Developer Guide
Project Structure
Overview of the Inferia LLM codebase and monorepo layout
Inferia LLM follows a structured monorepo pattern to separate executable applications from reusable core libraries.
Top-Level Directories
| Directory | Description |
|---|---|
apps/ | Deployment entry points (Gateways, Dashboard, Docs). |
package/ | The public Python package (inferia) containing all core logic and services. |
deploy/ | Infrastructure-as-Code (Docker Compose, Kubernetes manifests). |
1. Apps (apps/)
These are the runnable services that get deployed.
inference-gateway/: Entry point for API requests.filtration-gateway/: Entry point for safety guardrails.orchestration-gateway/: Entry point for orchestration.docs/: This documentation site (built with Fumadocs/Next.js).dashboard/: React/Next.js frontend for the management console.
2. Package (package/)
This directory contains the source code for the inferia package. All core logic lives here.
src/inferia/services/: Core business logic (inference, filtration, orchestration).src/inferia/cli.py: Theinferiacommand-line interface entry point.src/inferia/gateways/: Gateway implementation details.
3. Deploy (deploy/)
.env: Centralized environment configuration (found in root anddeploy/).docker-compose.yml: Definition for running the entire stack locally (found in root anddeploy/).