VS Code, Continue and Ollama: local coding setup

Abhishek explains how to connect VS Code to models running on the same machine through Ollama and the Continue extension. The tutorial presents an open source coding assistant setup that avoids external model API calls. His claims of unlimited tokens and performance similar to commercial tools are not established benchmarks.

The setup assigns separate models to chat, code completion and embeddings. The chat model handles questions and code generation, while the embedding model indexes repository files for search. After installing the official Continue extension, viewers select the local model option, download models through Ollama and connect them. Model choice depends on available compute; the tutorial does not establish a minimum hardware requirement.

For the demonstration, Abhishek selects GPT-OSS with 20 billion parameters and works with a cloned Argo CD repository. He requests a project overview, adds a selected Go snippet to chat for explanation, then submits it for editing. The resulting edit adds comments, which he can accept or reject. This illustrates the interaction rather than proving that the model can reliably fix bugs.

The final section covers separate chat sessions, MCP server configuration and model roles in the local configuration file. These controls extend the local LLM workflow, though the demonstrated MCP use can involve internet access.