Context
Context is the information the AI agent uses to understand your codebase and task. Limyrx Studio IDE provides a dedicated interface for viewing and managing context in real time.
How Context Works
When you start a session, the agent collects context from multiple sources:
- Project files - the file tree and its structure
- OpenCode config - rules, instructions, and preferences
- Terminal output - command results and logs
- Your prompt - the task description you provide
- Session history - previous turns in the current session
The Context Panel
Open the context panel from the sidebar to see what the agent is currently reading. Files are listed with their relevance score. You can:
- Pin files to keep them in context
- Exclude files you don't want the agent to see
- Refresh context when files change externally
Context Configuration
In your .opencode.jsonc, you can configure:
{
"context": {
"maxTokens": 64000,
"includePatterns": ["src/**/*.ts", "README.md"],
"excludePatterns": ["node_modules", "dist"],
"autoContext": true
}
}
Relevance Scoring
The agent scores each file for relevance to your current task. Files with higher scores stay in context longer. This helps manage token usage by prioritizing the most important information.
See Also
- Projects - project-level settings
- Providers, Models & Agents - model context limits
- Usage & Quotas - token tracking