triagent 0.1.0-alpha13 → 0.1.0-alpha18

package/src/mastra/tools/cost.ts

@@ -0,0 +1,389 @@
+import { createTool } from "@mastra/core/tools";
+import { z } from "zod";
+import { exec } from "child_process";
+import { promisify } from "util";
+import type { CostAnalysisConfig } from "../../cli/config.js";
+
+const execAsync = promisify(exec);
+
+// Default hourly rates (USD) - based on typical cloud pricing
+const DEFAULT_RATES = {
+  cpu: 0.03,     // per vCPU-hour
+  memory: 0.004, // per GB-hour
+  storage: 0.0001, // per GB-hour
+};
+
+interface ResourceUsage {
+  cpuCores: number;
+  memoryGB: number;
+  storageGB: number;
+}
+
+interface CostEstimate {
+  hourly: number;
+  daily: number;
+  monthly: number;
+  breakdown: {
+    cpu: number;
+    memory: number;
+    storage: number;
+  };
+}
+
+// Store config for cost calculations
+let costConfig: CostAnalysisConfig | null = null;
+
+export function initCostConfig(config?: CostAnalysisConfig): void {
+  costConfig = config || null;
+}
+
+function getRates(): typeof DEFAULT_RATES {
+  if (costConfig?.hourlyRates) {
+    return {
+      cpu: costConfig.hourlyRates.cpu || DEFAULT_RATES.cpu,
+      memory: costConfig.hourlyRates.memory || DEFAULT_RATES.memory,
+      storage: costConfig.hourlyRates.storage || DEFAULT_RATES.storage,
+    };
+  }
+  return DEFAULT_RATES;
+}
+
+function calculateCost(usage: ResourceUsage): CostEstimate {
+  const rates = getRates();
+
+  const cpuCost = usage.cpuCores * rates.cpu;
+  const memoryCost = usage.memoryGB * rates.memory;
+  const storageCost = usage.storageGB * rates.storage;
+
+  const hourly = cpuCost + memoryCost + storageCost;
+
+  return {
+    hourly,
+    daily: hourly * 24,
+    monthly: hourly * 24 * 30,
+    breakdown: {
+      cpu: cpuCost,
+      memory: memoryCost,
+      storage: storageCost,
+    },
+  };
+}
+
+function formatCurrency(amount: number): string {
+  return `$${amount.toFixed(4)}`;
+}
+
+async function getResourceUsage(kind: string, name: string, namespace: string): Promise<ResourceUsage | null> {
+  try {
+    // Get resource requests/limits
+    const { stdout } = await execAsync(
+      `kubectl get ${kind.toLowerCase()} ${name} -n ${namespace} -o jsonpath='{.spec.template.spec.containers[*].resources}' 2>/dev/null || kubectl get ${kind.toLowerCase()} ${name} -n ${namespace} -o jsonpath='{.spec.containers[*].resources}' 2>/dev/null`,
+      { timeout: 10000 }
+    );
+
+    // Parse CPU (convert from millicores to cores)
+    const cpuMatch = stdout.match(/"cpu":\s*"?(\d+)(m)?/);
+    let cpuCores = 0;
+    if (cpuMatch) {
+      cpuCores = cpuMatch[2] === "m"
+        ? parseInt(cpuMatch[1], 10) / 1000
+        : parseInt(cpuMatch[1], 10);
+    }
+
+    // Parse Memory (convert to GB)
+    const memMatch = stdout.match(/"memory":\s*"?(\d+)([KMGTPEi]+)?/);
+    let memoryGB = 0;
+    if (memMatch) {
+      const value = parseInt(memMatch[1], 10);
+      const unit = memMatch[2] || "";
+      const multipliers: Record<string, number> = {
+        "": 1 / (1024 * 1024 * 1024),
+        "Ki": 1 / (1024 * 1024),
+        "Mi": 1 / 1024,
+        "Gi": 1,
+        "Ti": 1024,
+      };
+      memoryGB = value * (multipliers[unit] || 1 / (1024 * 1024 * 1024));
+    }
+
+    // Get PVC storage if any
+    const { stdout: pvcStdout } = await execAsync(
+      `kubectl get pvc -n ${namespace} -l app=${name} -o jsonpath='{.items[*].spec.resources.requests.storage}' 2>/dev/null || echo ""`,
+      { timeout: 10000 }
+    );
+
+    let storageGB = 0;
+    const storageMatches = pvcStdout.matchAll(/(\d+)([KMGTPEi]+)?/g);
+    for (const match of storageMatches) {
+      const value = parseInt(match[1], 10);
+      const unit = match[2] || "";
+      const multipliers: Record<string, number> = {
+        "": 1 / (1024 * 1024 * 1024),
+        "Ki": 1 / (1024 * 1024),
+        "Mi": 1 / 1024,
+        "Gi": 1,
+        "Ti": 1024,
+      };
+      storageGB += value * (multipliers[unit] || 1 / (1024 * 1024 * 1024));
+    }
+
+    return { cpuCores, memoryGB, storageGB };
+  } catch {
+    return null;
+  }
+}
+
+async function getReplicaCount(kind: string, name: string, namespace: string): Promise<number> {
+  try {
+    const { stdout } = await execAsync(
+      `kubectl get ${kind.toLowerCase()} ${name} -n ${namespace} -o jsonpath='{.spec.replicas}' 2>/dev/null || echo "1"`,
+      { timeout: 10000 }
+    );
+    return parseInt(stdout.trim(), 10) || 1;
+  } catch {
+    return 1;
+  }
+}
+
+export const costTool = createTool({
+  id: "cost",
+  description: `Analyze resource costs and incident impact.
+Use this tool to:
+- Estimate resource costs for Kubernetes workloads
+- Calculate incident cost impact (downtime costs)
+- Suggest cost optimization opportunities
+
+Costs are calculated based on configured hourly rates or default cloud pricing.`,
+  inputSchema: z.object({
+    operation: z.enum(["resource", "incident", "optimization"]).describe(
+      "Operation: resource (estimate workload cost), incident (calculate downtime impact), optimization (find savings)"
+    ),
+    target: z.object({
+      kind: z.string().describe("Kubernetes resource kind"),
+      name: z.string().describe("Resource name"),
+      namespace: z.string().describe("Resource namespace"),
+    }).optional().describe("Target resource for cost analysis"),
+    timeRange: z.object({
+      start: z.string().describe("Incident start time (ISO 8601)"),
+      end: z.string().optional().describe("Incident end time (ISO 8601 or 'now')"),
+    }).optional().describe("Time range for incident cost calculation"),
+  }),
+  outputSchema: z.object({
+    success: z.boolean(),
+    data: z.string(),
+    costs: z.object({
+      hourly: z.number().optional(),
+      daily: z.number().optional(),
+      monthly: z.number().optional(),
+      incident: z.number().optional(),
+      business: z.number().optional(),
+    }).optional(),
+    error: z.string().optional(),
+  }),
+  // eslint-disable-next-line @typescript-eslint/no-explicit-any
+  execute: (async ({ operation, target, timeRange }: any) => {
+
+    try {
+      switch (operation) {
+        case "resource": {
+          if (!target) {
+            return {
+              success: false,
+              data: "",
+              error: "Target resource is required for cost estimation",
+            };
+          }
+
+          const usage = await getResourceUsage(target.kind, target.name, target.namespace);
+          if (!usage) {
+            return {
+              success: false,
+              data: "",
+              error: `Could not get resource usage for ${target.kind}/${target.name}`,
+            };
+          }
+
+          const replicas = await getReplicaCount(target.kind, target.name, target.namespace);
+          const totalUsage: ResourceUsage = {
+            cpuCores: usage.cpuCores * replicas,
+            memoryGB: usage.memoryGB * replicas,
+            storageGB: usage.storageGB,
+          };
+
+          const cost = calculateCost(totalUsage);
+          const rates = getRates();
+
+          const lines = [
+            `💰 Cost Analysis: ${target.kind}/${target.name}`,
+            `   Namespace: ${target.namespace}`,
+            `   Replicas: ${replicas}`,
+            ``,
+            `📊 Resource Usage (total):`,
+            `   CPU: ${totalUsage.cpuCores.toFixed(2)} cores`,
+            `   Memory: ${totalUsage.memoryGB.toFixed(2)} GB`,
+            `   Storage: ${totalUsage.storageGB.toFixed(2)} GB`,
+            ``,
+            `💵 Cost Estimate:`,
+            `   Hourly:  ${formatCurrency(cost.hourly)}`,
+            `   Daily:   ${formatCurrency(cost.daily)}`,
+            `   Monthly: ${formatCurrency(cost.monthly)}`,
+            ``,
+            `📈 Breakdown (hourly):`,
+            `   CPU:     ${formatCurrency(cost.breakdown.cpu)} (${formatCurrency(rates.cpu)}/core/hr)`,
+            `   Memory:  ${formatCurrency(cost.breakdown.memory)} (${formatCurrency(rates.memory)}/GB/hr)`,
+            `   Storage: ${formatCurrency(cost.breakdown.storage)} (${formatCurrency(rates.storage)}/GB/hr)`,
+          ];
+
+          return {
+            success: true,
+            data: lines.join("\n"),
+            costs: {
+              hourly: cost.hourly,
+              daily: cost.daily,
+              monthly: cost.monthly,
+            },
+          };
+        }
+
+        case "incident": {
+          if (!timeRange) {
+            return {
+              success: false,
+              data: "",
+              error: "Time range is required for incident cost calculation",
+            };
+          }
+
+          const startTime = new Date(timeRange.start);
+          const endTime = timeRange.end === "now" || !timeRange.end
+            ? new Date()
+            : new Date(timeRange.end);
+
+          const durationMs = endTime.getTime() - startTime.getTime();
+          const durationMinutes = durationMs / 60000;
+          const durationHours = durationMs / 3600000;
+
+          // Calculate resource cost during downtime if target provided
+          let resourceCost = 0;
+          let usageLines: string[] = [];
+
+          if (target) {
+            const usage = await getResourceUsage(target.kind, target.name, target.namespace);
+            if (usage) {
+              const replicas = await getReplicaCount(target.kind, target.name, target.namespace);
+              const totalUsage: ResourceUsage = {
+                cpuCores: usage.cpuCores * replicas,
+                memoryGB: usage.memoryGB * replicas,
+                storageGB: usage.storageGB,
+              };
+              const cost = calculateCost(totalUsage);
+              resourceCost = cost.hourly * durationHours;
+              usageLines = [
+                `   Affected: ${target.kind}/${target.name}`,
+                `   Resource cost during incident: ${formatCurrency(resourceCost)}`,
+              ];
+            }
+          }
+
+          // Calculate business impact if configured
+          let businessImpact = 0;
+          let businessLines: string[] = [];
+
+          if (costConfig?.businessImpact?.revenuePerMinute) {
+            businessImpact = costConfig.businessImpact.revenuePerMinute * durationMinutes;
+            businessLines = [
+              ``,
+              `📉 Business Impact:`,
+              `   Revenue rate: ${formatCurrency(costConfig.businessImpact.revenuePerMinute)}/min`,
+              `   Estimated lost revenue: ${formatCurrency(businessImpact)}`,
+            ];
+          }
+
+          const totalCost = resourceCost + businessImpact;
+
+          const lines = [
+            `⏱️ Incident Duration Analysis`,
+            ``,
+            `📅 Time Range:`,
+            `   Start: ${startTime.toISOString()}`,
+            `   End: ${endTime.toISOString()}`,
+            `   Duration: ${durationMinutes.toFixed(0)} minutes (${durationHours.toFixed(2)} hours)`,
+            ``,
+            `💰 Resource Cost:`,
+            ...usageLines,
+            ...businessLines,
+            ``,
+            `📊 Total Incident Cost: ${formatCurrency(totalCost)}`,
+          ];
+
+          return {
+            success: true,
+            data: lines.join("\n"),
+            costs: {
+              incident: resourceCost,
+              business: businessImpact,
+            },
+          };
+        }
+
+        case "optimization": {
+          // Get all deployments and analyze for optimization opportunities
+          const { stdout } = await execAsync(
+            `kubectl get deployments -A -o jsonpath='{range .items[*]}{.metadata.namespace},{.metadata.name},{.spec.replicas},{.spec.template.spec.containers[0].resources.requests.cpu},{.spec.template.spec.containers[0].resources.requests.memory}{" "}' 2>/dev/null || echo ""`,
+            { timeout: 30000 }
+          );
+
+          const suggestions: string[] = [];
+          const items = stdout.trim().split(" ").filter(Boolean);
+
+          for (const item of items) {
+            const [namespace, name, replicas, cpu, memory] = item.split(",");
+
+            // Check for over-provisioned resources
+            if (parseInt(replicas, 10) > 3) {
+              suggestions.push(`• ${namespace}/${name}: Consider autoscaling (currently ${replicas} replicas)`);
+            }
+
+            // Check for missing resource requests
+            if (!cpu || !memory) {
+              suggestions.push(`• ${namespace}/${name}: Add resource requests for better scheduling`);
+            }
+          }
+
+          if (suggestions.length === 0) {
+            return {
+              success: true,
+              data: "✅ No obvious cost optimization opportunities found.\n\nConsider:\n- Reviewing unused PVCs\n- Right-sizing node pools\n- Using spot/preemptible instances",
+            };
+          }
+
+          const lines = [
+            `💡 Cost Optimization Suggestions:`,
+            ``,
+            ...suggestions,
+            ``,
+            `General recommendations:`,
+            `- Review unused PVCs and delete if not needed`,
+            `- Consider using horizontal pod autoscaling`,
+            `- Use spot/preemptible instances for non-critical workloads`,
+            `- Right-size resource requests based on actual usage`,
+          ];
+
+          return {
+            success: true,
+            data: lines.join("\n"),
+          };
+        }
+
+        default:
+          return { success: false, data: "", error: `Unknown operation: ${operation}` };
+      }
+    } catch (error) {
+      return {
+        success: false,
+        data: "",
+        error: error instanceof Error ? error.message : String(error),
+      };
+    }
+  }) as any,
+});

package/src/mastra/tools/logs.ts

@@ -0,0 +1,210 @@
+import { createTool } from "@mastra/core/tools";
+import { z } from "zod";
+import { getElasticsearchClient } from "../../integrations/elasticsearch/client.js";
+import { getLokiClient } from "../../integrations/loki/client.js";
+
+export const logsTool = createTool({
+  id: "logs",
+  description: `Search and aggregate logs from centralized log systems (Elasticsearch or Loki).
+Use this tool for:
+- Searching logs beyond kubectl retention
+- Cross-pod log correlation
+- Historical log analysis
+- Log aggregation and statistics
+
+The query syntax depends on the configured log provider:
+- Elasticsearch: Lucene query syntax (e.g., "error AND pod:myapp*")
+- Loki: LogQL syntax (e.g., '{namespace="production"} |= "error"')`,
+  inputSchema: z.object({
+    operation: z.enum(["search", "tail", "aggregate"]).describe(
+      "Operation: search (query logs), tail (recent logs), aggregate (group by field)"
+    ),
+    query: z.string().describe(
+      "Search query. For ES use Lucene syntax, for Loki use LogQL"
+    ),
+    timeRange: z.object({
+      start: z.string().describe("Start time (ISO 8601 or relative like '1h')"),
+      end: z.string().optional().describe("End time (ISO 8601, 'now', or omit for current)"),
+    }).optional().describe("Time range for search"),
+    limit: z.number().default(100).describe("Maximum number of logs to return"),
+    aggregateField: z.string().optional().describe("Field to aggregate by (for aggregate operation)"),
+  }),
+  outputSchema: z.object({
+    success: z.boolean(),
+    data: z.string(),
+    count: z.number().optional(),
+    error: z.string().optional(),
+  }),
+  // eslint-disable-next-line @typescript-eslint/no-explicit-any
+  execute: (async ({ operation, query, timeRange, limit, aggregateField }: any) => {
+
+    try {
+      const esClient = getElasticsearchClient();
+      const lokiClient = getLokiClient();
+
+      // Determine which client to use
+      const client = esClient || lokiClient;
+      const clientType = esClient ? "elasticsearch" : lokiClient ? "loki" : null;
+
+      if (!client) {
+        return {
+          success: false,
+          data: "",
+          error: "No log provider configured. Set elasticsearch or loki in config.",
+        };
+      }
+
+      switch (operation) {
+        case "search": {
+          if (clientType === "elasticsearch") {
+            const logs = await esClient!.search({
+              query,
+              timeRange: timeRange ? {
+                start: parseRelativeTime(timeRange.start),
+                end: timeRange.end ? parseRelativeTime(timeRange.end) : undefined,
+              } : undefined,
+              limit,
+            });
+
+            return {
+              success: true,
+              data: logs.length > 0 ? esClient!.formatLogs(logs) : "No logs found",
+              count: logs.length,
+            };
+          } else {
+            const logs = await lokiClient!.query({
+              query,
+              start: timeRange?.start,
+              end: timeRange?.end,
+              limit,
+            });
+
+            return {
+              success: true,
+              data: logs.length > 0 ? lokiClient!.formatLogs(logs) : "No logs found",
+              count: logs.length,
+            };
+          }
+        }
+
+        case "tail": {
+          if (clientType === "elasticsearch") {
+            // For ES, tail is just search with latest logs
+            const logs = await esClient!.search({
+              query,
+              limit,
+              sort: "desc",
+            });
+
+            return {
+              success: true,
+              data: logs.length > 0 ? esClient!.formatLogs(logs) : "No recent logs",
+              count: logs.length,
+            };
+          } else {
+            const logs = await lokiClient!.tail({
+              query,
+              limit,
+            });
+
+            return {
+              success: true,
+              data: logs.length > 0 ? lokiClient!.formatLogs(logs) : "No recent logs",
+              count: logs.length,
+            };
+          }
+        }
+
+        case "aggregate": {
+          if (!aggregateField) {
+            return {
+              success: false,
+              data: "",
+              error: "aggregateField is required for aggregate operation",
+            };
+          }
+
+          if (clientType === "elasticsearch") {
+            const aggregations = await esClient!.aggregate({
+              query,
+              field: aggregateField,
+              timeRange: timeRange ? {
+                start: parseRelativeTime(timeRange.start),
+                end: timeRange.end ? parseRelativeTime(timeRange.end) : undefined,
+              } : undefined,
+            });
+
+            const lines = [`Aggregation by ${aggregateField}:\n`];
+            for (const agg of aggregations) {
+              lines.push(`  ${agg.key}: ${agg.count} logs`);
+            }
+
+            return {
+              success: true,
+              data: lines.join("\n"),
+              count: aggregations.length,
+            };
+          } else {
+            // Loki doesn't have native aggregations, use label cardinality
+            const series = await lokiClient!.getSeries([query]);
+            const counts = new Map<string, number>();
+
+            for (const s of series) {
+              const value = s[aggregateField] || "unknown";
+              counts.set(value, (counts.get(value) || 0) + 1);
+            }
+
+            const lines = [`Aggregation by ${aggregateField}:\n`];
+            for (const [key, count] of counts) {
+              lines.push(`  ${key}: ${count} streams`);
+            }
+
+            return {
+              success: true,
+              data: lines.join("\n"),
+              count: counts.size,
+            };
+          }
+        }
+
+        default:
+          return { success: false, data: "", error: `Unknown operation: ${operation}` };
+      }
+    } catch (error) {
+      return {
+        success: false,
+        data: "",
+        error: error instanceof Error ? error.message : String(error),
+      };
+    }
+  }) as any,
+});
+
+function parseRelativeTime(timeStr: string): string {
+  if (timeStr === "now") {
+    return new Date().toISOString();
+  }
+
+  // If it's already ISO format, return as-is
+  if (timeStr.includes("T") || timeStr.includes("-")) {
+    return timeStr;
+  }
+
+  // Parse relative time like "1h", "30m", "2d"
+  const match = timeStr.match(/^(\d+)([smhdw])$/);
+  if (!match) {
+    return timeStr;
+  }
+
+  const [, amount, unit] = match;
+  const now = new Date();
+  const ms = parseInt(amount, 10) * {
+    s: 1000,
+    m: 60 * 1000,
+    h: 60 * 60 * 1000,
+    d: 24 * 60 * 60 * 1000,
+    w: 7 * 24 * 60 * 60 * 1000,
+  }[unit as "s" | "m" | "h" | "d" | "w"]!;
+
+  return new Date(now.getTime() - ms).toISOString();
+}