Feature Flags & Extensibility
Relevant source files
The following files were used as context for generating this wiki page:
The Feature Flags & Extensibility system provides a governed framework for managing feature gates, experimental capabilities, and operational overrides within the Isabella AI platform. It ensures that new functionalities are deployed with risk assessment, environment restrictions, and auditability.
The system integrates directly with the platform's API to provide real-time configuration snapshots and supports complex routing logic, such as fallback models and quantum execution routes. This framework prevents uncontrolled feature releases and enables self-healing operations through kill switches.
Feature Flag Architecture
The architecture relies on a centralized catalog that serves as the source of truth for all feature definitions. A dedicated service evaluates these definitions based on the current execution environment and user context.
The diagram above represents the flow of feature flag definitions from the static catalog to active system components.
Sources: src/config/feature-flags/catalog.ts, src/platform/flags/feature-flags.ts
Core Components
- Feature Flag Catalog: Defines keys, risk levels, and ownership for every gate.
- Feature Flag Service: Processes snapshots and evaluates active states based on
NODE_ENV. - API Endpoint: Exposes the current flag state via
/api/v1/flagsfor external consumption.
Sources: src/config/feature-flags/catalog.ts:1-25, server.ts:511-517
Governance and Risk Assessment
Each feature flag includes metadata to enforce governance. The system classifies flags by type—such as release, experiment, or kill_switch—and assigns a risk level from low to critical.
Feature Flag Types
| Type | Description |
|---|---|
| release | Progressive enrutamiento for new system modules. |
| experiment | A/B testing or prototype features like new terminal canvases. |
| operational | Critical functional gates like AI model fallback mechanisms. |
| kill_switch | Emergency overrides to restrict system behavior during attacks. |
| Sources: src/config/feature-flags/catalog.ts:21-25, src/config/feature-flags/catalog.ts:40-100 |
Risk and Audit Requirements
Flags marked as high or critical risk require explicit approval and audit logging. For example, isabella_ai_model_fallback is classified as critical because it alters core inference behavior during failures.
Sources: src/config/feature-flags/catalog.ts:54-64, governance/system-cards/isabella-5-3-0.md:30-40
Extensibility Through Modularity
The platform extends its cognitive capabilities using a 12-module architecture. The isabella-inference-engine uses feature flag logic and weighted intent scoring to route requests to specific modules like ISA (empathy), SOPHIA (logic), or ORION (execution).
This diagram shows how the system routes requests across cognitive modules based on detected intent and weighted logic.
Sources: src/lib/isabella-inference-engine.ts:850-950
Dynamic Routing Logic
- Intent Matching: The engine tokenizes input to find the best intent match using weighted scoring.
- Module Weighting: The system adjusts weights based on the active preset or security needs (e.g., boosting ARGUS for legal queries).
- Telemetry Generation: Real-time metrics are generated for safety, resonance, and reasoning depth.
Sources: src/lib/isabella-inference-engine.ts:790-840, src/lib/isabella-inference-engine.ts:900-940
Operational Control and Fallbacks
The system utilizes operational flags to manage resilience. The model fallback flag enables the platform to switch between the sovereign inference engine and secondary models like Gemini during timeouts.
The sequence shows the fallback mechanism triggered by operational feature flags during inference failures.
Sources: server.ts:840-900, src/config/feature-flags/catalog.ts:54-64
Snapshot and Environment Constraints
The featureFlagService provides snapshots constrained by the environment. Certain flags, like the isabella_quantum_route, are restricted to development and staging environments to prevent unauthorized access to experimental hardware in production.
Sources: src/platform/flags/feature-flags.ts:1-20, src/config/feature-flags/catalog.ts:88-98
Summary
The Feature Flags & Extensibility system ensures Isabella AI remains modular and resilient. By combining a governed flag catalog with a tri-modal cognitive core, the platform allows for safe experimentation and automatic fallback procedures while maintaining strict audit trails and risk management.