v2.4
Memory
Long Memory Overview
Persistent, searchable conversation memory with summarization and heat scoring.
Long Memory provides persistent, semantically searchable conversation memory for your AI agents. Unlike short-term session buffers, long memory stores and indexes conversations in a vector database, enabling your agents to recall information across days, weeks, or months.
How It Works
When your agent stores a memory, the content is embedded and indexed in Qdrant for semantic search. PostgreSQL tracks metadata, heat scores, and compliance settings. An optional summarization pipeline consolidates related memories to reduce storage and improve search quality.
Key Concepts
| Concept | Description |
|---|---|
| Semantic Search | Memories are retrieved by meaning, not keywords — find related information even when wording differs |
| Heat Scoring | Relevance score that decays over time and boosts on access — keeps active memories prominent |
| Conflict Resolution | Strategies for handling semantically similar memories (newer wins, keep both, confidence-based) |
| Auto-Summarization | Periodic consolidation of related memories into compact summaries — reduces storage and noise |
| PII Redaction | Automatic detection and masking of personally identifiable information before storage |
| Permanent Facts | Critical information marked as immune to heat decay — preserved indefinitely |
Quick Start
# Store a new memory
curl -X POST https://api.guidedmind.ai/api/v1/memory/long/store \
-H "Content-Type: application/json" \
-H "X-Memory-Api-Key: mk_your_key_here" \
-d '{
"user_id": "user-456",
"role": "user",
"content": "I prefer the Professional plan at $199/month",
"metadata": {"category": "pricing_preference"}
}'
# Search memories semantically
curl -X POST https://api.guidedmind.ai/api/v1/memory/long/search \
-H "Content-Type: application/json" \
-H "X-Memory-Api-Key: mk_your_key_here" \
-d '{
"query": "What plan does the user prefer?",
"external_user_id": "user-456",
"limit": 10,
"threshold": 0.7
}'When to Use Long Memory
- Personalized assistants — Remember user preferences, history, and patterns across sessions
- Customer support — Recall past interactions and resolutions for consistent service
- Knowledge accumulation — Build a growing memory base that improves over time
- Compliance-aware storage — Automatic PII redaction and configurable retention policies
Next Steps
- Basic Settings — Configure project name and enable/disable
- Storage — Set TTL and conflict resolution strategy
- Summarization — Configure automatic memory consolidation
- Compliance — Set up PII redaction and heat scoring
- API Setup — Generate API keys and review endpoints