Convex Tracer

Powerful Observability and tracing for Convex applications. Track function calls across queries, mutations, and actions with detailed insights, nested spans, and automatic error tracking.

Why use Convex Tracer?

• Deep Visibility: See exactly how your Convex functions execute, including nested calls and cross-function traces
• Debug Production Issues: Preserve error traces and have a complete view of of what went wrong, were it went wrong and why
• Trace Sampling: Control costs with configurable sample rates while preserving important traces
• Zero Boilerplate: Simple wrapper functions that feel natural in Convex

Perfect for complex workflows like multi-step order processing, payment flows, or any scenario where you need to understand what's happening across multiple function calls.

Installation

Install the component:

npm install convex-tracer

Create a convex.config.ts file in your app's convex/ folder and install the component:

// convex/convex.config.ts
import { defineApp } from "convex/server";
import tracer from "convex-tracer/convex.config";

const app = defineApp();
app.use(tracer);

export default app;

Quick Start

Create a tracer instance in your Convex backend:

// convex/tracer.ts
import { Tracer } from "convex-tracer";
import { components } from "./_generated/api";
import { DataModel } from "./_generated/dataModel";

export const {
  tracedQuery,
  tracedMutation,
  tracedAction,
  internalTracedQuery,
  internalTracedMutation,
  internalTracedAction,
  tracer,
} = new Tracer<DataModel>(
  components.tracer,
  // Default options
  {
    sampleRate: 0.1, // Sample 10% of traces
    preserveErrors: true, // Always keep error traces
    retentionMinutes: 120, // Keep traces for 2 hours
  },
);

Use traced functions just like regular Convex functions:

// convex/shop.ts
import { v } from "convex/values";
import { tracedMutation } from "./tracer";

export const createOrder = tracedMutation({
  name: "createOrder",
  args: {
    customerId: v.id("customers"),
    items: v.array(
      v.object({
        productId: v.id("products"),
        quantity: v.number(),
      }),
    ),
  },
  handler: async (ctx, args) => {
    await ctx.tracer.info("Starting order creation", {
      customerId: args.customerId,
      itemCount: args.items.length,
    });

    // Your business logic here
    const orderId = await ctx.db.insert("orders", {
      customerId: args.customerId,
      items: args.items,
      status: "pending",
    });

    await ctx.tracer.info("Order created successfully", { orderId });

    return orderId;
  },
});

Core Features

Nested Spans

Create detailed traces with nested operations:

export const processPayment = tracedMutation({
  name: "processPayment",
  args: { orderId: v.id("orders"), amount: v.number() },
  handler: async (ctx, { orderId, amount }) => {
    for (const item of items) {
      const reservation = await ctx.tracer.withSpan(
        `reserveItem_${item.productId}`,
        async (span) => {
          await span.updateMetadata({
            productId: item.productId,
            requestedQty: item.quantity,
          });

          const inventory = await ctx.db
            .query("inventory")
            .withIndex("by_product", (q) => q.eq("productId", item.productId))
            .first();

          // More logic here
        },
      );
    }

    return result;
  },
});

Cross-Function Tracing

Automatically trace calls across multiple functions:

export const createOrder = tracedMutation({
  name: "createOrder",
  args: { customerId: v.id("customers"), items: v.array(...) },
  handler: async (ctx, args) => {
    // This call is automatically traced as part of the same trace
    const validation = await ctx.runTracedMutation(
      internal.shop.validateCustomer,
      { customerId: args.customerId }
    );

    // Process payment - also traced
    const payment = await ctx.runTracedMutation(
      internal.shop.processPayment,
      { orderId, amount: total }
    );

    return { orderId, status: "confirmed" };
  },
});

Lifecycle Hooks

Control trace behavior with lifecycle callbacks:

export const getProductWithInventory = tracedQuery({
  name: "getProductWithInventory",
  args: { productId: v.id("products") },
  onSuccess: async (ctx, args, result) => {
    if (result.inventory < 10) {
      await ctx.tracer.warn("Low inventory detected", {
        productId: args.productId,
        inventory: result.inventory,
      });
      await ctx.tracer.preserve(); // Keep this trace!
    }
  },
  onError: async (ctx, args, error) => {
    await ctx.tracer.error("Product fetch failed", {
      productId: args.productId,
      error: error.message,
    });
  },
  handler: async (ctx, { productId }) => {
    const product = await ctx.db.get(productId);
    const inventory = await checkInventory(ctx, productId);
    return { ...product, inventory };
  },
});

Logging and Metadata

Rich logging at different severity levels:

await ctx.tracer.info("Processing step complete", { step: 1 });
await ctx.tracer.warn("Approaching rate limit", { remaining: 10 });
await ctx.tracer.error("Validation failed", { reason: "INVALID_EMAIL" });

// Add metadata to the current span
await ctx.tracer.updateMetadata({
  userId: user._id,
  planType: "premium",
});

Configuration Options

Global Configuration

new Tracer<DataModel>(components.tracer, {
  sampleRate: 0.1, // Sample 10% of traces (0.0-1.0)
  preserveErrors: true, // Always preserve error traces
  retentionMinutes: 120, // Keep traces for 2 hours
});

Per-Function Configuration

export const myFunction = tracedMutation({
  name: "myFunction",
  args: { userId: v.id("users") },

  // Tracing options
  sampleRate: 1.0, // Override: trace 100% of calls
  logArgs: ["userId"], // Log specific arguments or all with "true"
  logReturn: true, // Log the return value

  // Lifecycle hooks
  onStart: async (ctx, args) => {
    await ctx.tracer.info("Function will start");
  },
  onSuccess: async (ctx, args, result) => {
    await ctx.tracer.info("Function succeeded", { result });
  },
  onError: async (ctx, args, error) => {
    await ctx.tracer.error("Function failed", { error: error.message });
  },

  handler: async (ctx, args) => {
    // Your logic here
  },
});

API Reference

Tracer Context Methods

All traced functions receive an enhanced context with these methods:

ctx.tracer.info(message, metadata?)

Log an info-level message with optional metadata.

ctx.tracer.warn(message, metadata?)

Log a warning-level message.

ctx.tracer.error(message, metadata?)

Log an error-level message.

ctx.tracer.updateMetadata(metadata)

Add metadata to the current span.

ctx.tracer.preserve()

Mark this trace to be preserved regardless of sample rate.

ctx.tracer.discard()

Discard this trace.

ctx.tracer.sample(sampleRate?)

Sample this trace with an optional override for the sample rate.

ctx.tracer.withSpan(name, callback)

Create a nested span for a block of code:

const result = await ctx.tracer.withSpan("spanName", async (span) => {
  await span.info("Inside nested span");
  await span.updateMetadata({ key: "value" });
  return someValue;
});

span.info(message, metadata?)

Log an info-level message with optional metadata.

span.warn(message, metadata?)

Log a warning-level message.

span.error(message, metadata?)

Log an error-level message.

span.updateMetadata(metadata)

Add metadata to the current span.

span.withSpan(name, callback)

Create a nested span for a block of code:

const result = await span.withSpan("createPost", async (span) => {
  await span.info("Creating post");
  return await ctx.db.insert("posts", args);
});

ctx.runTracedQuery(funcRef, args)

Call another traced query while maintaining the trace context.

ctx.runTracedMutation(funcRef, args)

Call another traced mutation while maintaining the trace context.

ctx.runTracedAction(funcRef, args)

Call another traced action while maintaining the trace context (actions only).

Retrieving Traces

Query traces from your frontend or other functions:

// In your convex functions
import { tracer } from "./tracer";

export const getTrace = query({
  args: { traceId: v.string() },
  handler: async (ctx, args) => {
    return await tracer.tracer.getTrace(ctx, args.traceId);
  },
});

export const listTraces = query({
  args: {
    status: v.optional(v.union(v.literal("success"), v.literal("error"))),
    limit: v.optional(v.number()),
  },
  handler: async (ctx, args) => {
    return await tracer.tracer.listTraces(ctx, args);
  },
});

React Hooks

Use traced functions in your React components:

import { useTracedMutation } from "convex-tracer/react";
import { api } from "../convex/_generated/api";

function MyComponent() {
  const createOrder = useTracedMutation(api.shop.createOrder);

  const handleOrder = async () => {
    const result = await createOrder({
      customerId: "...",
      items: [...]
    });

    if (result.success) {
      console.log("Order created:", result.data);
    } else {
      console.error("Order failed:", result.error);
    }
  };

  return <button onClick={handleOrder}>Create Order</button>;
}

Advanced Patterns

Multi-Step Workflows

Trace complex workflows across multiple functions:

export const processOrder = tracedMutation({
  name: "processOrder",
  handler: async (ctx, args) => {
    // Step 1: Validate customer
    const validation = await ctx.runTracedQuery(internal.validateCustomer, {
      customerId: args.customerId,
    });

    // Step 2: Reserve inventory
    await ctx.runTracedMutation(internal.reserveInventory, {
      items: args.items,
    });

    // Step 3: Process payment
    await ctx.runTracedMutation(internal.processPayment, { amount: total });

    // Step 4: Send notifications (async)
    await ctx.scheduler.runAfter(0, api.sendNotification, {
      orderId,
      __traceContext: {
        traceId: ctx.tracer.getTraceId(),
        spanId: ctx.tracer.getSpanId(),
      },
    });
  },
});

Conditional Preservation

Preserve traces based on business logic:

export const placeOrder = tracedMutation({
  name: "placeOrder",
  onSuccess: async (ctx, args, result) => {
    // Preserve high-value orders
    if (result.total > 1000) {
      await ctx.tracer.preserve();
    }

    // Preserve orders from VIP customers
    const customer = await ctx.db.get(args.customerId);
    if (customer.vipStatus) {
      await ctx.tracer.preserve();
    }
  },
  handler: async (ctx, args) => {
    // Process order...
  },
});

Examples

See more detailed examples in example/convex/shop.ts

Found a bug? Feature request? File it here.

Best Practices

1. Use descriptive span names: "validatePaymentMethod" not "step1"
2. Add relevant metadata: Include IDs, counts, and business-relevant data
3. Preserve strategically: Don't preserve everything. Focus on errors and edge cases
4. Sample appropriately: Use low sample rates in production (0.05-0.15)
5. Log at the right level: info for normal flow, warn for concerning but handled issues, error for failures

⚠️ Important: TracedQueries Run as Mutations

Critical difference from other traced functions: tracedQuery functions run as mutations, not queries. This is necessary to enable tracing (which requires writes to the trace tables), but it has important implications:

Breaking Change: Loss of Reactivity

Unlike regular Convex queries, tracedQuery results do not update reactively. This means:

• ❌ Your UI won't automatically re-render when data changes
• ❌ You lose Convex's real-time subscription benefits
• ❌ You must manually refetch to get updated data

When to Use TracedQueries

Use tracedQuery only when you specifically need tracing:

• ✅ Understanding complex data flows
• ✅ Data fetches that don't need reactivity
• ✅ Debugging production issues
• ✅ Performance profiling

For normal queries that need reactivity, use regular query() functions.

Pattern: Shared Logic Between Queries and TracedQueries

To maintain both reactive queries for your UI and traced queries for debugging, extract your business logic into a shared helper:

// convex/helpers/products.ts
import { QueryCtx } from "../_generated/server";
import { Id } from "./_generated/dataModel";

// Shared business logic - no tracing
export async function getProductWithInventoryLogic(
  ctx: Pick<QueryCtx, "db">,
  productId: Id<"products">,
) {
  const product = await ctx.db.get(productId);
  if (!product) {
    throw new Error("Product not found");
  }

  const inventory = await ctx.db
    .query("inventory")
    .withIndex("by_product", (q) => q.eq("productId", productId))
    .first();

  return {
    ...product,
    inventory: inventory?.quantity || 0,
  };
}

// convex/products.ts
import { v } from "convex/values";
import { query } from "./_generated/server";
import { tracedQuery } from "./tracer";
import { getProductWithInventoryLogic } from "./helpers/products";

// Regular query - USE THIS IN YOUR UI (reactive)
export const getProductWithInventory = query({
  args: { productId: v.id("products") },
  handler: async (ctx, { productId }) => {
    return await getProductWithInventoryLogic(ctx, productId);
  },
});

// Traced query (not reactive)
export const getProductWithInventoryTraced = tracedQuery({
  name: "getProductWithInventory",
  args: { productId: v.string() },
  logArgs: ["productId"],
  onSuccess: async (ctx, args, result) => {
    if (result.inventory < 10) {
      await ctx.tracer.warn("Low inventory", {
        productId: args.productId,
        inventory: result.inventory,
      });
      await ctx.tracer.preserve();
    }
  },
  handler: async (ctx, { productId }) => {
    await ctx.tracer.info("Fetching product", { productId });

    // Reuse the same business logic
    const result = await getProductWithInventoryLogic(ctx, productId);

    await ctx.tracer.info("Product fetched", {
      productId,
      inventory: result.inventory,
    });

    return result;
  },
});

For queries that sometimes need tracing, you can conditionally call either version

Run the example:

npm i
npm run dev