Workspaces

# Troubleshooting

# Troubleshooting and FAQ Hub

This central hub lists common issues, failure modes, and frequently asked questions across Curiosity Workspace. When troubleshooting, start by identifying which layer is failing:

  • ingestion (connector/integration)
  • graph correctness (schemas, edges, keys)
  • indexing (text/vector search)
  • NLP/AI configuration (pipelines, embeddings, LLM provider)
  • app logic (endpoints, UI, integrations)

# Ingestion: “My connector ran, but I don’t see data”

  • Confirm the connector is writing to the correct workspace/environment.
  • Validate schema existence (node types and edge types).
  • Check that nodes are being committed (batch commit and final commit).
  • Validate keys: if keys are empty or unstable, data may be deduplicated incorrectly or duplicated.

# Graph: “Relationships are missing”

  • Ensure ingestion explicitly creates edges (relationships are not inferred).
  • Validate edge type names and direction conventions.
  • Run a neighbor summary query (or inspect a node’s neighborhood) to verify edge counts.

# Search: “Search returns no results”

  • Confirm you have enabled indexing for the relevant node types and fields.
  • Rebuild/reindex after enabling fields.
  • If filtering is enabled, confirm facets aren’t excluding all results.

# Vector search: “Semantic results are bad”

  • Confirm embedding indexing is enabled for the right fields.
  • Enable chunking for long fields.
  • Validate similarity cutoffs and evaluate using a small query set.
  • Apply context constraints (facets/graph) to reduce noise.

# NLP: “Entities aren’t being captured”

  • Confirm pipelines are configured for the correct language and field.
  • Reparse/reprocess existing nodes after updating pipeline configuration.
  • Run extraction experiments on representative samples and iterate on rules/models.

# Endpoints: “My endpoint works in the UI but fails externally”

  • Confirm you are using the correct endpoint URL for external calls.
  • Ensure the endpoint requires the correct authorization mode.
  • Verify the endpoint token scope and expiration.
  • Check logs for input validation failures.

# Installation and Deployment FAQ

Q: Why can't I reach the Workspace UI after installation?

  • A: Ensure the service is running and binding to the correct port (default 8080). Check firewall settings and reverse proxy configurations.

Q: How do I recover from a lost admin password?

  • A: Use the command-line recovery tool or environment variables to reset the initial admin credentials.

# Security and Access FAQ

Q: Why are users seeing "Access Denied" errors?

  • A: Check the User and Role Management settings. Ensure users are assigned the appropriate roles and that their SSO group mappings are correct.

Q: Can I restrict access to specific nodes in the graph?

  • A: Yes, use node-level permissions and ensure your search/AI workflows are configured to respect these visibility rules.

# Performance and Scaling FAQ

Q: Ingestion is very slow. How can I speed it up?

  • A: Increase batch sizes, use parallel ingestion, and ensure your storage volume has high I/O performance. See Performance Tuning.

Q: Search queries are timing out. What should I do?

  • A: Optimize your search DSL queries, ensure relevant fields are indexed, and check for high CPU/Memory usage on the server.

# Next steps

© 2026 Curiosity. All rights reserved.