Execution Routing Engine

ERE.md

Execution Routing Engine

A modular prompt-driven execution runtime for routing user intent to commands, tools, and permission-aware execution pipelines.

Designed as a deterministic runtime layer that analyzes user intent, routes execution requests to the most relevant commands and tools, enforces permission boundaries before execution, and returns fully traceable runtime sessions.

---

Core Architecture

User Prompt
    ↓
PortRouter
    ↓
ExecutionEngine
    ↓
QueryEnginePort
    ↓
RuntimeSession

Component Responsibilities

Component	Responsibility
PortRouter	Tokenizes prompts, scores modules, and routes execution requests
ExecutionRegistry	Stores and resolves available commands and tools
ExecutionEngine	Executes matched components and enforces permission policies
QueryEnginePort	Aggregates execution results into structured runtime output
RuntimeSession	Stores prompt, routing decisions, execution logs, and final results

---

Example Flow

from query_runtime import CleanRuntime

runtime = CleanRuntime()

session = runtime.run(
    "Search how data processing using python?"
)

print(session.routed_matches)
print(session.command_execution_messages)
print(session.tool_execution_messages)

Output

[
    RoutedMatch(
        kind='command',
        name='search',
        score=1
    ),

    RoutedMatch(
        kind='tool',
        name='python',
        score=1
    )
]

---

Key Design Principles

Deterministic Routing

Prompt-to-module routing based on token scoring.

Permission-Aware Execution

Execution permissions are validated before tool invocation.

Full Traceability

Every execution produces a structured RuntimeSession.

Modular Extensibility

New commands, tools, scoring strategies, and policies can be added without changing the core runtime.

---

Current Features

• Prompt tokenization
• Command routing
• Tool routing
• Execution registry
• Permission enforcement
• Runtime session tracking
• Structured execution logging

---

Planned Evolution

• Semantic routing (Embeddings / BM25 / Hybrid retrieval)
• Role-based access control (RBAC)
• Async execution pipelines
• DAG-based dependency execution
• Adaptive confidence thresholds
• Multi-agent orchestration

---

Repository

Full source code:

https://github.com/Ahmad-Al-Dibo/execution-routing-engine

---

Technical Article

Architecture article:

https://www.linkedin.com/pulse/van-lost-data-naar-controle-het-herontwerpen-een-runtime-al-dibo-jjnse/

---

Author

Ahmad Al Dibo Student

This tool was inspired by the leaked Cloudcode code; I transformed it into a usable tool and used artificial intelligence to organize the code.