HPD-Agent.Framework 0.5.0

HPD AI Framework

👁 GitHub
👁 Docs
👁 NuGet

A C# framework for building production AI systems — AI agents, RAG pipelines, ML pipelines, authentication, and everything in between.

<picture> <source media="(prefers-color-scheme: dark)" srcset="overview-dark.svg"> <source media="(prefers-color-scheme: light)" srcset="overview.svg"> <img alt="HPD AI Framework Overview" src="overview.svg"> </picture>

HPD-Agent

Production-ready agent framework — tools, multi-turn conversations, middleware, sub-agents, multi-agent workflows, audio, and more. Paired with TypeScript/Svelte UI libraries for streaming chat interfaces.

<picture> <source media="(prefers-color-scheme: dark)" srcset="architecture-dark.svg"> <source media="(prefers-color-scheme: light)" srcset="architecture.svg"> <img alt="HPD-Agent Architecture" src="architecture.svg"> </picture>

Install

dotnet add package HPD-Agent.Framework

Quick Start

using HPD.Agent;

var agent = await new AgentBuilder()
 .WithProvider("openai", "gpt-4o")
 .WithInstructions("You are a helpful assistant.")
 .BuildAsync();

agent.On<TextDeltaEvent>(e =>
{
 Console.Write(e.Text);
 return ValueTask.CompletedTask;
});

await agent.RunAsync("Hello!");

What's Inside

• Tools — AIFunctions, Skills, SubAgents with source-generated wiring. MCP servers, OpenAPI specs, or runtime-provided.
• Sessions & Branches — Conversation history with forking. Automatic crash recovery.
• Sub-Agents — Child agents with own tools, provider, and memory. Stateful, per-session, or one-shot.
• Middleware — Intercept every stage. Built-in: CircuitBreaker, Retry, HistoryReduction, PII, FunctionTimeout, Logging.
• Multi-Agent Workflows — Directed graph composition with conditional routing.
• Evaluation — LLM-as-judge scoring, decompose-verify evaluators, score store, and human-in-the-loop annotation queue.
• Event Runtime — 50+ event types in real time. Send input events with RunAsync, handle output with On<TEvent>.
• Memory & Content Store — ISessionStore, IContentStore, and agent working memory.
• Audio — Native audio input/output for voice agents.
• Observability — OpenTelemetry integration with automatic span hierarchy.

TypeScript Libraries

Documentation

Full docs at hpd-ai.github.io — or browse .

HPD-RAG

Fully modular RAG framework — every node in every pipeline is swappable or removable. Build your own ingestion, retrieval, and evaluation pipelines by snapping blocks together.

<picture> <source media="(prefers-color-scheme: dark)" srcset="rag-architecture-dark.svg"> <source media="(prefers-color-scheme: light)" srcset="rag-architecture.svg"> <img alt="HPD-RAG Architecture" src="rag-architecture.svg"> </picture>

Coming soon. HPD-RAG is under active development.

HPD-ML

Fully modular machine learning framework — data ingestion, feature engineering, model training, and evaluation all composable and extensible. Universal data abstraction with pluggable learners and transforms.

<picture> <source media="(prefers-color-scheme: dark)" srcset="HPD-AI-Framework/dotnet/src/HPD-ML.Framework/architecture-dark.svg"> <source media="(prefers-color-scheme: light)" srcset="HPD-AI-Framework/dotnet/src/HPD-ML.Framework/architecture.svg"> <img alt="HPD-ML Architecture" src="HPD-AI-Framework/dotnet/src/HPD-ML.Framework/architecture.svg"> </picture>

Coming soon. HPD-ML is under active development.

HPD-Auth

Hosted-auth-service experience as an embedded .NET library. Wraps ASP.NET Core Identity and exposes a ready-made REST API — JWT + Cookie dual-auth, session management, 2FA, passkeys, OAuth, admin API, and event-driven audit logging. No separate service to run. No per-user pricing. No data leaving your infrastructure.

<picture> <source media="(prefers-color-scheme: dark)" srcset="HPD-AI-Framework/dotnet/src/HPD-Auth.Framework/architecture-dark.svg"> <source media="(prefers-color-scheme: light)" srcset="HPD-AI-Framework/dotnet/src/HPD-Auth.Framework/architecture.svg"> <img alt="HPD-Auth Architecture" src="HPD-AI-Framework/dotnet/src/HPD-Auth.Framework/architecture.svg"> </picture>

Coming soon. HPD-Auth is under active development.