Filter metric names with regex in list_metrics call

TOOLS.md

@@ -12,13 +12,15 @@ This MCP server exposes the following tools for interacting with Prometheus/Than
 **Usage Tips:**
 
 - YOU MUST CALL THIS TOOL BEFORE ANY OTHER QUERY TOOL
-- This tool MUST be called first for EVERY observability question to: 1. Discover what metrics actually exist in this environment 2. Find the EXACT metric name to use in queries 3. Avoid querying non-existent metrics
+- This tool MUST be called first for EVERY observability question to: 1. Discover what metrics actually exist in this environment 2. Find the EXACT metric name to use in queries 3. Avoid querying non-existent metrics 4. The 'name_regex' parameter should always be provided, and be a best guess of what the metric would be named like. 5. Do not use a blanket regex like .* or .+ in the 'name_regex' parameter. Use specific ones like kube.*, node.*, etc.
 - NEVER skip this step. NEVER guess metric names. Metric names vary between environments.
 - After calling this tool: 1. Search the returned list for relevant metrics 2. Use the EXACT metric name found in subsequent queries 3. If no relevant metric exists, inform the user
 
-|                |      |
-| :------------- | :--- |
-| **Parameters** | None |
+**Parameters:**
+
+| Parameter    | Type     | Required | Description                                                                                                                           |
+| :----------- | :------- | :------: | :------------------------------------------------------------------------------------------------------------------------------------ |
+| `name_regex` | `string` | ✅        | Regex pattern to filter metric names (e.g., 'http_.*', 'node_.*', 'kube.*'). This parameter is required. Don't pass in blanket regex. |
 
 **Output Schema:**
 

pkg/mcp/handlers.go

@@ -17,7 +17,7 @@ func ListMetricsHandler(opts ObsMCPOptions) func(context.Context, mcp.CallToolRe
 			return mcp.NewToolResultError(fmt.Sprintf("failed to create Prometheus client: %s", err.Error())), nil
 		}
 
-		return tools.ListMetricsHandler(ctx, promClient).ToMCPResult()
+		return tools.ListMetricsHandler(ctx, promClient, tools.BuildListMetricsInput(req.GetArguments())).ToMCPResult()
 	}
 }
 

pkg/mcp/handlers_test.go

@@ -16,17 +16,17 @@ import (
 
 // MockedLoader is a mock implementation of prometheus.PromClient for testing
 type MockedLoader struct {
-	ListMetricsFunc         func(ctx context.Context) ([]string, error)
+	ListMetricsFunc         func(ctx context.Context, nameRegex string) ([]string, error)
 	ExecuteRangeQueryFunc   func(ctx context.Context, query string, start, end time.Time, step time.Duration) (map[string]any, error)
 	ExecuteInstantQueryFunc func(ctx context.Context, query string, time time.Time) (map[string]any, error)
 	GetLabelNamesFunc       func(ctx context.Context, metricName string, start, end time.Time) ([]string, error)
 	GetLabelValuesFunc      func(ctx context.Context, label string, metricName string, start, end time.Time) ([]string, error)
 	GetSeriesFunc           func(ctx context.Context, matches []string, start, end time.Time) ([]map[string]string, error)
 }
 
-func (m *MockedLoader) ListMetrics(ctx context.Context) ([]string, error) {
+func (m *MockedLoader) ListMetrics(ctx context.Context, nameRegex string) ([]string, error) {
 	if m.ListMetricsFunc != nil {
-		return m.ListMetricsFunc(ctx)
+		return m.ListMetricsFunc(ctx, nameRegex)
 	}
 	return []string{}, nil
 }

pkg/mcp/tools_test.go

@@ -176,7 +176,7 @@ func TestToolParameters(t *testing.T) {
 	}{
 		{
 			tool:             CreateListMetricsTool(),
-			expectedRequired: []string{},
+			expectedRequired: []string{"name_regex"},
 			expectedOptional: []string{},
 		},
 		{

pkg/prometheus/loader.go

@@ -18,7 +18,7 @@ const (
 
 // Loader defines the interface for querying Prometheus
 type Loader interface {
-	ListMetrics(ctx context.Context) ([]string, error)
+	ListMetrics(ctx context.Context, nameRegex string) ([]string, error)
 	ExecuteRangeQuery(ctx context.Context, query string, start, end time.Time, step time.Duration) (map[string]any, error)
 	ExecuteInstantQuery(ctx context.Context, query string, time time.Time) (map[string]any, error)
 	GetLabelNames(ctx context.Context, metricName string, start, end time.Time) ([]string, error)
@@ -54,8 +54,16 @@ func (p *RealLoader) WithGuardrails(g *Guardrails) *RealLoader {
 	return p
 }
 
-func (p *RealLoader) ListMetrics(ctx context.Context) ([]string, error) {
-	labelValues, _, err := p.client.LabelValues(ctx, "__name__", []string{}, time.Now().Add(-ListMetricsTimeRange), time.Now())
+func (p *RealLoader) ListMetrics(ctx context.Context, nameRegex string) ([]string, error) {
+	var matches []string
+
+	// For blanket regex patterns like ".*", use empty matcher to get all metrics to not get 4xx.
+	if nameRegex != ".*" && nameRegex != ".+" && nameRegex != "" {
+		matcher := fmt.Sprintf("{__name__=~\"%s\"}", nameRegex) //nolint:gocritic
+		matches = []string{matcher}
+	}
+
+	labelValues, _, err := p.client.LabelValues(ctx, "__name__", matches, time.Now().Add(-ListMetricsTimeRange), time.Now())
 	if err != nil {
 		return nil, fmt.Errorf("error fetching metric names: %w", err)
 	}
@@ -81,7 +89,8 @@ func (p *RealLoader) ValidateMetricsExist(ctx context.Context, query string) err
 		return nil
 	}
 
-	availableMetricsList, err := p.ListMetrics(ctx)
+	// Use ".*" to match all metrics for validation
+	availableMetricsList, err := p.ListMetrics(ctx, ".*")
 	if err != nil {
 		return fmt.Errorf("failed to fetch available metrics: %w", err)
 	}

pkg/tools/definitions.go

@@ -6,7 +6,14 @@ var (
 		Name:        "list_metrics",
 		Description: ListMetricsPrompt,
 		Title:       "List Available Metrics",
-		Params:      []ParamDef{}, // no parameters
+		Params: []ParamDef{
+			{
+				Name:        "name_regex",
+				Type:        ParamTypeString,
+				Description: "Regex pattern to filter metric names (e.g., 'http_.*', 'node_.*', 'kube.*'). This parameter is required. Don't pass in blanket regex.",
+				Required:    true,
+			},
+		},
 		ReadOnly:    true,
 		Destructive: false,
 		Idempotent:  true,

pkg/tools/handlers.go

@@ -35,6 +35,12 @@ func GetBoolPtr(params map[string]any, key string) *bool {
 	return nil
 }
 
+func BuildListMetricsInput(args map[string]any) ListMetricsInput {
+	return ListMetricsInput{
+		NameRegex: GetString(args, "name_regex", ""),
+	}
+}
+
 func BuildInstantQueryInput(args map[string]any) InstantQueryInput {
 	return InstantQueryInput{
 		Query: GetString(args, "query", ""),
@@ -95,10 +101,16 @@ func BuildSilencesInput(args map[string]any) SilencesInput {
 }
 
 // ListMetricsHandler handles the listing of available Prometheus metrics.
-func ListMetricsHandler(ctx context.Context, promClient prometheus.Loader) *resultutil.Result {
+func ListMetricsHandler(ctx context.Context, promClient prometheus.Loader, input ListMetricsInput) *resultutil.Result {
 	slog.Info("ListMetricsHandler called")
+	slog.Debug("ListMetricsHandler params", "input", input)
+
+	// Validate required parameters
+	if input.NameRegex == "" {
+		return resultutil.NewErrorResult(fmt.Errorf("name_regex parameter is required and must be a string"))
+	}
 
-	metrics, err := promClient.ListMetrics(ctx)
+	metrics, err := promClient.ListMetrics(ctx, input.NameRegex)
 	if err != nil {
 		slog.Error("failed to list metrics", "error", err)
 		return resultutil.NewErrorResult(fmt.Errorf("failed to list metrics: %w", err))

pkg/tools/prompt.go

@@ -15,6 +15,7 @@ If the user mentions a specific alert by name, use get_alerts with a filter to r
 - This is NON-NEGOTIABLE for EVERY question
 - NEVER skip this step, even if you think you know the metric name
 - NEVER guess metric names - they vary between environments
+- Always pass in a name_regex param to it with a best guess of what the metric would be named like.
 - Search the returned list to find the exact metric name that exists
 
 **STEP 2: Call get_label_names for the metric you found**
@@ -48,6 +49,8 @@ This tool MUST be called first for EVERY observability question to:
 1. Discover what metrics actually exist in this environment
 2. Find the EXACT metric name to use in queries
 3. Avoid querying non-existent metrics
+4. The 'name_regex' parameter should always be provided, and be a best guess of what the metric would be named like.
+5. Do not use a blanket regex like .* or .+ in the 'name_regex' parameter. Use specific ones like kube.*, node.*, etc.
 
 NEVER skip this step. NEVER guess metric names. Metric names vary between environments.
 

pkg/tools/schema.go

@@ -99,6 +99,11 @@ type Matcher struct {
 
 // Input structs for handler parameters
 
+// ListMetricsInput defines the input parameters for ListMetricsHandler.
+type ListMetricsInput struct {
+	NameRegex string `json:"name_regex"`
+}
+
 // RangeQueryInput defines the input parameters for ExecuteRangeQueryHandler.
 type RangeQueryInput struct {
 	Query    string `json:"query"`

pkg/toolset/tools/handlers.go

@@ -15,7 +15,7 @@ func ListMetricsHandler(params api.ToolHandlerParams) (*api.ToolCallResult, erro
 		return api.NewToolCallResult("", fmt.Errorf("failed to create Prometheus client: %w", err)), nil
 	}
 
-	return tools.ListMetricsHandler(params.Context, promClient).ToToolsetResult()
+	return tools.ListMetricsHandler(params.Context, promClient, tools.BuildListMetricsInput(params.GetArguments())).ToToolsetResult()
 }
 
 // ExecuteInstantQueryHandler handles the execution of Prometheus instant queries.

tests/e2e/e2e_test.go

@@ -66,25 +66,34 @@ func TestHealthEndpoint(t *testing.T) {
 	}
 }
 
-func TestListMetrics(t *testing.T) {
-	resp, err := mcpClient.CallTool(t, 1, "list_metrics", map[string]any{})
+func TestListMetricsReturnsKnownMetrics(t *testing.T) {
+	resp, err := mcpClient.CallTool(t, 2, "list_metrics", map[string]any{
+		"name_regex": ".*",
+	})
 	if err != nil {
 		t.Fatalf("Failed to call list_metrics: %v", err)
 	}
 
 	if resp.Error != nil {
-		t.Errorf("MCP error: %s", resp.Error.Message)
+		t.Fatalf("MCP error: %s", resp.Error.Message)
 	}
 
-	if resp.Result == nil {
-		t.Error("Expected result, got nil")
-	}
+	// Verify known metrics from kube-prometheus are present
+	resultJSON, _ := json.Marshal(resp.Result)
+	resultStr := string(resultJSON)
 
-	t.Logf("list_metrics returned successfully")
+	expectedMetrics := []string{"up", "prometheus_build_info"}
+	for _, metric := range expectedMetrics {
+		if !strings.Contains(resultStr, metric) {
+			t.Errorf("Expected metric %q not found in results", metric)
+		}
+	}
 }
 
-func TestListMetricsReturnsKnownMetrics(t *testing.T) {
-	resp, err := mcpClient.CallTool(t, 2, "list_metrics", map[string]any{})
+func TestListMetricsReturnsKnownMetricsWithMatcher(t *testing.T) {
+	resp, err := mcpClient.CallTool(t, 2, "list_metrics", map[string]any{
+		"name_regex": "prometheus.*",
+	})
 	if err != nil {
 		t.Fatalf("Failed to call list_metrics: %v", err)
 	}
@@ -97,7 +106,7 @@ func TestListMetricsReturnsKnownMetrics(t *testing.T) {
 	resultJSON, _ := json.Marshal(resp.Result)
 	resultStr := string(resultJSON)
 
-	expectedMetrics := []string{"up", "prometheus_build_info"}
+	expectedMetrics := []string{"prometheus_build_info"}
 	for _, metric := range expectedMetrics {
 		if !strings.Contains(resultStr, metric) {
 			t.Errorf("Expected metric %q not found in results", metric)