Add semantic vector search via Qdrant + Ollama

[renezander030]

Problem

The current tasks search command fires N+1 API calls on every query (1 to list projects, then 1 per project to fetch and filter tasks). With hundreds of tasks across many projects, this takes 3-5 seconds per search. Worse, substring matching misses semantically related results: searching "deployment" won't find a task titled "push release to prod".

Solution

Add optional vector search using Qdrant (vector database) and Ollama (local embeddings with nomic-embed-text). Both run locally, no external API calls, no data leaves the machine.

How it works

1. vector-sync fetches all active tasks, embeds them via Ollama, and stores vectors in Qdrant
2. Incremental sync uses MD5 content hashing (title|content|tags) to skip unchanged tasks
3. Metadata-only changes (priority, dueDate) update the Qdrant payload without re-embedding
4. semantic queries embed the search string and match by cosine similarity in <100ms
5. similar finds semantically related tasks for deduplication or discovering related work
6. If Qdrant/Ollama are down, semantic search falls back to keyword search automatically

What I changed

• lib/vector.js (new): Qdrant + Ollama integration, incremental sync, search, similar tasks, health checks
• lib/tasks.js: Added semanticSearch, findSimilar, vectorSync, vectorStatus with keyword fallback
• bin/ticktick.js: New subcommands: semantic, similar, vector-sync, vector-status
• lib/mcp.js: 4 new MCP tools for AI assistant integration
• lib/cli.js: Score-based result formatting, updated help text
• README.md: Full documentation with prerequisites, usage, cron setup

Production context

I've been running this architecture (same stack: Qdrant + Ollama + nomic-embed-text, same incremental sync pattern) on my own TickTick MCP server for several months with ~500 tasks. The speed difference is significant, and the semantic matching catches things keyword search never will.

New CLI commands

ticktick tasks vector-sync                           # Build/update the index
ticktick tasks semantic "anything about deployments"  # Semantic search
ticktick tasks similar TASK_ID                        # Find similar tasks
ticktick tasks vector-status                          # Index health check

New MCP tools

Tool	Description
ticktick_tasks_semantic_search	Semantic search via vector similarity
ticktick_tasks_similar	Find semantically similar tasks
ticktick_vector_sync	Sync tasks into vector index
ticktick_vector_status	Check vector index health

Prerequisites (optional)

Vector search is entirely optional. If Qdrant/Ollama aren't running, everything else works exactly as before.

docker run -d --name qdrant -p 6333:6333 qdrant/qdrant
docker run -d --name ollama -p 11434:11434 ollama/ollama
docker exec ollama ollama pull nomic-embed-text

Test plan

• Existing tests pass (5 pre-existing failures unrelated to this change)
• ticktick tasks vector-sync indexes tasks from the API
• ticktick tasks semantic "query" returns scored results
• ticktick tasks similar TASK_ID finds related tasks
• ticktick tasks vector-status reports index health
• Semantic search falls back to keyword when Qdrant is down
• MCP tools appear and work in Claude Desktop/Code