feat: Entity extraction + auto-edge suggestions

[ajianaz]

What

Wire /extract endpoint to auto-extract entities from memory content and suggest graph edges between related memories.

Why

Manual edge creation is tedious. Entity extraction can identify people, places, concepts, technologies in memories and suggest connections automatically — building the knowledge graph faster.

How

Endpoint

Endpoint	Method	Return
POST /extract	{text} -> {entities: [{name, type, confidence}]}	Existing

Implementation

Layer 0-2: Standard pattern.

Layer 3 — Integration points:

1. Memory Detail panel: [Extract Entities] button on each memory

   • Shows extracted entities as chips/badges
   • Each entity clickable -> search for other memories with same entity
   • [Create Edge] suggestion if related memory found

2. Batch extraction (optional): Settings -> [Extract All]

   • Runs extraction across all memories
   • Shows progress bar
   • Generates edge suggestions for review

3. Entity types: Color-coded chips

   • Person (blue), Organization (green), Technology (purple), Concept (orange), Location (teal)

Testing

• Extract entities from single memory
• Entities render as color-coded chips
• Click entity -> search related memories
• Edge suggestion flow works
• Batch extraction (if implemented) shows progress

Related Issues

• Part of: Phase 2B endpoint gap fill

Checklist

• Layer 0-3 for /extract
• Entity chips in Memory Detail
• Click-to-search per entity
• Edge suggestion flow