| MCP Tool Framework | The Build - Agents as MCP Tools | Frameworks | Architecture | - | 6/28/2025 | Using agents as MCP tools to integrate and orchestrate workflows provides a structured method for multi-agent tasks. |
| Tool Swarm Integration | The Build - Agents as MCP Tools | Frameworks | AI Development | - | 6/28/2025 | Tom explained the process of transferring Tool Swarm to SQL Agent by integrating the Langgraph agent as an MCP tool. |
| A2A Design Pattern | The Build - Agents as MCP Tools | Frameworks | Architecture | - | 6/28/2025 | They discussed using the A2A design pattern for agent-to-agent communication in Langgraph agents. |
| LLM-as-Judge Evaluation | The Build - LangChain Open Deep Research | Frameworks | AI Development | - | 6/28/2025 | Use an LLM as a judge to evaluate reports on conciseness, pleasantness, and ease of use |
| Trace-based Dataset Creation | The Build - LangChain Open Deep Research | Frameworks | Architecture | - | 6/28/2025 | You can filter traces (e.g., errors) and add them to a dataset for analysis |
| Specs Adequacy Evaluation | The Build - LangChain Open Deep Research | Frameworks | Architecture | - | 6/28/2025 | Assess whether the provided content section meets the given specifications |
| Evaluation Limitations Discussion | The Build - LangChain Open Deep Research | Frameworks | Architecture | - | 6/28/2025 | The discussion highlights limitations present in the evaluation segment |
| Evaluation Process Segment | The Build - LangChain Open Deep Research | Frameworks | Architecture | - | 6/28/2025 | This segment is specifically focused on describing the evaluation process |
| Step-by-step Trace Outline | The Build - LangChain Open Deep Research | Frameworks | Architecture | - | 6/28/2025 | The trace outlines every step, covering tool calls and agent transfers |
| Trace Graph Explanation | The Build - LangChain Open Deep Research | Frameworks | Architecture | - | 6/28/2025 | Traces in the graph represent each step of the process, including generating the report plan |
| Interrupt Step Mechanism | The Build - LangChain Open Deep Research | Frameworks | Architecture | - | 6/28/2025 | Tom outlines that a multi-agent system should return with an interrupt step to handle asynchronous tasks. |
| Plan-Then-Present Prompt | The Build - LangChain Open Deep Research | Frameworks | Architecture | - | 6/28/2025 | Cameron proposes that the initial assistant prompt should include a directive to create a plan before presenting it. |
| System Section Architecture | The Build - LangChain Open Deep Research | Frameworks | Architecture | - | 6/28/2025 | Covers the architecture of different sections within the system to illustrate overall design |
| Input-Adaptive Process | The Build - LangChain Open Deep Research | Frameworks | Architecture | - | 6/28/2025 | Describes how the system interprets user input and adjusts its processing accordingly |
| Supervisor-Tool Relationship | The Build - LangChain Open Deep Research | Frameworks | AI Development | - | 6/28/2025 | Explains the relationship between the supervisor component and various tools, showcasing their functionalities |
| Client App Workflow | The Build - LangChain Open Deep Research | Frameworks | Architecture | - | 6/28/2025 | Presents a typical client application workflow in a practical context demonstration |
| Multi-Agent Workflow Introduction | The Build - LangChain Open Deep Research | Frameworks | Architecture | - | 6/28/2025 | Introduces a multi-agent workflow to address the system's inability to recall previous data |
| Auto Report Regeneration | The Build - LangChain Open Deep Research | Frameworks | Architecture | - | 6/28/2025 | System can regenerate the report on demand to accommodate updates or corrections |
| Report Plan Workflow | The Build - LangChain Open Deep Research | Frameworks | Architecture | - | 6/28/2025 | Introduces a structured workflow process that starts with generating the report plan |
| Structured Output Retries | The Build - LangChain Open Deep Research | Frameworks | Architecture | - | 6/28/2025 | You can set the maximum number of structured output retries to automatically reattempt on errors. |
