Vector Vault
Build AI agents that think, remember, and act. Start free on your machine. Scale to production with zero infrastructure.
Start Free, Scale When Ready
Vector Vault gives you two ways to build:
🏠 Local Mode (Free) — Run entirely on your machine. No account needed. No limits. Perfect for learning, prototyping, and projects where your data stays local.
☁️ Cloud Platform — When you're ready for production, deploy to our Persistent Agentic Runtime (PAR). Sub-second responses, 99.9% uptime, visual workflow builder, and agents that can pause for days and resume instantly.
pip install vector-vault
Quick Start (Local Mode)
No signup. No API keys (except OpenAI for embeddings). Just code.
from vectorvault import Vault
# Create a local vault
vault = Vault(
vault='my_knowledge_base',
openai_key='YOUR_OPENAI_KEY',
local=True # Everything stays on your machine
)
# Add your data
vault.add("The mitochondria is the powerhouse of the cell")
vault.add("Neural networks are inspired by biological brains")
vault.add("Vector databases enable semantic search")
vault.get_vectors()
vault.save()
# Search by meaning, not keywords
results = vault.get_similar("How do AI systems learn?")
# → Returns: "Neural networks are inspired by biological brains"
# Or chat with your data
response = vault.get_chat(
"What powers the cell?",
get_context=True # Automatically retrieves relevant context
)
What You Can Build
RAG Applications
Give any LLM access to your knowledge base with automatic context retrieval.
response = vault.get_chat(
"How do I configure authentication?",
get_context=True,
n_context=5
)
Semantic Search
Find content by meaning. Search "budget issues" and find documents about "financial constraints."
results = vault.get_similar("budget issues", n=10)
AI Memory Systems
Give your agents persistent memory across conversations.
# Store conversation
vault.add(f"User asked about {topic}. Agent responded with {response}")
vault.get_vectors()
vault.save()
# Later, retrieve relevant context
context = vault.get_similar(new_user_message)
Document Q&A
Turn any document collection into a question-answering system.
# Load documents
for doc in documents:
vault.add(doc.text, meta={'source': doc.filename})
vault.get_vectors()
vault.save()
# Answer questions
answer = vault.get_chat("What's the refund policy?", get_context=True)
Going to Production
When you're ready to scale, Vector Vault Cloud provides:
Persistent Agentic Runtime (PAR)
Agents that pause for days, branch into parallel tasks, and resume instantly — without you managing servers.
Vector Flow
Design agent workflows visually with drag-and-drop. Branching logic, approvals, integrations, all in the browser.
Production Performance
- Sub-second streaming responses
- 99.9% uptime SLA
- Auto-scaling to thousands of concurrent conversations
Enterprise Ready
- SOC 2 compliant infrastructure
- Team collaboration
- Usage-based pricing
# Switch to cloud mode
vault = Vault(
user='you@company.com',
api_key='YOUR_VECTORVAULT_KEY',
openai_key='YOUR_OPENAI_KEY',
vault='production_kb'
)
# Same API, production infrastructure
response = vault.get_chat("Customer question here", get_context=True)
# Or run visual workflows
result = vault.run_flow('customer_support_agent', user_message="...")
Get started at vectorvault.io →
Core API
Initialization
# Local mode (free, no account)
vault = Vault(
vault='vault_name',
openai_key='sk-...',
local=True
)
# Cloud mode (production)
vault = Vault(
user='email',
api_key='vv_...',
openai_key='sk-...',
vault='vault_name'
)
Essential Methods
| Method | Description |
|---|---|
add(text, meta=None) |
Add text to the vault |
get_vectors() |
Generate embeddings |
save() |
Persist to storage |
get_similar(text, n=4) |
Semantic search |
get_chat(text, get_context=True) |
RAG chat |
get_items(ids) |
Retrieve by ID |
edit_item(id, text) |
Update item |
delete_items(ids) |
Remove items |
Convenience
# Add + embed + save in one call
vault.add_n_save("Your text here")
# Stream responses
for chunk in vault.get_chat_stream("Your question"):
print(chunk, end='')
How It Works
Vector Vault uses FAISS (Facebook AI Similarity Search) for fast, accurate vector operations:
- Add your text data
- Embed using OpenAI's embedding models
- Search by semantic similarity
- Chat with automatic context retrieval
Local mode stores everything in ~/.vectorvault/. Cloud mode syncs to our managed infrastructure.
Requirements
- Python 3.8+
- OpenAI API key (for embeddings)
Resources
- Website: vectorvault.io
- Vector Flow: app.vectorvault.io/vector-flow
- Full API Docs: Documentation
- Discord: Join the community
- JavaScript SDK: VectorVault-js
Contributing
git clone https://github.com/John-Rood/VectorVault.git
cd VectorVault
pip install -e .
# Run tests
cd VectorVault-Testing
python run_tests.py # Cloud tests
python test_local_mode.py # Local tests
License
MIT License
Start free. Scale infinitely. vectorvault.io
Model-compatible thinking levels
VectorVault ships a canonical model capability table and validates each selection before a provider call. Existing callers can omit the setting to preserve provider defaults.
from vectorvault import Vault, get_allowed_thinking_levels
print(get_allowed_thinking_levels("gpt-5.6"))
response = vault.get_chat(
"Compare these options carefully",
model="gpt-5.6",
thinking_level="high",
)
1 maintainer
Download files
Download the file for your platform. If you're not sure which to choose, learn more about installing packages.
Source Distribution
Built Distribution
Filter files by name, interpreter, ABI, and platform.
If you're not sure about the file name format, learn more about wheel file names.
Copy a direct link to the current filters
File details
Details for the file vector_vault-7.4.9.13.tar.gz.
File metadata
- Download URL: vector_vault-7.4.9.13.tar.gz
- Upload date:
- Size: 61.6 kB
- Tags: Source
- Uploaded using Trusted Publishing? No
- Uploaded via: twine/6.1.0 CPython/3.10.16
File hashes
| Algorithm | Hash digest | |
|---|---|---|
| SHA256 |
af8ef532abf2e6a1ac34c57c78e1f219064e9d1bf017534d7734d25101128c11
|
|
| MD5 |
71ceb45953cd7c330516d9ff55f38311
|
|
| BLAKE2b-256 |
0a9dc3d4903931fc80872f135e81cd3bc02b738278465321e9cf847e09e74006
|
File details
Details for the file vector_vault-7.4.9.13-py3-none-any.whl.
File metadata
- Download URL: vector_vault-7.4.9.13-py3-none-any.whl
- Upload date:
- Size: 64.2 kB
- Tags: Python 3
- Uploaded using Trusted Publishing? No
- Uploaded via: twine/6.1.0 CPython/3.10.16
File hashes
| Algorithm | Hash digest | |
|---|---|---|
| SHA256 |
1b861de1d02ef658906cc03be52e0ef1e51fb791b58afbd0b162e8298f23470a
|
|
| MD5 |
dac6b9671bbe8898aabda2f3b81ed610
|
|
| BLAKE2b-256 |
f955a1cc412ab971db1fc609ed9f5ccd27f77a960172dcb5c9a9aaa4bb5ee707
|
