Skip to content

mikeparcewski/wicked-garden / skills/agentic/frameworks

wicked-garden-agentic-frameworks

Verified

Use when selecting or comparing agentic frameworks (LangChain, LangGraph, CrewAI, AutoGen, etc.) — curated comparison by use case, language, and maturity. Gets latest context via Context7 when available. NOT for reviewing existing agentic code (use review-methodology) or architecture patterns (use agentic-patterns).

refs/framework-profiles-2.md

View on GitHub
---
phase_relevance: ["design", "review"]
archetype_relevance: ["*"]
---
# Framework Profiles (Part 2)

Detailed profiles of AutoGen, Pydantic AI, and LlamaIndex Agents.

## 4. AutoGen

### Overview
Microsoft's framework for multi-agent conversations with flexible dialogue patterns.

### Architecture
- Agents communicate via messages
- Group chat support
- Human-in-the-loop conversations
- Code execution environments

### Code Example
```python
from autogen import AssistantAgent, UserProxyAgent

assistant = AssistantAgent(
    name="assistant",
    llm_config={"model": "gpt-4"}
)

user_proxy = UserProxyAgent(
    name="user",
    human_input_mode="TERMINATE",
    code_execution_config={"work_dir": "coding"}
)

user_proxy.initiate_chat(
    assistant,
    message="Write a Python function to calculate fibonacci"
)
```

### Strengths
- **Conversation Patterns:** Excellent for multi-agent discussions
- **Code Execution:** Built-in safe code execution
- **Human-in-Loop:** Natural conversation with humans
- **Group Chat:** Multiple agents can collaborate
- **Backed by Microsoft:** Research-backed, well-maintained

### Weaknesses
- **Verbose:** Can require lot of code for simple tasks
- **Conversation Complexity:** Managing long conversations is tricky
- **Less Structured:** More flexible but less opinionated

### Best Use Cases
- Agents that need to debate/discuss
- Code generation and execution
- Research and exploration tasks
- Interactive applications

### Not Ideal For
- Simple linear workflows
- Applications not needing conversation
- Quick prototypes (can be verbose)

### Pricing Considerations
- Framework is free
- Pay for LLM API costs
- No additional services

### Learning Resources
- Official docs: microsoft.github.io/autogen
- Paper: arxiv.org/abs/2308.08155
- Examples: github.com/microsoft/autogen/examples

## 5. Pydantic AI

### Overview
Type-safe Python framework for building agents with Pydantic validation.

### Architecture
- Agents are Python classes with type hints
- Pydantic validates all inputs/outputs
- Dependency injection for tools
- Multi-provider support

### Code Example
```python
from pydantic_ai import Agent
from pydantic import BaseModel

class ReviewOutput(BaseModel):
    approved: bool
    issues: list[str]
    confidence: float

agent = Agent(
    'claude-opus-4',
    result_type=ReviewOutput,
    system_prompt='You review code for security issues.'
)

result = agent.run_sync('Review this code: ...')
# result is type-checked ReviewOutput
```

### Strengths
- **Type Safety:** Pydantic validation on all I/O
- **Simple API:** Clean, minimal boilerplate
- **Multi-Provider:** OpenAI, Anthropic, Gemini, etc.
- **Dependency Injection:** Clean tool management

### Weaknesses
- **New Framework:** Still maturing
- **Limited Patterns:** No built-in workflows
- **Small Community:** Fewer examples and resources

### Best Use Cases
- Type-safe applications
- Simple agent use cases
- Teams already using Pydantic
- Data validation requirements

### Not Ideal For
- Complex multi-agent workflows
- Applications needing mature framework
- Teams not familiar with type hints

### Pricing Considerations
- Framework is free
- Pay for LLM API costs
- No additional services

### Learning Resources
- Official docs: ai.pydantic.dev
- Examples: github.com/pydantic/pydantic-ai/examples
- Discord: Pydantic Discord

## 6. LlamaIndex Agents

### Overview
Framework for building agents that reason over data, optimized for RAG applications.

### Architecture
- Query engines for data retrieval
- Agents plan and execute queries
- Tool abstraction for data access
- Multi-step reasoning over documents

### Code Example
```python
from llama_index.agent import OpenAIAgent
from llama_index import VectorStoreIndex

# Build index over documents
index = VectorStoreIndex.from_documents(documents)

# Create query engine tool
query_tool = index.as_query_engine().as_tool(
    name="knowledge_base",
    description="Search company knowledge base"
)

# Create agent with tool
agent = OpenAIAgent.from_tools(
    [query_tool],
    system_prompt="You are a helpful assistant."
)

response = agent.chat("What is our refund policy?")
```

### Strengths
- **RAG Excellence:** Best-in-class retrieval capabilities
- **Query Planning:** Intelligent query decomposition
- **Data Connectors:** 100+ data source connectors
- **Multi-Step Reasoning:** Complex reasoning over documents

### Weaknesses
- **RAG-Focused:** Heavier if not using RAG
- **Complexity:** Learning curve for advanced features
- **Framework Weight:** Larger dependency footprint

### Best Use Cases
- RAG applications (Q&A over docs)
- Knowledge base agents
- Document analysis
- Research assistants

### Not Ideal For
- Non-RAG applications
- Simple workflows
- Applications without data retrieval needs

### Pricing Considerations
- Framework is free
- LlamaCloud (managed service): $99-$999/month
- Pay for LLM API costs

### Learning Resources
- Official docs: docs.llamaindex.ai
- Course: llamaindex.ai/bootcamp
- Examples: github.com/run-llama/llama_index/examples

## Framework Comparison Matrix

| Feature | ADK | LangGraph | CrewAI | AutoGen | Pydantic AI | LlamaIndex |
|---------|-----|-----------|--------|---------|-------------|------------|
| State Management | ★★★☆☆ | ★★★★★ | ★★☆☆☆ | ★★★☆☆ | ★☆☆☆☆ | ★★★☆☆ |
| Type Safety | ★★★★★ | ★★☆☆☆ | ★★☆☆☆ | ★★☆☆☆ | ★★★★★ | ★★☆☆☆ |
| Ease of Use | ★★★★☆ | ★★★☆☆ | ★★★★★ | ★★★☆☆ | ★★★★☆ | ★★★☆☆ |
| Production Ready | ★★★★★ | ★★★★★ | ★★★☆☆ | ★★★★☆ | ★★★☆☆ | ★★★★☆ |
| Observability | ★★★★☆ | ★★★★★ | ★★☆☆☆ | ★★★☆☆ | ★★☆☆☆ | ★★★★☆ |
| Community Size | ★★★☆☆ | ★★★★★ | ★★★☆☆ | ★★★★☆ | ★★☆☆☆ | ★★★★☆ |
| Multi-Provider | ★☆☆☆☆ | ★★★★★ | ★★★★★ | ★★★★★ | ★★★★★ | ★★★★★ |
| RAG Support | ★★☆☆☆ | ★★★★☆ | ★★★☆☆ | ★★★☆☆ | ★★☆☆☆ | ★★★★★ |
reverse-engineer-anythingmorluto/reaReverse engineer native, managed, Electron/JavaScript, packaged, firmware, and browser targets with REA, and build agent environments (AgentEnv bundles) of web applications from a live instance or a capture, and of Linux desktop apps with a task from a request. Use shipped-artifact or requested runtime evidence to explain features, compare versions, decompile code, or guide a reconstruction. Skip REA for ordinary source-repository architecture analysis.Stars: 472kUpdated Claude CodeCodexVerifiedInstall
diagnosing-superpowersobra/superpowersUse when a superpowers session went wrong and your human partner wants to know why — repeated work, ignored plans, stumbles, poor results, a skill that didn't fire, "it took too long", "why is it so expensive", "what is it doing" — or wants to build a bug report for the superpowers maintainers, for the current session or a past one identified by id or path, on any harness.Stars: 322kUpdated Claude CodeCodexVerifiedInstall
dispatching-parallel-agentsobra/superpowersUse when facing 2+ independent tasks that can be worked on without shared state or sequential dependenciesStars: 322kUpdated Claude CodeCodexVerifiedInstall
executing-plansobra/superpowersUse when executing an implementation plan in the current session as the implementer yourself — your human partner chose inline execution, or no subagent tool is availableStars: 322kUpdated Claude CodeCodexVerifiedInstall