Seeker: DeepSeek-R1 Integration for VS Code

Seeker enables seamless interaction with DeepSeek-R1 reasoning models directly within Visual Studio Code, leveraging the power of Ollama's local runtime. Designed for privacy, performance, and usability, Seeker offers a modern AI-assisted coding experience with no internet connection required. Whether you’re working on resource-constrained hardware or high-end systems, Seeker supports a range of DeepSeek-R1 models to fit your needs. Published here.

Key Features

• Local Model Execution: Run DeepSeek-R1 models entirely on your local machine using Ollama runtime. No data ever leaves your environment.
• Model Flexibility: Supports all DeepSeek-R1 model sizes—from 1.5B to 671B parameters.
• [NEW] LLM Agnostic: Compatible with any model available through Ollama (Llama, DeepSeek, Gemma, Mistral, etc.)
• Privacy-First AI: Offline functionality ensures complete data privacy.
• Integrated Chat Interface: Engage with a clean, modern chat interface embedded directly into VS Code.
• Hardware-Based Performance: Select models based on your machine’s capabilities.
• Context-Aware Assistance: Enhance productivity with AI-driven insights tailored to your coding environment.

DeepSeek-R1 Model Overview

DeepSeek-R1 is DeepSeek’s first-generation reasoning model series, rivaling the performance of OpenAI-o1 across diverse tasks like mathematics, programming, and logical reasoning. The series includes distilled variants optimized for smaller hardware footprints, as well as massive server-grade models for maximum capability.

Available Models

Requirements

1. Ollama: Install Ollama from ollama.ai to enable local model execution.
2. Hardware: Ensure sufficient system resources based on your chosen model.
3. Visual Studio Code: Ensure you have the latest version installed.

Installation

Step 1: Install Seeker

• Download and install Seeker from the VS Code Marketplace or via the VSIX file.

Step 2: Set Up Ollama

• Download and install Ollama from ollama.ai.
• [Optional] Use the following command to download your preferred DeepSeek-R1 or any other model:

ollama run deepseek-r1:1.5b   # Smallest model
ollama run deepseek-r1:7b     # Recommended model

# Seeker supports all Open Source LLMs available on Ollama
ollama pull llama2:7b      # Meta's Llama 2
ollama pull mistral:7b     # Mistral AI
ollama pull gemma:7b       # Google's Gemma

Usage

1. Open Seeker in VS Code:

   • Launch Visual Studio Code.
   • Press Ctrl+Shift+P (Windows/Linux) or Cmd+Shift+P (Mac).
   • Select Seeker: Query.

2. Model Selection:

   • Default model: deepseek-r1:1.5b.
   • Change model via VS Code settings:
     • Open settings and search for “Seeker”.
     • Select from preset models or choose "custom"
     • For custom models, enter any Ollama model name (e.g., qwen:0.5b, mistral:7b).

3. Start Chatting:

   • Engage with chosen models in a modern, intuitive interface.
   • Real-time responses with streaming.
   • Keyboard shortcuts (Enter to send, Shift+Enter for newline)

Running Seeker

Known Issues

• First-Time Setup or Model Switching: Initial model downloads may take some time.
• Memory Usage: Ensure your system meets the requirements of your chosen model.
• Download Issues: If downloads fail or get interrupted, you may need to:
  1. Stop Ollama service (or quit the extension)
  2. Delete contents of ~\.ollama\models\blobs to remove partial downloads
  3. Restart Ollama service
  4. Try downloading again
• Manual Model Cleanup: To free up disk space, you can:
  1. Stop Ollama service (or quit the extension)
  2. Delete contents of ~\.ollama\models\blobs
  3. Restart Ollama service

Why Seeker?

Modern Interface

• Sleek sidebar integration with custom icon
• Clean chat UI with message timestamps
• Auto-resizing message input
• VS Code theme integration

Model Management

• Live download progress in status bar
• Automatic model switching
• Download size and progress tracking
• Custom model support

Privacy & Performance

• All processing done locally
• No internet connection required
• Efficient stream-based responses
• Automatic resource cleanup

License

Seeker is licensed under the MIT License, allowing for commercial use, modifications, and derivative works. For details, see the LICENSE file. Individual model licenses apply based on your chosen LLM.

DeepSeek-R1 models are similarly available for commercial use and modification, with additional licensing information for Qwen and Llama-derived models:

• Qwen Models: Based on Apache 2.0 License.
• Llama Models: Licensed under Llama3 series licenses.

Contributing

Contributions are welcome! If you have ideas, bug fixes, or feature requests, feel free to submit a Pull Request or open an issue.

Acknowledgments

• DeepSeek AI: For developing the cutting-edge DeepSeek-R1 models.
• Ollama: For providing a seamless local runtime.
• Model Creators : For the open-source models.

Experience privacy-focused AI assistance with any LLM, right in your editor with Seeker.