Project: AI Companion Suite
Philosophy: Seamless, anticipatory AI interactions
- Hybrid Agentic Logic: Lightweight models make decisions before engaging heavy resources
- Auto-Detection: Minimal explicit commands - system anticipates needs
- Robust Infrastructure: Smart polling, retry logic, and cold-boot handling
| Component |
Type |
Version |
Description |
| Auto Image |
Filter |
5.7.0 |
Auto-detect image requests, smart prompt merging, spontaneous selfies |
| Image Button |
Action |
5.4.0 |
Manual image generation with same robust logic as Auto Image |
Key Features:
- NanoGPT prompt enhancement with outfit/location override logic
- RunPod ComfyUI integration with smart cold-boot handling
- Truly random seeds, local image storage
| Component |
Type |
Version |
Description |
| Auto Web Search |
Filter |
5.6.0 |
Detects factual queries, syntax-highlighted "hacker terminal" logs |
| Vision Proxy |
Filter |
1.1.1 |
Enables vision for non-multimodal models via image-to-text |
| Component |
Type |
Version |
Description |
| Auto Memory |
Filter |
1.1.0-alpha1 |
Intelligent memory extraction with consolidation rules, async task handling |
Key Features:
- Automatic extraction of valuable long-term information
- Smart consolidation to prevent duplicates and conflicts
- Cross-message context linking
| Component |
Type |
Version |
Description |
| Antigravity Pipe |
Manifold |
0.3.5 |
Bridges to Google Cloud Code models (Gemini 3, Claude 4.5) with load balancing |
| NanoGPT Integration |
Manifold |
3.2.1 |
Subscription-aware model fetching with proper timeout handling |
| Usage Dashboard |
Filter |
2.0.0 |
Visual API usage tracking for NanoGPT and RunPod with cyberpunk styling |
- Version Control: All functions must update version numbers in file headers when changed
- Naming Convention: Use descriptive version tags (e.g.,
5.7.0, 1.1.0-alpha1)
- Error Handling: Implement retry logic and graceful degradation