How dani12po/zevora works
ZEVORA is a hybrid AI coding workspace: users submit chat or coding tasks through its browser UI, the gateway coordinates agent execution, local context and adaptive model routing, and the selected provider returns a response that may include workspace changes and verification. The README documents cache, memory, project context, evolution, provider management and approval-gated tools as additional capabilities. The excerpts show selected internals; subsystem links not established by those excerpts are based on the README’s documented workflow.
Main entry points: execution.py Agent Execution · main.py API Gateway · app.js Web UI · hybrid_router.py Adaptive Router
Components 22
Workspace Experience
- Web UI
static/app.js - API Gateway
main.py - Gateway Controller
zevora/cli.py
Agent Workflow
- Agent Execution
agent/core/execution.py - Task Planning
agent/core/planning.py - Workflow Journal
main.py
Local Intelligence
- Project Context
agent/core/project_index.py - Local Intelligence
agent/intelligence/engine.py - Memory Store
agent/memory/store.py - Exact Cache
agent/intelligence/engine.py - Experience Evolution
agent/evolution/engine.py - Skill Registry
agent/skills/registry.py
Inference and Providers
- Adaptive Router
agent/routing/hybrid_router.py - Quality and Fallback
agent/routing/quality_gate.py - Local AI Provider
agent/providers/local_provider.py - Cloud Providers
agent/providers/registry.py - Provider Management
agent/providers/service.py - Local Model Manager
agent/models/manager.py
Workspace Tools
- MCP Tools
agent/tools/mcp_gateway.py - Workspace Access
agent/core/workspace.py - Terminal Sessions
agent/tools/terminal_sessions.py
Other components
- User
Connections 25
- User → Web UI: submits task
- Web UI → API Gateway: sends request
- API Gateway → Agent Execution: dispatches chat
- Agent Execution → Task Planning: plans task
- Agent Execution → Project Context: retrieves context
- Agent Execution → Local Intelligence: checks cache and context
- Local Intelligence → Exact Cache: reads and writes
- Local Intelligence → Memory Store: retrieves memory
- Agent Execution → Adaptive Router: requests candidate
- Adaptive Router → Local AI Provider: selects local
- Adaptive Router → Cloud Providers: selects cloud
- Agent Execution → Quality and Fallback: checks response
- Quality and Fallback → Adaptive Router: requests fallback
- Agent Execution → MCP Tools: dispatches tools
- MCP Tools → Workspace Access: accesses files
- MCP Tools → Terminal Sessions: runs commands
- API Gateway → Workflow Journal: records activity
- Workflow Journal → Web UI: streams progress
- API Gateway → Web UI: returns response
- Agent Execution → Experience Evolution: records outcome
- Experience Evolution → Skill Registry: registers approved skill
- Web UI → Provider Management: manages providers
- Provider Management → Cloud Providers: updates registry
- Local Model Manager → Local AI Provider: reports runtime state
- Gateway Controller → API Gateway: starts and stops