A high-performance fleet router for running the full suite of DeepSeek models across local hardware with zero API costs.
Install & Download
1. ClawHub CLI
The fastest way to install a skill directly from the registry.
npx clawhub@latest install deepseek-deepseek-coder
2. Manual Installation
Copy the skill folder to one of these locations
~/.openclaw/skills/ <project>/skills/ Priority: Workspace > Local > Bundled
3. Prompt Installation
Copy this prompt to OpenClaw to install it automatically.
Help me install deepseek-deepseek-coder using Clawhub. If Clawhub is not installed, install it first (npm i -g clawhub).
Prefer to download?
Get the raw skill files in a ZIP archive.
What is DeepSeek Local Fleet Manager?
This skill enables developers to harness the power of DeepSeek-V3, DeepSeek-R1, and DeepSeek-Coder on their own local infrastructure. By utilizing the Openclaw Skills ecosystem, users can eliminate per-token cloud costs and ensure data privacy while maintaining high availability. The system acts as a smart router, distributing requests to the most capable device in your local network, whether you are running a lightweight 1.5B reasoning model or a massive 671B MoE powerhouse.
The core value of this skill lies in its ability to transform a collection of computers into a unified AI compute cluster. It supports a wide range of DeepSeek variants, from coding-specific models to advanced reasoning engines, all while keeping your proprietary code and data strictly within your private network.
DeepSeek Local Fleet Manager Use Cases
- Running private, local coding assistants using DeepSeek-Coder-V2 without external data leaks.
- Executing complex reasoning tasks and mathematical proofs with DeepSeek-R1 without API rate limits.
- Deploying a cost-effective AI inference cluster across multiple home or office machines using Openclaw Skills.
- Developing applications that require high-throughput LLM access without recurring cloud expenses or subscription fees.
How DeepSeek Local Fleet Manager Works
- The user installs the herd package and initializes the central router on a primary machine to manage the Openclaw Skills traffic.
- Individual nodes on different machines are connected to the fleet, broadcasting their hardware capabilities and available VRAM.
- A request is sent to the local endpoint (port 11435) using an OpenAI-compatible SDK or standard HTTP clients.
- The 7-signal scoring algorithm evaluates all active nodes to find the optimal device for the specific DeepSeek model requested.
- The system automatically handles model loading, request routing, and provides transparent failover if a node becomes unresponsive.
DeepSeek Local Fleet Manager Setup
To get started with these Openclaw Skills, ensure you have Python installed and run the following commands:
pip install ollama-herd
# Start the central fleet router
herd
# On each machine in your fleet, start a node
herd-node
DeepSeek Local Fleet Manager Data Schema & Taxonomy
The skill manages its operations through a specific directory structure and database to ensure high performance:
| File/Path | Description |
|---|---|
~/.fleet-manager/latency.db |
SQLite database tracking node response times and performance metrics. |
~/.fleet-manager/logs/herd.jsonl |
Structured JSON logs for monitoring requests, model pulls, and fleet health. |
The system also supports request tagging, allowing users to categorize and track project-specific usage within the Openclaw Skills framework.
DeepSeek Local Fleet Manager Advanced Features
- 7-signal scoring for intelligent request distribution based on real-time latency and hardware load.
- VRAM-aware fallback mechanisms that route requests to already-loaded models to avoid cold-start delays.
- Context protection features that prevent expensive model reloads when context window parameters change.
- Full compatibility with the OpenAI SDK, allowing for easy integration into existing AI workflows and Openclaw Skills agents.
- Built-in dashboard at
http://localhost:11435/dashboardfor monitoring token throughput and hardware health.
SKILL.md
Loading
Related Openclaw Skills
Twitter Command Center (Search + Post + Interact)
An all-in-one Twitter intelligence and automation tool for autonomous agents to search, monitor, and interact with X data.
Spec-Kit
Spec-Kit is a framework for Spec-Driven Development that turns specifications into first-class citizens for AI agent workflows.
DeepSeek Local Fleet
A local fleet management skill for running high-performance DeepSeek models across multiple devices without cloud dependencies.
Ollama Herd
A high-performance multimodal router that unifies multiple local Ollama instances into a single, smart API endpoint.
City of Toronto Open Data
A specialized data extraction and analysis tool for accessing the City of Toronto's comprehensive open data portal.
dgent
A specialized utility to identify and remove AI tells like naming patterns, catch-rethrow blocks, and verbose vocabulary from code and commit messages.
Featured*
