You’ll need a Nebula API key before starting. Get one at
trynebula.ai
Using Claude Code, Cursor, or Codex? Run npx skills add nebula-agi/skills to integrate Nebula automatically.
Prerequisites
1. Install
# Python
pip install nebula-client
# Node.js
npm install @nebula-ai/sdk
2. Initialize Client
from nebula import Nebula
# Uses NEBULA_API_KEY environment variable
nebula = Nebula()
3. Store & Search
from nebula import Nebula, Memory
nebula = Nebula()
# Create a collection
collection = nebula.create_collection(name="my_notes", description="Personal notes")
# Store a memory
memory_id = nebula.store_memory(
Memory(
collection_id=collection.id,
content="Machine learning is transforming healthcare",
metadata={"topic": "AI", "importance": "high"}
)
)
# Search memories
results = nebula.search(
query="machine learning healthcare",
collection_ids=[collection.id]
)
# Results contain semantics, procedures, episodes, and sources
for fact in results.semantics:
print(f"{fact['subject']} → {fact['predicate']} → {fact['value']}")
for source in results.sources:
print(source['text'])
What You’ll Get Back
When you store a memory, Nebula returns the memory ID:
memory_id = "eng_abc123def456"
When you search, Nebula returns a MemoryResponse with structured memory:
{
"semantics": [
{
"category": "fact",
"subject": "Machine Learning",
"predicate": "is transforming",
"value": "healthcare",
"activation_score": 0.92
}
],
"procedures": [],
"episodes": [],
"sources": [
{
"text": "Machine learning is transforming healthcare",
"activation_score": 0.91
}
]
}
- Semantics: Subject-predicate-value assertions (facts, inferences, tasks)
- Procedures: User preferences and behavioral patterns
- Episodes: Temporally clustered events
- Sources: The original source text that grounds each assertion
See
Core Concepts for a deeper explanation of the memory response.
Next Steps