Thanks for asking! Here's how the extended context support works:
Context Management Features:
• Dynamic File Selection - Instead of loading entire files, we intelligently select relevant code sections based on the query
• Automatic Conversation Summarization - When approaching context limits, conversations are automatically summarized to preserve essential information while staying within bounds
• Provider-Specific Optimization - Automatically adjusts to each model's context window: Claude (200K), GPT-4 (128K), Gemini (2M tokens)
• ENFIY.md Memory Files - Hierarchical context files (global, project, component-level) that provide persistent context across sessions
How it works in practice:
# Even with massive codebases
enfiy "find performance issues in this 1000-file React poject"
# → Analyzes only relevant files, dynamically loading what's needed
The system monitors token usage and triggers summarization when approaching 95% of the model's limit. For models like Gemini with 2M context, you can process entire large projects. The key is leveraging each provider's strengths while working within their constraints.
Looks nice - How does the extended context support work?
Thanks for asking! Here's how the extended context support works:
Context Management Features:
• Dynamic File Selection - Instead of loading entire files, we intelligently select relevant code sections based on the query
• Automatic Conversation Summarization - When approaching context limits, conversations are automatically summarized to preserve essential information while staying within bounds
• Provider-Specific Optimization - Automatically adjusts to each model's context window: Claude (200K), GPT-4 (128K), Gemini (2M tokens)
• ENFIY.md Memory Files - Hierarchical context files (global, project, component-level) that provide persistent context across sessions
How it works in practice:
# Even with massive codebases
enfiy "find performance issues in this 1000-file React poject"
# → Analyzes only relevant files, dynamically loading what's needed
The system monitors token usage and triggers summarization when approaching 95% of the model's limit. For models like Gemini with 2M context, you can process entire large projects. The key is leveraging each provider's strengths while working within their constraints.
More technical details in our GitHub README!