AI Agent Knowledge Base

This is an old revision of the document!

---

AgentWiki

Core Concepts

• Planning
• Memory
• Tool Use
• Structured Outputs
• Autonomy
• Agent Safety

Reasoning & Planning

• Chain-of-Thought
• Tree of Thoughts
• ReAct
• Reflexion
• Prompt Chaining
• Prompt Engineering
• Latent Reasoning
• Test-Time Compute

Memory Systems

• Hierarchical Memory
• Short-Term
• Long-Term
• Augmentation
• Memory Frameworks
• RAG
• Knowledge Graphs
• Embeddings
• Continual Learning

Retrieval

• FAISS
• HNSW
• ScaNN

Agent Types

• Autonomous
• ReAct Agents
• Plan & Execute
• Conversational
• Tool-Using
• Code Agents
• Web Browsing
• Vision Agents
• Voice Agents
• AI Scientists
• GUI Grounding
• Long-Horizon

Design Patterns

• Agent Loop
• Orchestration
• Multi-Agent
• Modular Architectures
• Function Calling
• Tool Integration
• Parallel Calls
• Speculative Execution
• Durable Execution
• Context Management

Training & Alignment

• RLHF / DPO
• RLVR
• Agentic RL
• Process Rewards
• Reasoning Rewards
• Fine-Tuning
• Trajectory Optimization
• World Models
• Neurosymbolic

Frameworks

• MCP
• A2A Protocol
• LangGraph
• CrewAI
• AutoGen
• DSPy
• LlamaIndex
• OpenAI Agents SDK
• Claude Agent SDK
• Semantic Kernel
• Smolagents
• OpenHands
• Langroid
• Flowise
• LiteLLM
• GraphRAG

Tools & Products

• Devin
• Cursor
• Claude Code
• Computer Use
• Bolt.new

Safety & Governance

• Sandbox Security
• Governance
• Threat Modeling

Evaluation

• Overview
• SWE-bench
• WebArena
• Agent-as-Judge
• Agent Index

Development

• Agentic Coding
• Debugging

Meta

• Wiki API Skill