievalMetrics:
+    """Tests for retrieval-based evaluation metrics."""
+
+    def test_make_contextual_precision_metric_returns_correct_type(self, mock_model):
+        metric = make_contextual_precision_metric(mock_model)
+        assert isinstance(metric, ContextualPrecisionMetric)
+
+    def test_make_contextual_precision_metric_threshold(self, mock_model):
+        metric = make_contextual_precision_metric(mock_model)
+        assert metric.threshold == PRECISION_THRESHOLD
+
+    def test_make_contextual_precision_metric_includes_reason(self, mock_model):
+        metric = make_contextual_precision_metric(mock_model)
+        assert metric.include_reason is True
+
+    def test_make_contextual_recall_metric_returns_correct_type(self, mock_model):
+        metric = make_contextual_recall_metric(mock_model)
+        assert isinstance(metric, ContextualRecallMetric)
+
+    def test_make_contextual_recall_metric_threshold(self, mock_model):
+        metric = make_contextual_recall_metric(mock_model)
+        assert metric.threshold == RECALL_THRESHOLD
+
+    def test_make_contextual_recall_metric_includes_reason(self, mock_model):
+        metric = make_contextual_recall_metric(mock_model)
+        assert metric.include_reason is True
+
+    def test_make_hallucination_metric_returns_correct_type(self, mock_model):
+        metric = make_hallucination_metric(mock_model)
+        assert isinstance(metric, HallucinationMetric)
+
+    def test_make_hallucination_metric_threshold(self, mock_model):
+        metric = make_hallucination_metric(mock_model)
+        assert metric.threshold == HALLUCINATION_THRESHOLD
+
+    def test_make_hallucination_metric_includes_reason(self, mock_model):
+        metric = make_hallucination_metric(mock_model)
+        assert metric.include_reason is True
+
+    def test_make_contextual_relevancy_metric_raises_not_implemented(self, mock_model):
+        """ContextualRelevancyMetric is disabled due to protobuf incompatibility."""
+        with pytest.raises(NotImplementedError, match="protobuf incompatability"):
+            make_contextual_relevancy_metric(mock_model)
+
+    def test_make_faithfulness_metric_raises_not_implemented(self, mock_model):
+        """FaithfulnessMetric is disabled due to protobuf incompatibility."""
+        with pytest.raises(NotImplementedError, match="protobuf incompatability"):
+            make_faithfulness_metric(mock_model)
+
+
+class TestContentMetrics:
+    """Tests for content-based evaluation metrics."""
+
+    def test_make_answer_relevancy_metric_returns_correct_type(self, mock_model):
+        metric = make_answer_relevancy_metric(mock_model)
+        assert isinstance(metric, AnswerRelevancyMetric)
+
+    def test_make_answer_relevancy_metric_threshold(self, mock_model):
+        metric = make_answer_relevancy_metric(mock_model)
+        assert metric.threshold == ANSRELEVANCY_THRESHOLD
+
+    def test_make_answer_relevancy_metric_includes_reason(self, mock_model):
+        metric = make_answer_relevancy_metric(mock_model)
+        assert metric.include_reason is True
+
+    def test_make_bias_metric_returns_correct_type(self, mock_model):
+        metric = make_bias_metric(mock_model)
+        assert isinstance(metric, BiasMetric)
+
+    def test_make_bias_metric_threshold(self, mock_model):
+        metric = make_bias_metric(mock_model)
+        assert metric.threshold == BIAS_THRESHOLD
+
+    def test_make_bias_metric_includes_reason(self, mock_model):
+        metric = make_bias_metric(mock_model)
+        assert metric.include_reason is True
+
+    def test_make_toxicity_metric_returns_correct_type(self, mock_model):
+        metric = make_toxicity_metric(mock_model)
+        assert isinstance(metric, ToxicityMetric)
+
+    def test_make_toxicity_metric_threshold(self, mock_model):
+        metric = make_toxicity_metric(mock_model)
+        assert metric.threshold == TOXICITY_THRESHOLD
+
+    def test_make_toxicity_metric_includes_reason(self, mock_model):
+        metric = make_toxicity_metric(mock_model)
+        assert metric.include_reason is True
+
+
+class TestGEvalMetrics:
+    """Tests for GEval custom LLM-based metrics."""
+
+    def test_make_correctness_metric_returns_geval(self, mock_model):
+        metric = make_correctness_metric(mock_model)
+        assert isinstance(metric, GEval)
+
+    def test_make_correctness_metric_name(self, mock_model):
+        metric = make_correctness_metric(mock_model)
+        assert metric.name == "Correctness"
+
+    def test_make_correctness_metric_has_evaluation_steps(self, mock_model):
+        metric = make_correctness_metric(mock_model)
+        assert metric.evaluation_steps is not None
+        assert len(metric.evaluation_steps) > 0
+
+    def test_make_correctness_metric_has_criteria(self, mock_model):
+        metric = make_correctness_metric(mock_model)
+        assert metric.criteria is not None
+        assert "factually correct" in metric.criteria
+
+
+class TestThresholdValues:
+    """Tests to verify threshold constants are within valid range."""
+
+    def test_precision_threshold_valid(self):
+        assert 0.0 <= PRECISION_THRESHOLD <= 1.0
+
+    def test_recall_threshold_valid(self):
+        assert 0.0 <= RECALL_THRESHOLD <= 1.0
+
+    def test_hallucination_threshold_valid(self):
+        assert 0.0 <= HALLUCINATION_THRESHOLD <= 1.0
+
+    def test_answer_relevancy_threshold_valid(self):
+        assert 0.0 <= ANSRELEVANCY_THRESHOLD <= 1.0
+
+    def test_bias_threshold_valid(self):
+        assert 0.0 <= BIAS_THRESHOLD <= 1.0
+
+    def test_toxicity_threshold_valid(self):
+        assert 0.0 <= TOXICITY_THRESHOLD <= 1.0