Search & Discovery

Definition: Search & Discovery in Taskade is a multi-layer search system that combines full-text indexing, semantic vector search (HNSW), and file content OCR to provide comprehensive discovery across your entire workspace - projects, agents, automations, and uploaded files.

Multi-Layer Search Architecture

Taskade uses three complementary search layers to ensure you always find what you need:

Full-Text Search: Traditional keyword matching across all project content, task titles, descriptions, and metadata. Fast, exact-match results for when you know the specific terms.

Semantic Search (HNSW): AI-powered vector search that understands meaning, not just keywords. Ask "how do we handle refunds?" and find the relevant policy even if the word "refund" doesn't appear. Uses HNSW (Hierarchical Navigable Small World) indexing for fast approximate nearest-neighbor retrieval.

File Content OCR: Optical character recognition extracts text from uploaded images, PDFs, and scanned documents, making their content searchable alongside native project content.

How Search Works

Instant Results: Start typing and see results immediately with relevance ranking
Cross-Workspace: Search spans all projects, agents, automations, and files you have access to
Context-Aware: Results are weighted by recency, relevance, and your interaction patterns
Permission-Respecting: You only see results from content you have access to

Search Use Cases

Find Anything Fast:

• Project names, task titles, and content
• AI agent configurations and training data
• Automation workflow descriptions
• Uploaded documents and files

Semantic Questions:

• "What's our pricing strategy?" → Finds relevant strategy documents
• "Client onboarding process" → Surfaces related projects and templates
• "Last quarter revenue analysis" → Locates financial reports

File Discovery:

• Search inside uploaded PDFs and images
• Find specific data within spreadsheets
• Locate scanned documents by their content

AI Agent + Search Integration

AI agents leverage the search system to answer questions from your workspace knowledge:

Agent Knowledge Retrieval: When you ask an agent a question, it searches your workspace's full-text and semantic indexes to find relevant context before responding.

Continuous Learning: As you add more projects and content, agents automatically gain access to the expanded knowledge base through search.

Cross-Project Intelligence: Agents can synthesize information from multiple projects found through search to provide comprehensive answers.

Search Best Practices

Organize for Discovery: Use clear, descriptive titles for projects and tasks
Add Context: Write meaningful descriptions that help both humans and search find content
Upload Documents: Add relevant PDFs and images to make their content searchable
Use Tags: Consistent tagging improves both keyword and semantic search relevance

Related Wiki Pages: Workspaces, Projects, AI Agents