VecminDB SDK
The official SDK for VecminDB — The Sovereign Memory OS for AI Agents.
Stop letting your AI Agents hallucinate from memory rot. VecminDB naturally decays outdated memories, distills knowledge via PCA, and provides 100% Air-Gapped cryptographic data sovereignty.
⚠️ License Note: VecminDB is a commercial Cognitive Vector Database. The Free Tier supports up to 5 agents and 100K vectors/agent forever. For enterprise scale-out or clusters, please visit our official website to register and obtain a license key: https://lingxinmind.com.
Deployment & Installation
VecminDB can be run via Docker or as optimized, standalone pre-compiled native binary packages. No local compilers, dependencies, or Python runtimes are needed.
Method A: Docker Deployment (All Platforms - Windows, macOS, Linux)
The fastest way to spin up VecminDB with automatic in-database bilingual embedding support.
# For Global / Overseas users:
docker run -d --name vecmindb-trial -p 5520:5520 ghcr.io/lingxinmind/vecmindb:latest
# For Domestic users (China Aliyun Mirror):
# docker run -d --name vecmindb-trial -p 5520:5520 crpi-ngtfnt7d3tsnwk7l.cn-shanghai.personal.cr.aliyuncs.com/vecmindb/vecmindb:latest
Method B: Pre-Compiled Native Binary Bundles (Zero-Docker / Zero-Python)
Ideal for high-performance, air-gapped on-premise or private cloud servers. Download the appropriate package from our official website Downloads portal:
- Windows (AMD64):
Download
vecmindb-1.0.0-beta-x86_64-pc-windows-msvc.zip. Extract the ZIP archive, open Command Prompt or PowerShell in the directory, and run:.\vecmindb-server.exe
- macOS (Apple Silicon M1/M2/M3):
Download
vecmindb-1.0.0-beta-aarch64-apple-darwin.tar.gz. Open Terminal, extract and run:tar -xzf vecmindb-1.0.0-beta-aarch64-apple-darwin.tar.gz cd vecmindb-1.0.0-beta-aarch64-apple-darwin ./vecmindb-server
- Linux (AMD64):
Download
vecmindb-offline-linux-amd64.tar.gz. Extract and run:tar -xzf vecmindb-offline-linux-amd64.tar.gz cd vecmindb-offline-linux-amd64 ./vecmindb-server
SDK Quickstart
First, install the target client SDK:
# Install core client
pip install vecmindb
# Install with LangChain integration
pip install vecmindb[langchain]
# Install with CrewAI integration
pip install vecmindb[crewai]
Using with LangChain
from vecmindb.memory_plugin import VecminDBMemoryPlugin
from langchain_openai import ChatOpenAI
from langchain.chains import ConversationChain
# Initialize Sovereign Agent Memory
memory = VecminDBMemoryPlugin.for_langchain(agent_id="support_agent_01", base_url="http://localhost:5520")
llm = ChatOpenAI(temperature=0)
conversation = ConversationChain(llm=llm, memory=memory)
conversation.predict(input="Hi, I need help with my billing.")
Using with CrewAI
from vecmindb.memory_plugin import VecminDBMemoryPlugin
from crewai import Agent, Crew
# Initialize Sovereign Agent Memory
memory_storage = VecminDBMemoryPlugin.for_crewai(agent_id="finance_agent_01", base_url="http://localhost:5520")
agent = Agent(
role='Financial Analyst',
goal='Analyze billing data',
backstory='An expert in financial data.',
memory=True,
memory_config={"storage": memory_storage} # Inject VecminDB memory
)
Why VecminDB?
Unlike generic vector databases that act as static drives, VecminDB acts as a cognitive memory operating system with native lifecycle management and cryptographic isolation:
- Biological Forgetting (LTSM): Episodic memories decay dynamically following $W(t) = \exp(-\lambda \times \Delta t)$ with automatic 90-day semantic pruning (
let semantic_prune_threshold_secs = 90 * 86400;on disk). Frequently accessed memories persist; transient noise is permanently retired. - Welford & PCA Memory Distillation: Fuses decaying memory clusters into stable Abstract Centroids using real-time Welford online variance and DP-Federated PCA. Storage converges and scales logarithmically, locking in long-term TCO budgets.
- 3-Sigma Sentinel Guard: Performs real-time anomaly detection and adversarial injection pruning. Evaluates cosine outlier distance with dynamic cutoffs: $\text{Threshold} = \max(\text{Mean}_s - 3 \times \text{Std}_s, 0.7)$.
- Sovereign Federation: Shares collective intelligence across multiple agent domains or VPCs without raw data leak. Fuses PCA Candidate Centroids with differential privacy and a 10% principal bias: $\vec{v}{\text{centroid}} = \text{Mean}{\text{global}} + P_0 \times 0.1$.
- Raft Consensus & 1024-Bucket Anti-Entropy: Combines strong consensus replication with self-healing topology. Employs monotonic lock validation (
pub fencing_token: u64) and an adaptive sync cap:(resolution * 2).min(1024). - 100% Air-Gapped Single-Binary: Built-in BGE-M3 ONNX runtime. No Python, PyTorch, or external embedding API keys needed. Bounded tightly to machine-level HAI hardware fingerprints.
Enterprise Licensing: For multi-node SOC-2 compliant deployments, please purchase subscriptions or contact us at sulingqi@hotmail.com.
Metadata
Release files for vecmindb 1.0.1
For a detailed explanation of source distributions (sdists) and built distributions (wheels), please see the package formats documentation.
Source distribution (sdist)
| File | Size | Uploaded | Provenance | Actions |
|---|---|---|---|---|
| vecmindb-1.0.1.tar.gz | 41.7 kB | Without provenance | Details for vecmindb-1.0.1.tar.gz |
Built distribution (wheel)
| File | Interpreter | ABI | Platform | Actions Reset |
|---|---|---|---|---|
| vecmindb-1.0.1-py3-none-any.whl | Python 3 | none | any | Details for vecmindb-1.0.1-py3-none-any.whl |
Total release size: 83.5 kB
Release files / vecmindb-1.0.1.tar.gz
| Download URL | vecmindb-1.0.1.tar.gz |
|---|---|
| Size | 41.7 kB |
| Tags | Source |
|
SHA-256 checksum How to use checksums |
0c31ff95e0db40f95aaaa832fa81971b66b94069b6b46e61ddfd6f8f814ad242
|
|
BLAKE2b-256 checksum How to use checksums |
b0f006b0e5c13c8676b212350c8eeb1257695c39e27de3a9724e621ea0dcc6f3
|
| Upload date | |
|
Uploaded using Trusted Publishing? What is trusted publishing? |
No |
| Uploaded via |
twine/6.2.0 CPython/3.10.12
|
Release files / vecmindb-1.0.1-py3-none-any.whl
| Download URL | vecmindb-1.0.1-py3-none-any.whl |
|---|---|
| Size | 41.7 kB |
| Tags | Python 3 |
|
SHA-256 checksum How to use checksums |
3413f4978a2fd16ac473018260f6467728ffbd97e8398cca860a359670a8198a
|
|
BLAKE2b-256 checksum How to use checksums |
4a4a31f3856b2446500c0bdd6c343b9ce81dfeda54f99f2d15dc92bec22e140b
|
| Upload date | |
|
Uploaded using Trusted Publishing? What is trusted publishing? |
No |
| Uploaded via |
twine/6.2.0 CPython/3.10.12
|
Security for vecmindb 1.0.1
Provenance
Provenance describes where a file came from. On PyPI, provenance is shared via attestations, which provide a verifiable record of the build or publishing details. View details, limitations and caveats.
No provenance found
None of the files in this release were uploaded with provenance metadata.
A note on package security
PyPI is a repository for distributing packages, not a service for auditing them. While third-party security researchers and our staff do their best to detect and remove malicious packages, PyPI cannot guarantee the quality, security or suitability of any given package.
To learn more about protecting yourself from malicious actors, visit our package security documentation.