feat: add Loki log query support to lab-monitoring

Add 3 opt-in Loki tools (query_logs, list_labels, list_label_values)
that are registered when LOKI_URL is configured. Includes Loki HTTP
client, CLI commands (logs, labels), NixOS module option, formatting,
and tests.

Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>
This commit is contained in:
2026-02-05 20:55:39 +01:00
parent f4f859fefa
commit 859e35ab5c
13 changed files with 1014 additions and 20 deletions

View File

@@ -62,7 +62,7 @@ type HandlerOptions struct {
}
// RegisterHandlers registers all monitoring tool handlers on the MCP server.
func RegisterHandlers(server *mcp.Server, prom *PrometheusClient, am *AlertmanagerClient, opts HandlerOptions) {
func RegisterHandlers(server *mcp.Server, prom *PrometheusClient, am *AlertmanagerClient, loki *LokiClient, opts HandlerOptions) {
server.RegisterTool(listAlertsTool(), makeListAlertsHandler(am))
server.RegisterTool(getAlertTool(), makeGetAlertHandler(am))
server.RegisterTool(searchMetricsTool(), makeSearchMetricsHandler(prom))
@@ -73,6 +73,11 @@ func RegisterHandlers(server *mcp.Server, prom *PrometheusClient, am *Alertmanag
if opts.EnableSilences {
server.RegisterTool(createSilenceTool(), makeCreateSilenceHandler(am))
}
if loki != nil {
server.RegisterTool(queryLogsTool(), makeQueryLogsHandler(loki))
server.RegisterTool(listLabelsTool(), makeListLabelsHandler(loki))
server.RegisterTool(listLabelValuesTool(), makeListLabelValuesHandler(loki))
}
}
// Tool definitions
@@ -485,3 +490,183 @@ func makeCreateSilenceHandler(am *AlertmanagerClient) mcp.ToolHandler {
func parseJSON(s string, v interface{}) error {
return json.Unmarshal([]byte(s), v)
}
// Loki tool definitions
func queryLogsTool() mcp.Tool {
return mcp.Tool{
Name: "query_logs",
Description: "Execute a LogQL range query against Loki to search and retrieve log entries",
InputSchema: mcp.InputSchema{
Type: "object",
Properties: map[string]mcp.Property{
"logql": {
Type: "string",
Description: `LogQL query expression (e.g., '{job="varlogs"}', '{job="nginx"} |= "error"')`,
},
"start": {
Type: "string",
Description: "Start time: relative duration (e.g., '1h', '30m'), RFC3339 timestamp, or Unix epoch seconds. Default: 1h ago",
},
"end": {
Type: "string",
Description: "End time: relative duration (e.g., '5m'), RFC3339 timestamp, or Unix epoch seconds. Default: now",
},
"limit": {
Type: "integer",
Description: "Maximum number of log entries to return (default: 100)",
Default: 100,
},
"direction": {
Type: "string",
Description: "Sort order for log entries: 'backward' (newest first) or 'forward' (oldest first)",
Enum: []string{"backward", "forward"},
},
},
Required: []string{"logql"},
},
}
}
func listLabelsTool() mcp.Tool {
return mcp.Tool{
Name: "list_labels",
Description: "List available label names from Loki",
InputSchema: mcp.InputSchema{
Type: "object",
Properties: map[string]mcp.Property{},
},
}
}
func listLabelValuesTool() mcp.Tool {
return mcp.Tool{
Name: "list_label_values",
Description: "List values for a specific label from Loki",
InputSchema: mcp.InputSchema{
Type: "object",
Properties: map[string]mcp.Property{
"label": {
Type: "string",
Description: "Label name to get values for (e.g., 'job', 'instance')",
},
},
Required: []string{"label"},
},
}
}
// Loki handler constructors
func makeQueryLogsHandler(loki *LokiClient) mcp.ToolHandler {
return func(ctx context.Context, args map[string]interface{}) (mcp.CallToolResult, error) {
logql, _ := args["logql"].(string)
if logql == "" {
return mcp.ErrorContent(fmt.Errorf("logql is required")), nil
}
now := time.Now()
start := now.Add(-time.Hour)
end := now
if startStr, ok := args["start"].(string); ok && startStr != "" {
parsed, err := parseTimeArg(startStr, now.Add(-time.Hour))
if err != nil {
return mcp.ErrorContent(fmt.Errorf("invalid start time: %w", err)), nil
}
start = parsed
}
if endStr, ok := args["end"].(string); ok && endStr != "" {
parsed, err := parseTimeArg(endStr, now)
if err != nil {
return mcp.ErrorContent(fmt.Errorf("invalid end time: %w", err)), nil
}
end = parsed
}
limit := 100
if l, ok := args["limit"].(float64); ok && l > 0 {
limit = int(l)
}
direction := "backward"
if d, ok := args["direction"].(string); ok && d != "" {
direction = d
}
data, err := loki.QueryRange(ctx, logql, start, end, limit, direction)
if err != nil {
return mcp.ErrorContent(fmt.Errorf("log query failed: %w", err)), nil
}
return mcp.CallToolResult{
Content: []mcp.Content{mcp.TextContent(formatLogStreams(data))},
}, nil
}
}
func makeListLabelsHandler(loki *LokiClient) mcp.ToolHandler {
return func(ctx context.Context, args map[string]interface{}) (mcp.CallToolResult, error) {
labels, err := loki.Labels(ctx)
if err != nil {
return mcp.ErrorContent(fmt.Errorf("failed to list labels: %w", err)), nil
}
return mcp.CallToolResult{
Content: []mcp.Content{mcp.TextContent(formatLabels(labels))},
}, nil
}
}
func makeListLabelValuesHandler(loki *LokiClient) mcp.ToolHandler {
return func(ctx context.Context, args map[string]interface{}) (mcp.CallToolResult, error) {
label, _ := args["label"].(string)
if label == "" {
return mcp.ErrorContent(fmt.Errorf("label is required")), nil
}
values, err := loki.LabelValues(ctx, label)
if err != nil {
return mcp.ErrorContent(fmt.Errorf("failed to list label values: %w", err)), nil
}
return mcp.CallToolResult{
Content: []mcp.Content{mcp.TextContent(formatLabelValues(label, values))},
}, nil
}
}
// parseTimeArg parses a time argument that can be:
// - A relative duration (e.g., "1h", "30m", "2h30m") — interpreted as that duration ago from now
// - An RFC3339 timestamp (e.g., "2024-01-15T10:30:00Z")
// - A Unix epoch in seconds (e.g., "1705312200")
// If parsing fails, returns the provided default time.
func parseTimeArg(s string, defaultTime time.Time) (time.Time, error) {
// Try as relative duration first
if d, err := time.ParseDuration(s); err == nil {
return time.Now().Add(-d), nil
}
// Try as RFC3339
if t, err := time.Parse(time.RFC3339, s); err == nil {
return t, nil
}
// Try as Unix epoch seconds
var epoch int64
validDigits := true
for _, c := range s {
if c >= '0' && c <= '9' {
epoch = epoch*10 + int64(c-'0')
} else {
validDigits = false
break
}
}
if validDigits && len(s) > 0 {
return time.Unix(epoch, 0), nil
}
return defaultTime, fmt.Errorf("cannot parse time '%s': use relative duration (e.g., '1h'), RFC3339, or Unix epoch seconds", s)
}