Continue is an open-source AI coding assistant for VS Code and JetBrains. It supports chat, inline editing, autocomplete, and agentic workflows — all configurable against any OpenAI-compatible endpoint.
More information is available here: API Access
Continue is configured via ~/.continue/config.yaml. The easiest way to open it is from the Continue sidebar — click the gear icon, then Open Config.
Replace the contents with the following, substituting your API key:
name: Local Config
version: 1.0.0
schema: v1
models:
- name: UCSB Gemma 4 31B
provider: openai
model: "gemma-4-31b"
apiBase: "https://api.ai.college.ucsb.edu/v1"
apiKey: sk-YOURKEYHERE
capabilities:
- tool_use
roles:
- chat
- edit
- apply
- autocomplete
- name: UCSB Qwen 3 Coder Next
provider: openai
model: "Qwen3-Coder-Next"
apiBase: "https://api.ai.college.ucsb.edu/v1"
apiKey: sk-YOURKEYHERE
capabilities:
- tool_use
roles:
- chat
- edit
- apply
- autocomplete
- name: UCSB Granite 4.1 30B
provider: openai
model: "granite-4.1-30b"
apiBase: "https://api.ai.college.ucsb.edu/v1"
apiKey: sk-YOURKEYHERE
capabilities:
- tool_use
roles:
- chat
- edit
- apply
- autocomplete
- name: UCSB GPT-OSS 120B
provider: openai
model: "gpt-oss-120b"
apiBase: "https://api.ai.college.ucsb.edu/v1"
apiKey: sk-YOURKEYHERE
capabilities:
- tool_use
roles:
- chat
- edit
- apply
- autocomplete
The config above assigns roles to each model:
Ctrl+I / Cmd+I)The first model for a given role (chat, edit, etc.) as listed in the file is what will be set as the default for that capability. You can switch the active chat model from the model picker in the Continue panel but otherwise Continue will use the first model listed for chat to chat, the first model listed with the edit role, as the default to edit, and so forth.
If you wanted to set Qwen3 as the default for autocomplete and gemma4 as the default for chat, edit, apply, and a fallback for autocomplete, you might have this in your config.yaml:
models:
# ---- Autocomplete default (fast local) ----
- name: UCSB Qwen 3 Coder Next # fast FIM model
provider: openai
model: "Qwen3-Coder-Next"
apiBase: "https://api.ai.college.ucsb.edu/v1"
apiKey: sk-YOURKEYHERE
capabilities: [tool_use]
roles: [autocomplete]
# ---- Chat / edit / apply default (big model) ----
- name: UCSB Gemma 4 31B
provider: openai
model: "gemma-4-31b"
apiBase: "https://api.ai.college.ucsb.edu/v1"
apiKey: sk-YOURKEYHERE
capabilities: [tool_use]
roles: [chat, edit, apply, autocomplete] # autocomplete listed here too,
# but the *first* autocomplete entry
# (Qwen‑Coder‑Next) will be used.
@Codebase) indexes your project locally for RAG-style queries — no data leaves your machine for indexing, but queries are sent to the API.