Cline supports running models locally using Ollama. This approach offers privacy, offline access, and potentially reduced costs. It requires some initial setup and a sufficiently powerful computer. Because of the present state of consumer hardware, it’s not recommended to use Ollama with Cline as performance will likely be poor for average hardware configurations. Website: https://ollama.com/

​
Setting up Ollama

  1. Download and Install Ollama: Obtain the Ollama installer for your operating system from the Ollama website and follow their installation guide. Ensure Ollama is running. You can typically start it with: 
    ollama serve
  2. Download a Model: Ollama supports a wide variety of models. A list of available models can be found on the Ollama model library. Some models recommended for coding tasks include:
    • codellama:7b-code (a good, smaller starting point)
    • codellama:13b-code (offers better quality, larger size)
    • codellama:34b-code (provides even higher quality, very large)
    • qwen2.5-coder:32b
    • mistralai/Mistral-7B-Instruct-v0.1 (a solid general-purpose model)
    • deepseek-coder:6.7b-base (effective for coding)
    • llama3:8b-instruct-q5_1 (suitable for general tasks)
    To download a model, open your terminal and execute: 
    ollama pull <model_name>
    For instance: 
    ollama pull qwen2.5-coder:32b
  3. Configure the Model’s Context Window: By default, Ollama models often use a context window of 2048 tokens, which can be insufficient for many Cline requests. A minimum of 12,000 tokens is advisable for decent results, with 32,000 tokens being ideal. To adjust this, you’ll modify the model’s parameters and save it as a new version. First, load the model (using qwen2.5-coder:32b as an example): 
    ollama run qwen2.5-coder:32b
    Once the model is loaded within the Ollama interactive session, set the context size parameter: 
    /set parameter num_ctx 32768
    Then, save this configured model with a new name: 
    /save your_custom_model_name
    (Replace your_custom_model_name with a name of your choice.)
  4. Configure Cline:
    • Open the Cline sidebar (usually indicated by the Cline icon).
    • Click the settings gear icon (⚙️).
    • Select “ollama” as the API Provider.
    • Enter the Model name you saved in the previous step (e.g., your_custom_model_name).
    • (Optional) Adjust the base URL if Ollama is running on a different machine or port. The default is http://localhost:11434.
    • (Optional) Configure the Model context size in Cline’s Advanced settings. This helps Cline manage its context window effectively with your customized Ollama model.

​
Tips and Notes

  • Resource Demands: Running large language models locally can be demanding on system resources. Ensure your computer meets the requirements for your chosen model.
  • Model Choice: Experiment with various models to discover which best fits your specific tasks and preferences.
  • Offline Capability: After downloading a model, you can use Cline with that model even without an internet connection.
  • Token Usage Tracking: Cline tracks token usage for models accessed via Ollama, allowing you to monitor consumption.
  • Ollama’s Own Documentation: For more detailed information, consult the official Ollama documentation.