Spaces:
Running
Running
| { | |
| "id": "agents", | |
| "concept": "Agents & Tool Calling", | |
| "fiction": "The knight discovers the highest art of the realm: breathing autonomy into golems, allowing them to wield tools and plan their own conquests.", | |
| "questions": [ | |
| { | |
| "id": "agt_q1", | |
| "shown": { | |
| "question": "In the context of modern AI, explain what an 'Agent' is.", | |
| "concept_brief": "Ask the knight to describe these autonomous golems. What makes them different from spirits that merely answer questions?" | |
| }, | |
| "check": { | |
| "strategy": "llm_judge", | |
| "spec": { | |
| "rubric": [ | |
| "Mentions an AI/LLM acting autonomously", | |
| "Mentions the ability to use external tools or write code", | |
| "Mentions reasoning or planning steps to achieve a goal" | |
| ] | |
| } | |
| }, | |
| "reveal": "An agent is an AI system that uses an LLM as its engine to autonomously reason, plan, and use tools to achieve a goal." | |
| }, | |
| { | |
| "id": "agt_q2", | |
| "shown": { | |
| "question": "What is the name of the lightweight Hugging Face library specifically designed to build and run these autonomous AI agents?", | |
| "concept_brief": "What is the name of the guild dedicated entirely to the construction and command of these autonomous golems?" | |
| }, | |
| "check": { | |
| "strategy": "keyword", | |
| "spec": { | |
| "correct_any": [ | |
| "smolagents", | |
| "smol agents" | |
| ] | |
| } | |
| }, | |
| "reveal": "The `smolagents` library." | |
| }, | |
| { | |
| "id": "agt_q3", | |
| "shown": { | |
| "question": "Agents cannot interact with the outside world on their own. What do we call the external functions (like web search or calculators) given to an agent?", | |
| "concept_brief": "A golem has no hands of its own. What do we call the physical instruments (like pickaxes or compasses) we attach to it?" | |
| }, | |
| "check": { | |
| "strategy": "keyword", | |
| "spec": { | |
| "correct_any": [ | |
| "tools", | |
| "tool" | |
| ] | |
| } | |
| }, | |
| "reveal": "Tools." | |
| }, | |
| { | |
| "id": "agt_q4", | |
| "shown": { | |
| "question": "In `smolagents`, what is the name of the specific agent class that achieves its goals by writing and executing Python code snippets?", | |
| "concept_brief": "What is the specific classification for a golem that solves problems by writing its own mechanical spells?" | |
| }, | |
| "check": { | |
| "strategy": "keyword", | |
| "spec": { | |
| "correct_any": [ | |
| "codeagent", | |
| "CodeAgent" | |
| ] | |
| } | |
| }, | |
| "reveal": "The `CodeAgent` class." | |
| }, | |
| { | |
| "id": "agt_q5", | |
| "shown": { | |
| "question": "When an agent is given a complex task, it first breaks it down into logical steps before taking action. What is this internal cognitive process called?", | |
| "concept_brief": "Before the golem swings its tool, it pauses to map out its actions. What is this act of foresight called?" | |
| }, | |
| "check": { | |
| "strategy": "keyword", | |
| "spec": { | |
| "correct_any": [ | |
| "planning", | |
| "reasoning", | |
| "chain of thought", | |
| "cot" | |
| ] | |
| } | |
| }, | |
| "reveal": "Planning (or Reasoning / Chain of Thought)." | |
| }, | |
| { | |
| "id": "agt_q6", | |
| "shown": { | |
| "question": "What is the term for the foundational set of instructions given to an agent that defines its persona, rules, and how it should format its outputs?", | |
| "concept_brief": "What is the master scroll placed inside the golem's head that dictates its personality and absolute laws?" | |
| }, | |
| "check": { | |
| "strategy": "keyword", | |
| "spec": { | |
| "correct_any": [ | |
| "system prompt", | |
| "system message" | |
| ] | |
| } | |
| }, | |
| "reveal": "The system prompt." | |
| }, | |
| { | |
| "id": "agt_q7", | |
| "shown": { | |
| "question": "Explain why providing a highly detailed docstring (description) is critical when creating a custom tool for an agent.", | |
| "concept_brief": "Why must the knight engrave exactly what a tool does onto its handle before handing it to the golem?" | |
| }, | |
| "check": { | |
| "strategy": "llm_judge", | |
| "spec": { | |
| "rubric": [ | |
| "Mentions that the LLM/Agent reads the description to understand what the tool does", | |
| "Mentions that without it, the agent won't know when or how to use the tool" | |
| ] | |
| } | |
| }, | |
| "reveal": "The agent's LLM relies entirely on the docstring to understand what the tool does, what inputs it requires, and when it is appropriate to use it." | |
| }, | |
| { | |
| "id": "agt_q8", | |
| "shown": { | |
| "question": "What specific method do you call on an initialized agent object in `smolagents` to start it on a task?", | |
| "concept_brief": "What is the single command uttered by the knight to awaken the golem and send it on its quest?" | |
| }, | |
| "check": { | |
| "strategy": "keyword", | |
| "spec": { | |
| "correct_any": [ | |
| "run", | |
| "run()" | |
| ] | |
| } | |
| }, | |
| "reveal": "The `run()` method (e.g., `agent.run(\"task\")`)." | |
| }, | |
| { | |
| "id": "agt_q9", | |
| "shown": { | |
| "question": "What is the name of the built-in tool in the `smolagents` library that allows an agent to browse the internet for real-time information?", | |
| "concept_brief": "What is the name of the specific tool that grants the golem the power of the all-seeing digital eye?" | |
| }, | |
| "check": { | |
| "strategy": "keyword", | |
| "spec": { | |
| "correct_any": [ | |
| "duckduckgosearchtool", | |
| "DuckDuckGoSearchTool" | |
| ], | |
| "partial_any": [ | |
| "search" | |
| ] | |
| } | |
| }, | |
| "reveal": "The `DuckDuckGoSearchTool`." | |
| }, | |
| { | |
| "id": "agt_q10", | |
| "shown": { | |
| "question": "Advanced workflows sometimes involve an overarching 'Manager' agent delegating tasks to smaller, specialized 'Worker' agents. What is this architecture called?", | |
| "concept_brief": "When a grand golem commands a legion of lesser golems to finish a castle, what is this army structure named?" | |
| }, | |
| "check": { | |
| "strategy": "keyword", | |
| "spec": { | |
| "correct_any": [ | |
| "multi-agent", | |
| "multiagent", | |
| "multi agent system" | |
| ] | |
| } | |
| }, | |
| "reveal": "Multi-agent architecture." | |
| } | |
| ] | |
| } |