LLM Agents & Ecosystem Handbook

A unified handbook for building, deploying and understanding LLM agents and the wider ecosystem

A polished, curated collection of Large Language Model (LLM) agents, tutorials and ecosystem insights. This handbook highlights projects that push the boundaries of generative AI, multi-agent collaboration, retrieval-augmented generation (RAG), voice and game agents, and more. It goes beyond simple link aggregation, aiming to be a one-stop reference for building, deploying, and understanding LLM applications across the entire stack.

Tip: If you enjoy this list, please consider starring the repository to help others discover it!

---

Table of Contents

• Top Agent Frameworks
• Agent Toolkits & Platforms
• Starter AI Agents
• Advanced AI & Domain-Specific Agents
• Multi-Agent Teams
• Voice & Game Agents
• RAG & Memory Examples
• MCP Agent Integrations
• LLM Evaluation Frameworks
• Example Projects
• Tutorials & Learning Resources
  • RAG Tutorials
  • Memory Apps Tutorials
  • Chat with X Tutorials
  • LLM Fine-Tuning Tutorials
• Other Educational Spaces
• Unique Features
• Languages & Multilingual Support
• Interactive Demos & Resources
  • Web Apps
  • Jupyter Notebooks
• Datasets & Design Assets
• Documentation & Roadmap
• Complete Applications
• Beginner’s Guide
• Contributing
• License
• Maintainer

---

Why this repository stands out

This curated collection aims to be a comprehensive resource for developers and researchers building their own LLM applications. In addition to code examples, it provides:

• Comparative analysis of leading agent frameworks: A quick matrix contrasting frameworks (LangGraph, AutoGen, CrewAI, Smolagents, etc.) with key features to help you choose.
• Guidance on framework selection: Practical advice based on task complexity, collaboration needs and ecosystem integrations.
• LLM evaluation toolbox: Summaries of tools like Promptfoo, DeepEval, MLflow LLM Evaluate, RAGAs and Langfuse to measure performance and safety.
• 60+ skeleton projects: The agents folder contains scaffolded agents across many domains (blogging, medical imaging, music generation, multimodal input, news, finance, research, scraping, consultancy, system design, compliance, marketing, scheduling, supply-chain, healthcare, education). Each skeleton includes a README.md and main.py.
• Agent skeleton generator: scripts/create_agent.py to spin up new agent skeletons in seconds.

---

Top Agent Frameworks

Framework	Description & Key Features
LangGraph	Graph/DAG-based orchestration for complex multi-step workflows.
OpenAI Agents SDK	Structured runtime with tool-calling and role-based agents.
AutoGen (AG2)	Event-driven multi-agent conversations and human-in-the-loop.
CrewAI	Role-based “crew” collaboration with memory and error handling.
Google AgentKit (ADK)	Modular Gemini/Vertex AI agent kit with hierarchical tools.
Dify	Low-code builder with RAG and function calling.
LangChain & Tools	Mature chains, memory and 3rd-party integrations.
Smolagents	Minimal, code-centric loop (agents write & execute code).
Semantic Kernel	.NET-first skills/plans; enterprise-friendly.
LlamaIndex Agents	Retrieval-focused agents for data-heavy apps.
Strands Agents	Provider-agnostic SDK with OpenTelemetry.
Pydantic AI	Type-safe IO/tool signatures with great DX.

---

Agent Toolkits & Platforms

Project	Description
AutoGPT	Toolkit for autonomous agents (creation, benchmarking, UI/CLI).
Ollama	Run LLMs locally across macOS/Windows/Linux/Docker.
Lobe Chat	Open-source chat UI with plugins and multimodal support.
OpenDevin	Open initiative towards an AI software engineer.
Open Interpreter	Natural-language coding & local computer control.
MetaGPT	Multi-agent “virtual company” for complex tasks.
PrivateGPT	Secure offline Q&A over your documents.
GPT-Engineer	From natural-language spec to code.
LlamaIndex Tools	Connectors/tools for data agents.
Flowise	Drag-and-drop builder for LLM workflows.
FastChat	Train/serve/evaluate chatbots.
Mem0	Memory layer for personalised LLMs.
Cal.ai	Scheduling assistant with email handling.
Aider	CLI pair-programming agent with Git integration.
Jan	Offline ChatGPT-style desktop app.

---

Starter AI Agents

Agent	Description
AI Blog to Podcast Agent	Convert blog posts into podcasts.
AI Data Analysis Agent	Insights from CSV/structured data.
AI Travel Agent	Trip itineraries (local/cloud).
AI Music Generator	Compose via generative models.
AI Meme Generator (Browser)	Creates memes by overlaying captions on images.
AI Breakup Recovery Agent	Supportive advice for emotional situations.
AI Health & Fitness Agent	Health metrics & coaching.
Gemini Multimodal Agent	Text+image multimodal demo.

---

Advanced AI & Domain-Specific Agents

Agent	Description
AI Deep Research Agent	Multi-source research & synthesis.
AI Consultant Agent	Domain-expert strategy & advice.
AI System Architect Agent	From requirements to architecture.
AI Lead Generation Agent	Identify & qualify prospects.
AI Meeting Agent	Summaries & action items.
OpenAI Research Agent	Research workflows with tools.
Explainable AI Finance Agent	Finance with interpretability.
Web Scraping Agent	Crawl & extract structured data.
Document Processing Agent	OCR + analysis/summarisation.
Sentiment Analysis Agent	Classify sentiment at scale.
Technical Translation Agent	Preserve domain terminology.
Research Synthesizer Agent	RAG + coherent reporting.

---

Multi-Agent Teams

Team	Description
Competitor Intelligence Team	Market/competitor research & reporting.
Finance Agent Team	Budgeting, forecasting and reporting.
Teaching Agent Team	Lesson planning, delivery and assessment.
Multi-Agent Team Demo	Cross-role collaboration patterns.
Mixture of Agents Demo	Specialised agent ensembles.

---

Voice & Game Agents

Agent	Description
Voice Summary Agent	Transcribe and summarise audio.
AI Audio Tour Agent	Generate audio tours for museums/cities.
Customer Support Voice Agent	Handle spoken queries and log issues.
Voice RAG Agent	Voice input + retrieval + TTS.
Tic-Tac-Toe Agent	Autonomous gameplay template.

---

RAG & Memory Examples

Example	Description
Agentic RAG with Reasoning	Retrieval → reasoning → generation.
Hybrid Search RAG	Vector + keyword retrieval.
Vision RAG	Apply RAG to visual data.
CRAG (Corrective RAG)	Human-feedback corrective loop.
Local RAG Agent	Fully offline retrieval pipeline.

---

MCP Agent Integrations

Agent	Description
Browser MCP Agent	Drive a browser (search/click/forms).
GitHub MCP Agent	Read/write/manage repositories.
Notion MCP Agent	Create/update/query Notion pages & DBs.
Travel Planner MCP Agent Team	Multi-agent trip planning via MCP.

---

LLM Evaluation Frameworks

See evaluation_frameworks/README.md for details about Promptfoo, DeepEval, MLflow LLM Evaluate, RAGAs, Deepchecks, LangSmith, TruLens, Arize Phoenix and Langfuse.

---

Example Projects

The agents directory contains many agent skeletons organised by category. Category folders (starter, advanced, teams, rag) act as indexes, while each agent lives in its own top-level folder under agents/.

Quick picks

• Summarization Agent – agents/summarization_agent
• Data Analysis Agent – agents/data_analysis_agent
• Travel Itinerary Agent – agents/travel_itinerary_agent
• Voice Assistant Demo – agents/voice_agent_demo
• Meme Generator Agent – agents/meme_generator_agent
• Health & Fitness Agent – agents/health_fitness_agent
• Breakup Recovery Agent – agents/breakup_recovery_agent
• AI Blog to Podcast Agent – agents/ai_blog_to_podcast_agent
• AI Medical Imaging Agent – agents/ai_medical_imaging_agent
• AI Music Generator Agent – agents/ai_music_generator_agent
• Local News Agent – agents/local_news_agent
• Gemini Multimodal Demo – agents/gemini_multimodal_agent_demo

---

Tutorials & Learning Resources

Hands-on tutorials live in tutorials/ directory.

• RAG Tutorials → tutorials/rag_tutorials
• Memory Apps Tutorials → tutorials/memory_apps
• Chat with X Tutorials → tutorials/chat_with_x_tutorials
• Fine-Tuning Tutorials → tutorials/fine_tuning_tutorials

---

Other Educational Spaces

• Interactive demos & notebooks: web_apps, notebooks
• Datasets & design assets: datasets, design
• LLM ecosystem overview: ecosystem/overview.md
• Complete applications: complete_apps

---

Unique Features

• Educational focus: Detailed tutorials (RAG, memory, chat with X, fine-tuning) + 60+ scaffolded agents.
• Framework comparison & guidance: Practical, vendor-neutral advice.
• Agent skeleton generator: scripts/create_agent.py.
• Evaluation toolbox: Promptfoo, DeepEval, RAGAs, etc.
• Ecosystem breadth: Training, tools, production, local inference, operations, interpretability.
• Community roadmap: See docs/roadmap.md.

---

Languages & Multilingual Support

We welcome translations. See TRANSLATION.md. A Technical Translation Agent exists under agents/technical_translation_agent.

---

Supported LLM Providers

The utilities/llm_provider.py module provides a single complete() function that works with multiple cloud LLM providers. Switch providers without changing agent code — just set LLM_PROVIDER:

Provider	LLM_PROVIDER value	API Key variable	Default model
OpenAI	openai (default)	OPENAI_API_KEY	gpt-4o-mini
Anthropic	anthropic	ANTHROPIC_API_KEY	claude-3-5-haiku-20241022
MiniMax	minimax	MINIMAX_API_KEY	MiniMax-M2.7

MiniMax (platform.minimaxi.com) offers the M2.7 and M2.5 model families with a 204 K-token context window via an OpenAI-compatible API endpoint at https://api.minimax.io/v1.

from utilities.llm_provider import complete

# Use MiniMax M2.7 (set MINIMAX_API_KEY first)
response = complete(
    "Summarise the main agent frameworks in the LLM ecosystem.",
    provider="minimax",
    model="MiniMax-M2.7",
)
print(response)

---

Interactive Demos & Resources

Web Apps

• Streamlit Summariser – web_apps/streamlit_summarizer (supports openai, anthropic, minimax via LLM_PROVIDER)
• Gradio FAQ Bot – web_apps/gradio_faq_bot

Jupyter Notebooks

• Getting Started – notebooks/getting_started.ipynb

---

Datasets & Design Assets

• Sample dataset – datasets/sample_products.csv (+ datasets/README.md)
• Architecture diagram – design/architecture_diagram.png (+ design/README.md)

---

Documentation & Roadmap

• Best practices – docs/best_practices.md
• Framework comparison – docs/framework_comparison.md
• Evaluation frameworks guide – evaluation_frameworks/README.md
• Quickstart – tutorials/quickstart.md
• Roadmap – docs/roadmap.md
• Changelog – CHANGELOG.md

---

Complete Applications

• Task Planner – complete_apps/task_planner
• Health Coach – complete_apps/health_coach

---

Beginner’s Guide

If you’re new to LLMs, start here → docs/beginners_guide.md

---

Contributing

Contributions are welcome! Please open an issue or PR. Ensure new entries are permissively licensed (MIT/Apache-2.0) and well-documented.

---

License

MIT — see LICENSE.

---

Maintainer

Curated & maintained by Sayed Allam (oxbshw). If you find this helpful, please ⭐ star the repo and share feedback via issues/PRs.