Skip to main content

Troubleshooting

Info
Last Updated: 2026-01-08

Common issues and solutions for Cortex Memory Provider with SDK v0.21.0+.

SDK v0.17.0+ Breaking Changes​

"agentId is required"​

Error:

Error: agentId is required. User-agent conversations require both a user and an agent participant. Provide agentId in your configuration, e.g., agentId: 'my-assistant'
Info

The actual error message includes additional configuration help text to guide users.

Cause: Since SDK v0.17.0, all conversations require an agentId to identify the AI agent participant.

Solution: Add agentId to your configuration:

// Old way (throws error)
const cortexMemory = createCortexMemory({
  convexUrl: process.env.CONVEX_URL!,
  memorySpaceId: "my-chatbot",
  userId: "user-123",
});

// New way (v0.17.0+)
const cortexMemory = createCortexMemory({
  convexUrl: process.env.CONVEX_URL!,
  memorySpaceId: "my-chatbot",
  userId: "user-123",
  agentId: "my-assistant", // Required!
});

"agentId must be a non-empty string"​

Cause: You provided an empty string or invalid value for agentId.

Solution: Provide a valid agent identifier:

const cortexMemory = createCortexMemory({
  // ...
  agentId: "my-assistant", // Valid non-empty string
});

Installation Issues​

"Cannot find module '@cortexmemory/sdk'"​

Terminal
$ 
npm install @cortexmemory/sdk@latest

"Peer dependency warnings"​

Ensure compatible versions:

Terminal
$ 
npm install ai@^4.0.0 convex@^1.28.2
Info

The AI SDK supports versions 3, 4, 5, and 6 (^3.0.0 || ^4.0.0 || ^5.0.0 || ^6.0.0).

Connection Issues​

"Failed to connect to Convex"​

  1. Check CONVEX_URL is set correctly in .env.local:

    CONVEX_URL=https://your-project.convex.cloud

  2. Ensure Convex is running: npx convex dev

  3. Verify Cortex backend is deployed to Convex

"Convex function not found"​

Cause: Cortex schema not deployed to Convex.

Solution: Run npx create-cortex-memories to set up the backend, or manually copy schema files from @cortexmemory/sdk/convex-dev/.

Memory Issues​

"Memory search returns empty"​

Common causes:

  1. No prior conversations - Expected for first use, send a message first
  2. Wrong memory space - Check memorySpaceId matches between requests
  3. Using local Convex - Vector search not supported locally
  4. No embeddings - Using keyword search without embedding provider

Solution:

import { embed } from 'ai';
import { createOpenAI } from '@ai-sdk/openai';

const openai = createOpenAI({ apiKey: process.env.OPENAI_API_KEY });

// Add embeddings for better search
const cortexMemory = createCortexMemory({
  convexUrl: process.env.CONVEX_URL!,
  memorySpaceId: "my-agent",
  userId: "user-123",
  agentId: "my-assistant",

  embeddingProvider: {
    generate: async (text) => {
      const { embedding } = await embed({
        model: openai.embedding("text-embedding-3-small"),
        value: text,
      });
      return embedding;
    },
  },
});

"Memory not persisting across sessions"​

Causes:

  1. enableMemoryStorage set to false
  2. Different memorySpaceId between requests
  3. Different userId between requests
  4. Convex deployment issue

Solution:

// Ensure consistent configuration
const cortexMemory = createCortexMemory({
  convexUrl: process.env.CONVEX_URL!, // Same URL
  memorySpaceId: "my-agent", // Same space ID
  userId: currentUser.id, // Same user ID
  agentId: "my-assistant",
  enableMemoryStorage: true, // Default is true
});

Graph Memory Issues​

"Graph sync not working"​

Cause: Graph database not configured correctly.

Solution: Check environment variables:

CORTEX_GRAPH_SYNC=true
NEO4J_URI=bolt://localhost:7687
NEO4J_USERNAME=neo4j
NEO4J_PASSWORD=your-password

Or use explicit configuration:

const cortexMemory = createCortexMemory({
  // ...
  enableGraphMemory: true,
  graphConfig: {
    uri: "bolt://localhost:7687",
    username: "neo4j",
    password: "your-password",
    type: "neo4j",
  },
});

"Graph connection timeout"​

Cause: Graph database not running or unreachable.

Solution:

  1. Start your Neo4j/Memgraph instance
  2. Check the URI is correct and accessible
  3. Verify firewall allows connections

Fact Extraction Issues​

"Facts not being extracted"​

Cause: Fact extraction not enabled or misconfigured.

Solution:

CORTEX_FACT_EXTRACTION=true
CORTEX_FACT_EXTRACTION_MODEL=gpt-4o-mini

Or explicitly enable:

const cortexMemory = createCortexMemory({
  // ...
  enableFactExtraction: true,
});

"Fact extraction slow"​

Cause: Using a large model for extraction.

Solution: Use a faster model:

factExtractionConfig: {
  model: "gpt-4o-mini", // Faster than gpt-4o
}

Type Issues​

"Type errors with LanguageModelV1"​

Cause: Version mismatch between AI SDK versions.

Solution:

Terminal
$ 
npm update ai @ai-sdk/openai @cortexmemory/vercel-ai-provider

"Property 'agentId' does not exist"​

Cause: Using old type definitions.

Solution: Update to latest version:

Terminal
$ 
npm update @cortexmemory/vercel-ai-provider

Edge Runtime Issues​

"process is not defined"​

Cause: Using Node.js APIs in edge runtime.

Solution: Use edge-compatible environment variable access:

// Edge-compatible
const convexUrl = process.env.CONVEX_URL!;

Most Next.js edge runtimes support process.env for build-time variables.

"Cannot use fs module"​

Cause: Edge runtimes don't have fs.

Solution: Use web-standard APIs only. Cortex is fully edge-compatible.

Performance Issues​

"Slow responses"​

Causes:

  • Too many memories searched
  • Slow embedding generation
  • Network latency to Convex

Solutions:

const cortexMemory = createCortexMemory({
  // ...
  memorySearchLimit: 3, // Reduce from default 5
  minMemoryRelevance: 0.8, // Higher = fewer, more relevant results
});

"High latency on first request"​

Cause: Cold start for Convex functions.

Solution: This is normal. Subsequent requests will be faster. Consider Convex's "scheduled functions" for warming.

Streaming Issues​

"Stream interrupted"​

Cause: Network issues or timeout.

Solution: Enable error recovery:

streamingOptions: {
  partialFailureHandling: 'store-partial', // Optional
  maxRetries: 3, // Default: 3
  generateResumeToken: true, // Default: false
  streamTimeout: 60000, // Default: 30000, increasing for reliability
}

"Streaming metrics not appearing"​

Cause: Metrics collection disabled.

Solution:

enableStreamMetrics: true, // Default is true

Debugging​

Enable Debug Logging​

const cortexMemory = createCortexMemory({
  // ...
  debug: true,
});

This will log:

  • Memory search queries and results
  • Context injection details
  • Storage operations
  • Timing information

Custom Logger​

const cortexMemory = createCortexMemory({
  // ...
  logger: {
    debug: (...args) => console.log("[DEBUG]", ...args),
    info: (...args) => console.log("[INFO]", ...args),
    warn: (...args) => console.warn("[WARN]", ...args),
    error: (...args) => console.error("[ERROR]", ...args),
  },
});

Getting Help​

Next Steps​