Advanced Configuration
For power users and developers, Rephlo offers granular control over performance and AI behavior. These settings are distributed across various tabs in the Settings panel.
┌─────────────────────────────────────────────────────────────────────────┐
│ ADVANCED SETTINGS (LLM Providers Tab) │
├─────────────────────────────────────────────────────────────────────────┤
│ │
│ ┌─────────────────────────────────────────────────────────────────┐ │
│ │ [General] [Providers] [Hotkeys] [License] [Billing] [Usage] │ │
│ │ ─────────── │ │
│ │ [Privacy] [Account] [Environment] [About] │ │
│ │ │ │
│ │ PERFORMANCE │ │
│ │ ┌───────────────────────────────────────────────────────────┐ │ │
│ │ │ │ │ │
│ │ │ Compaction Threshold [4000 ] tokens │ │ │
│ │ │ │ │ │
│ │ │ Compaction Strategy [Balanced v] │ │ │
│ │ │ ( ) Lossless │ │ │
│ │ │ (*) Balanced │ │ │
│ │ │ ( ) Aggressive │ │ │
│ │ │ │ │ │
│ │ └───────────────────────────────────────────────────────────┘ │ │
│ │ │ │
│ │ AI TUNING │ │
│ │ ┌───────────────────────────────────────────────────────────┐ │ │
│ │ │ │ │ │
│ │ │ Temperature [0.0]────|────[0.7]────|────[1.0+] │ │ │
│ │ │ Precise Balanced Creative │ │ │
│ │ │ ^ │ │ │
│ │ │ │ │ │
│ │ │ Max Response Tokens [2000 ] │ │ │
│ │ │ │ │ │
│ │ │ Context Window Limit [Auto v] │ │ │
│ │ │ │ │ │
│ │ └───────────────────────────────────────────────────────────┘ │ │
│ │ │ │
│ │ DEVELOPER │ │
│ │ ┌───────────────────────────────────────────────────────────┐ │ │
│ │ │ │ │ │
│ │ │ Debug Mode [O=====] OFF │ │ │
│ │ │ Prompt Inspection [O=====] OFF │ │ │
│ │ │ │ │ │
│ │ │ Data Directory: C:\Users\...\AppData\Roaming\Rephlo │ │ │
│ │ │ [Clear Cache] [Export Logs] [Open Data Folder] │ │ │
│ │ │ │ │ │
│ │ └───────────────────────────────────────────────────────────┘ │ │
│ │ │ │
│ └─────────────────────────────────────────────────────────────────┘ │
│ │
└─────────────────────────────────────────────────────────────────────────┘
Performance & Memory
Space Compaction Strategy
- Threshold: Set the token count at which a Space triggers compaction (default: 4000 tokens).
- Aggressiveness:
- Lossless: Only removes exact duplicates.
- Balanced: Summarizes older documents.
- Aggressive: Keeps only the most recent key points (saves money/time, loses detail).
Cache Management
Rephlo caches processed Space content and API responses to improve performance.
- Clear Cache: Forces a refresh of cached data. Use this if content seems outdated after updating Space files.
- Disk Location: You can move the data folder to a different drive if you are low on C: drive space.
Provider Tuning
Temperature
(Available per-provider or per-command)
- 0.0 - 0.3: Precise, deterministic. Good for code and facts.
- 0.7: Balanced. Good for writing.
- 1.0+: Creative, random. Good for brainstorming.
Max Tokens
Limit the length of the AI's response to prevent it from writing essays when you only want a sentence.
Context Window Limit
Manually cap the amount of text sent to the AI.
- Why? To control costs (for paid APIs) or prevent "Out of Memory" errors (for local Ollama models).
Developer Tools
- Debug Mode: Enables verbose logging to
logs/debug.log. Useful for troubleshooting provider connection issues. - Prompt Inspection: Shows the exact raw JSON payload sent to the API provider in the History details.
Next Section: Put it all together with Tutorials.