fred-runtime 0.1.15

Runtime adapters and infrastructure wiring for Fred v2 agents.

Fred Runtime

fred-runtime provides the execution layer and infrastructure adapters for Fred v2 agents.

It complements:

• fred-sdk (https://pypi.org/project/fred-sdk/) — agent authoring APIs
• Backend implementations — databases, Knowledge Flow, MCP, tracing, security, and other platform services

---

Overview

The Fred architecture is based on a clear separation between agent definition and agent execution.

fred-sdk — Authoring layer

Developers use fred-sdk to define and structure agents. This includes:

• Chat-oriented profile agents
• Multi-agent or team-based architectures (comparable to frameworks such as Agno)
• Workflow-driven agents

This layer provides:

• Agent interfaces and abstractions
• Tool integration patterns
• Prompt and behavior definitions
• Runtime contracts (in abstract form)

It is intentionally designed to be lightweight, portable, and independent of any specific backend or infrastructure.

---

fred-runtime — Execution layer

fred-runtime provides the concrete implementation required to execute agents defined with fred-sdk.

Its responsibilities include:

• Checkpointing (e.g., LangGraph state persistence)
• Tracing and observability
• Security integration
• Runtime wiring for tools and external systems (MCP, Knowledge Flow, etc.)
• Infrastructure adapters (SQL, storage, and related services)

In summary:

• fred-sdk defines agent behavior and structure
• fred-runtime enables their execution in real environments

---

Backend / platform layer

The hosting application or platform is responsible for:

• Data storage and persistence
• API exposure
• Deployment and operations (local environments, Kubernetes, etc.)
• Integration with external systems

---

Purpose

This package exists to:

• Keep fred-sdk clean, stable, and backend-agnostic
• Centralize runtime logic that would otherwise be duplicated across projects
• Provide a consistent and production-ready execution layer
• Enable external agent repositories (such as rags-v2) to run Fred agents with minimal setup

---

Implementation details

fred-runtime relies on:

• LangChain and LangGraph for execution and orchestration
• Runtime contracts defined in fred-sdk

These dependencies are abstracted so that agent developers can focus on authoring rather than infrastructure concerns.

---

Scope (initial)

• SQL-backed checkpointer for LangGraph runtimes
• Runtime helper utilities extracted from the agentic backend

Additional runtime adapters will be progressively integrated as the platform evolves.

---

Developer workflow

Typical usage:

pip install fred-sdk fred-runtime

1. Use fred-sdk to define agents
2. Use fred-runtime to execute them within a configured environment

---

Conceptual model

• fred-sdk: defines what the agent does
• fred-runtime: defines how the agent runs