The food intelligence core behind sNAKr and Takia
Cella is a deterministic, safety-first food intelligence engine designed to help households and communities reduce waste, prioritize safety, and plan with dignity.
It powers:
- sNAKr → shared household inventory, spoilage prevention, and meal planning
- Takia → community kitchens, crisis readiness, and surplus routing
Cella is not a chatbot.
Cella is not a medical tool.
Cella is the shared food brain both systems rely on.
Cella provides a rules-first intelligence layer for food systems that need to operate:
- Safely
- Clearly
- Calmly
- Under both everyday and crisis conditions
It focuses on spoilage risk, inventory awareness, and decision support, without drifting into health diagnostics, medical claims, or biotech complexity.
Spillage and safety perception
ShelfSense evaluates the spoilage and safety risk of stored food items using deterministic rules.
- Category-based
- Time-aware
- Storage-aware
- Fully explainable
Outputs clear risk levels:
LOW · MEDIUM · HIGH · CRITICAL
Every assessment includes a human-readable explanation describing why an item is considered safe or risky.
Inventory completeness and usability
StockView understands what exists in a kitchen or storage system and how usable it is.
It answers:
- What do we have?
- What are we missing?
- What should be used first?
StockView detects:
- Missing staples
- Low quantities
- Duplicates
- Items nearing spoilage
The internal rulebook powering Cella
SafeKeep is the deterministic foundation beneath all spoilage logic.
It defines:
- Expected shelf-life per food category
- Storage-specific safety windows
- Opened vs unopened behavior
SafeKeep is:
- Internal only
- Rules-based
- Not overridable by LLMs
ShelfSense relies entirely on SafeKeep as its ground truth.
Cella is strict by design.
- Deterministic rules always override LLM output
- Risk levels cannot be softened or reinterpreted
- Unsafe food handling suggestions are blocked
- Cella fails safe when uncertain
- All critical outputs include explanation strings
Cella operates near food safety and nutrition, but makes no medical claims.
Cella supports multiple AI personalities without fragmenting intelligence.
-
Chuma 🤖 (Takia)
- Calm
- Crisis-aware
- Dignity-first
-
Fasoolya 🦝 (sNAKr)
- Playful
- Household-focused
- Supportive
Both rely on the same Cella outputs and safety rules.
They differ only in tone and context.
Cella uses a unified CellaLLM wrapper so providers can be swapped without changing application logic.
Supported providers include:
- OpenAI
- Anthropic
- Google Gemini
- DeepSeek
- Perplexity
- Local models via provider interface
LLMs are used for explanations and planning, never as safety authorities.
- ShelfSense
- StockView
- SafeKeep
- Basic LLM explanations
- Meal planning for soon-to-spoil items (sNAKr)
- Kitchen priority summaries (Takia)
- Budget-aware restock logic
- Multi-step planning agent
- Nutrition summaries (non-medical)
- Scenario simulation
- Offline-friendly inference for crisis environments
Cella strengthens food systems through:
- Clarity
- Simplicity
- Transparency
- Dignity
- Resilience
It is built for real kitchens, real households, and real communities.
Cella is under active development as the shared intelligence core for sNAKr and Takia.
Specifications and design documents live in /docs.