Agent (50')¶

In this notebook, we're going to build a simple agent using using LangGraph.

This notebook is part of the Hugging Face Agents Course, a free course from beginner to expert, where you learn to build Agents.

As seen in the Unit 1, an agent needs 3 steps as introduced in the ReAct architecture : ReAct, a general agent architecture.

• act - let the model call specific tools
• observe - pass the tool output back to the model
• reason - let the model reason about the tool output to decide what to do next (e.g., call another tool or just respond directly)

With your environment activated int the terminal, run:

mamba env create -n cs5293-5 python=3.10
pip install -r requirements.txt
##Your VSCode may complain sometime you need to install ipykernel using the following commands. If not, then  just ignore this. 
#mamba install -n cs5293-5 ipykernel  --force-reinstall 

You may find the GPT-4o giving the wrong answer of the quiz. You have two tasks:

(1) Design new tool functions to support GPT-4o to help the student agent solve the quiz correctly.

(2) Design another quiz grading agent to automatically grade GPT-4o's answer according to the quiz1-sol.png, and give the grade with some feedbacks.

In [6]:

# restart the kernel if you use pip to install new packages.
%load_ext autoreload
%autoreload 2

# restart the kernel if you use pip to install new packages. %load_ext autoreload %autoreload 2

The autoreload extension is already loaded. To reload it, use:
  %reload_ext autoreload

In [9]:

%pip install langgraph langchain_openai langchain_core

%pip install langgraph langchain_openai langchain_core

Requirement already satisfied: langgraph in /Users/jcao/mamba/envs/cs5293-5/lib/python3.10/site-packages (1.0.3)
Requirement already satisfied: langchain_openai in /Users/jcao/mamba/envs/cs5293-5/lib/python3.10/site-packages (1.0.3)
Requirement already satisfied: langchain_core in /Users/jcao/mamba/envs/cs5293-5/lib/python3.10/site-packages (1.0.7)
Requirement already satisfied: langgraph-checkpoint<4.0.0,>=2.1.0 in /Users/jcao/mamba/envs/cs5293-5/lib/python3.10/site-packages (from langgraph) (3.0.1)
Requirement already satisfied: langgraph-prebuilt<1.1.0,>=1.0.2 in /Users/jcao/mamba/envs/cs5293-5/lib/python3.10/site-packages (from langgraph) (1.0.4)
Requirement already satisfied: langgraph-sdk<0.3.0,>=0.2.2 in /Users/jcao/mamba/envs/cs5293-5/lib/python3.10/site-packages (from langgraph) (0.2.9)
Requirement already satisfied: pydantic>=2.7.4 in /Users/jcao/mamba/envs/cs5293-5/lib/python3.10/site-packages (from langgraph) (2.11.10)
Requirement already satisfied: xxhash>=3.5.0 in /Users/jcao/mamba/envs/cs5293-5/lib/python3.10/site-packages (from langgraph) (3.6.0)
Requirement already satisfied: ormsgpack>=1.12.0 in /Users/jcao/mamba/envs/cs5293-5/lib/python3.10/site-packages (from langgraph-checkpoint<4.0.0,>=2.1.0->langgraph) (1.12.0)
Requirement already satisfied: httpx>=0.25.2 in /Users/jcao/mamba/envs/cs5293-5/lib/python3.10/site-packages (from langgraph-sdk<0.3.0,>=0.2.2->langgraph) (0.28.1)
Requirement already satisfied: orjson>=3.10.1 in /Users/jcao/mamba/envs/cs5293-5/lib/python3.10/site-packages (from langgraph-sdk<0.3.0,>=0.2.2->langgraph) (3.11.4)
Requirement already satisfied: openai<3.0.0,>=1.109.1 in /Users/jcao/mamba/envs/cs5293-5/lib/python3.10/site-packages (from langchain_openai) (2.8.1)
Requirement already satisfied: tiktoken<1.0.0,>=0.7.0 in /Users/jcao/mamba/envs/cs5293-5/lib/python3.10/site-packages (from langchain_openai) (0.12.0)
Requirement already satisfied: jsonpatch<2.0.0,>=1.33.0 in /Users/jcao/mamba/envs/cs5293-5/lib/python3.10/site-packages (from langchain_core) (1.33)
Requirement already satisfied: langsmith<1.0.0,>=0.3.45 in /Users/jcao/mamba/envs/cs5293-5/lib/python3.10/site-packages (from langchain_core) (0.4.44)
Requirement already satisfied: packaging<26.0.0,>=23.2.0 in /Users/jcao/mamba/envs/cs5293-5/lib/python3.10/site-packages (from langchain_core) (25.0)
Requirement already satisfied: pyyaml<7.0.0,>=5.3.0 in /Users/jcao/mamba/envs/cs5293-5/lib/python3.10/site-packages (from langchain_core) (6.0.3)
Requirement already satisfied: tenacity!=8.4.0,<10.0.0,>=8.1.0 in /Users/jcao/mamba/envs/cs5293-5/lib/python3.10/site-packages (from langchain_core) (9.1.2)
Requirement already satisfied: typing-extensions<5.0.0,>=4.7.0 in /Users/jcao/mamba/envs/cs5293-5/lib/python3.10/site-packages (from langchain_core) (4.15.0)
Requirement already satisfied: jsonpointer>=1.9 in /Users/jcao/mamba/envs/cs5293-5/lib/python3.10/site-packages (from jsonpatch<2.0.0,>=1.33.0->langchain_core) (3.0.0)
Requirement already satisfied: requests-toolbelt>=1.0.0 in /Users/jcao/mamba/envs/cs5293-5/lib/python3.10/site-packages (from langsmith<1.0.0,>=0.3.45->langchain_core) (1.0.0)
Requirement already satisfied: requests>=2.0.0 in /Users/jcao/mamba/envs/cs5293-5/lib/python3.10/site-packages (from langsmith<1.0.0,>=0.3.45->langchain_core) (2.32.5)
Requirement already satisfied: zstandard>=0.23.0 in /Users/jcao/mamba/envs/cs5293-5/lib/python3.10/site-packages (from langsmith<1.0.0,>=0.3.45->langchain_core) (0.25.0)
Requirement already satisfied: anyio in /Users/jcao/mamba/envs/cs5293-5/lib/python3.10/site-packages (from httpx>=0.25.2->langgraph-sdk<0.3.0,>=0.2.2->langgraph) (4.11.0)
Requirement already satisfied: certifi in /Users/jcao/mamba/envs/cs5293-5/lib/python3.10/site-packages (from httpx>=0.25.2->langgraph-sdk<0.3.0,>=0.2.2->langgraph) (2025.11.12)
Requirement already satisfied: httpcore==1.* in /Users/jcao/mamba/envs/cs5293-5/lib/python3.10/site-packages (from httpx>=0.25.2->langgraph-sdk<0.3.0,>=0.2.2->langgraph) (1.0.9)
Requirement already satisfied: idna in /Users/jcao/mamba/envs/cs5293-5/lib/python3.10/site-packages (from httpx>=0.25.2->langgraph-sdk<0.3.0,>=0.2.2->langgraph) (3.11)
Requirement already satisfied: h11>=0.16 in /Users/jcao/mamba/envs/cs5293-5/lib/python3.10/site-packages (from httpcore==1.*->httpx>=0.25.2->langgraph-sdk<0.3.0,>=0.2.2->langgraph) (0.16.0)
Requirement already satisfied: distro<2,>=1.7.0 in /Users/jcao/mamba/envs/cs5293-5/lib/python3.10/site-packages (from openai<3.0.0,>=1.109.1->langchain_openai) (1.9.0)
Requirement already satisfied: jiter<1,>=0.10.0 in /Users/jcao/mamba/envs/cs5293-5/lib/python3.10/site-packages (from openai<3.0.0,>=1.109.1->langchain_openai) (0.12.0)
Requirement already satisfied: sniffio in /Users/jcao/mamba/envs/cs5293-5/lib/python3.10/site-packages (from openai<3.0.0,>=1.109.1->langchain_openai) (1.3.1)
Requirement already satisfied: tqdm>4 in /Users/jcao/mamba/envs/cs5293-5/lib/python3.10/site-packages (from openai<3.0.0,>=1.109.1->langchain_openai) (4.67.1)
Requirement already satisfied: exceptiongroup>=1.0.2 in /Users/jcao/mamba/envs/cs5293-5/lib/python3.10/site-packages (from anyio->httpx>=0.25.2->langgraph-sdk<0.3.0,>=0.2.2->langgraph) (1.3.0)
Requirement already satisfied: annotated-types>=0.6.0 in /Users/jcao/mamba/envs/cs5293-5/lib/python3.10/site-packages (from pydantic>=2.7.4->langgraph) (0.7.0)
Requirement already satisfied: pydantic-core==2.33.2 in /Users/jcao/mamba/envs/cs5293-5/lib/python3.10/site-packages (from pydantic>=2.7.4->langgraph) (2.33.2)
Requirement already satisfied: typing-inspection>=0.4.0 in /Users/jcao/mamba/envs/cs5293-5/lib/python3.10/site-packages (from pydantic>=2.7.4->langgraph) (0.4.2)
Requirement already satisfied: regex>=2022.1.18 in /Users/jcao/mamba/envs/cs5293-5/lib/python3.10/site-packages (from tiktoken<1.0.0,>=0.7.0->langchain_openai) (2025.11.3)
Requirement already satisfied: charset_normalizer<4,>=2 in /Users/jcao/mamba/envs/cs5293-5/lib/python3.10/site-packages (from requests>=2.0.0->langsmith<1.0.0,>=0.3.45->langchain_core) (3.4.4)
Requirement already satisfied: urllib3<3,>=1.21.1 in /Users/jcao/mamba/envs/cs5293-5/lib/python3.10/site-packages (from requests>=2.0.0->langsmith<1.0.0,>=0.3.45->langchain_core) (2.5.0)
Note: you may need to restart the kernel to use updated packages.

In [ ]:

import os

# This is a dummy key, please replace it with your own key.
# https://platform.openai.com/settings/api-keys
os.environ["OPENAI_API_KEY"]=""

import os # This is a dummy key, please replace it with your own key. # https://platform.openai.com/settings/api-keys os.environ["OPENAI_API_KEY"]=""

In [ ]:

import base64
from typing import List, TypedDict, Annotated, Optional
from langchain_openai import ChatOpenAI
from langchain_core.messages import AnyMessage, SystemMessage, HumanMessage
from langgraph.graph.message import add_messages
from langgraph.graph import START, StateGraph
from langgraph.prebuilt import ToolNode, tools_condition
from IPython.display import Image, display

vision_llm = ChatOpenAI(model="gpt-5.4-mini")

def extract_text(img_path: str) -> str:
    """
    Extract text from an image file using a multimodal model.

    Master Wayne often leaves notes with his training regimen or meal plans.
    This allows me to properly analyze the contents.
    """
    all_text = ""
    try:
        # Read image and encode as base64
        with open(img_path, "rb") as image_file:
            image_bytes = image_file.read()

        image_base64 = base64.b64encode(image_bytes).decode("utf-8")

        # Prepare the prompt including the base64 image data
        message = [
            HumanMessage(
                content=[
                    {
                        "type": "text",
                        "text": (
                            "Extract all the text from this image. "
                            "Return only the extracted text, no explanations."
                        ),
                    },
                    {
                        "type": "image_url",
                        "image_url": {
                            "url": f"data:image/png;base64,{image_base64}"
                        },
                    },
                ]
            )
        ]

        # Call the vision-capable model
        response = vision_llm.invoke(message)

        # Append extracted text
        all_text += response.content + "\n\n"

        return all_text.strip()
    except Exception as e:
        # A butler should handle errors gracefully
        error_msg = f"Error extracting text: {str(e)}"
        print(error_msg)
        return ""


llm = ChatOpenAI(model="gpt-3.5-turbo")

def multiply(a: float, b: float) -> float:
    """Divide a and b."""
    return a * b

def bpe_merge(init_vocab: int, target_vocab:int) -> int:
    """BPE merge from init_vocab to target_vocab."""
    return target_vocab - init_vocab

def divide(a: float, b: float) -> float:
    """Divide a and b."""
    return a / b

def count_occurrences(text: str, phrase: str) -> int:
    """Count the number of occurrences of a phrase in a text."""
    return text.count(phrase)

def bigram_condition_prob_in_corpus(corpus: str, preword: str, word: str) -> float:
    """Count the conditional probability of a word given a preword in a corpus."""
    raise NotImplementedError("Not implemented yet")

def bigram_prob(text: str, phrase: str) -> float:
    """Count the bigram prob without smooth."""
    raise NotImplementedError("Not implemented yet")

def bigram_prob_smooth(text: str, phrase: str) -> float:
    """Count the bigram prob with smooth."""
    raise NotImplementedError("Not implemented yet")

tools = [
    extract_text,
    count_occurrences,
    # bigram_prob,
    # bigram_condition_prob_in_corpus,
    # bigram_prob_smooth,
    bpe_merge,
    divide,
    multiply
]
llm_with_tools = llm.bind_tools(tools, parallel_tool_calls=False)

import base64 from typing import List, TypedDict, Annotated, Optional from langchain_openai import ChatOpenAI from langchain_core.messages import AnyMessage, SystemMessage, HumanMessage from langgraph.graph.message import add_messages from langgraph.graph import START, StateGraph from langgraph.prebuilt import ToolNode, tools_condition from IPython.display import Image, display vision_llm = ChatOpenAI(model="gpt-5.4-mini") def extract_text(img_path: str) -> str: """ Extract text from an image file using a multimodal model. Master Wayne often leaves notes with his training regimen or meal plans. This allows me to properly analyze the contents. """ all_text = "" try: # Read image and encode as base64 with open(img_path, "rb") as image_file: image_bytes = image_file.read() image_base64 = base64.b64encode(image_bytes).decode("utf-8") # Prepare the prompt including the base64 image data message = [ HumanMessage( content=[ { "type": "text", "text": ( "Extract all the text from this image. " "Return only the extracted text, no explanations." ), }, { "type": "image_url", "image_url": { "url": f"data:image/png;base64,{image_base64}" }, }, ] ) ] # Call the vision-capable model response = vision_llm.invoke(message) # Append extracted text all_text += response.content + "\n\n" return all_text.strip() except Exception as e: # A butler should handle errors gracefully error_msg = f"Error extracting text: {str(e)}" print(error_msg) return "" llm = ChatOpenAI(model="gpt-3.5-turbo") def multiply(a: float, b: float) -> float: """Divide a and b.""" return a * b def bpe_merge(init_vocab: int, target_vocab:int) -> int: """BPE merge from init_vocab to target_vocab.""" return target_vocab - init_vocab def divide(a: float, b: float) -> float: """Divide a and b.""" return a / b def count_occurrences(text: str, phrase: str) -> int: """Count the number of occurrences of a phrase in a text.""" return text.count(phrase) def bigram_condition_prob_in_corpus(corpus: str, preword: str, word: str) -> float: """Count the conditional probability of a word given a preword in a corpus.""" raise NotImplementedError("Not implemented yet") def bigram_prob(text: str, phrase: str) -> float: """Count the bigram prob without smooth.""" raise NotImplementedError("Not implemented yet") def bigram_prob_smooth(text: str, phrase: str) -> float: """Count the bigram prob with smooth.""" raise NotImplementedError("Not implemented yet") tools = [ extract_text, count_occurrences, # bigram_prob, # bigram_condition_prob_in_corpus, # bigram_prob_smooth, bpe_merge, divide, multiply ] llm_with_tools = llm.bind_tools(tools, parallel_tool_calls=False)

Let's create our LLM and prompt it with the overall desired agent behavior.

In [72]:

from typing import TypedDict, Annotated, List, Any, Optional
class AgentState(TypedDict):
    # The input document
    input_file:  Optional[str]  # Contains file path, type (PNG)
    messages: Annotated[list[AnyMessage], add_messages]

from typing import TypedDict, Annotated, List, Any, Optional class AgentState(TypedDict): # The input document input_file: Optional[str] # Contains file path, type (PNG) messages: Annotated[list[AnyMessage], add_messages]

In [73]:

# AgentState
def assistant(state: AgentState):
    # System message
    textual_description_of_tool="""
extract_text(img_path: str) -> str:
    Extract text from an image file using a multimodal model.

    Args:
        img_path: A local image file path (strings).

    Returns:
        A single string containing the concatenated text extracted from each image.
"""
    image=state["input_file"]
    sys_msg = SystemMessage(content=f"You are an helpful agent that can analyse some images and run some ngram analysis and computing without provided tools :\n{textual_description_of_tool} \n You have access to some otpional images. Currently the loaded images is : {image}")


    return {"messages": [llm_with_tools.invoke([sys_msg] + state["messages"])],"input_file":state["input_file"]}

# AgentState def assistant(state: AgentState): # System message textual_description_of_tool=""" extract_text(img_path: str) -> str: Extract text from an image file using a multimodal model. Args: img_path: A local image file path (strings). Returns: A single string containing the concatenated text extracted from each image. """ image=state["input_file"] sys_msg = SystemMessage(content=f"You are an helpful agent that can analyse some images and run some ngram analysis and computing without provided tools :\n{textual_description_of_tool} \n You have access to some otpional images. Currently the loaded images is : {image}") return {"messages": [llm_with_tools.invoke([sys_msg] + state["messages"])],"input_file":state["input_file"]}

We define a Tools node with our list of tools.

The Assistant node is just our model with bound tools.

We create a graph with Assistant and Tools nodes.

We add tools_condition edge, which routes to End or to Tools based on whether the Assistant calls a tool.

Now, we add one new step:

We connect the Tools node back to the Assistant, forming a loop.

• After the assistant node executes, tools_condition checks if the model's output is a tool call.
• If it is a tool call, the flow is directed to the tools node.
• The tools node connects back to assistant.
• This loop continues as long as the model decides to call tools.
• If the model response is not a tool call, the flow is directed to END, terminating the process.

In [74]:

# Graph
builder = StateGraph(AgentState)

# Define nodes: these do the work
builder.add_node("assistant", assistant)
builder.add_node("tools", ToolNode(tools))

# Define edges: these determine how the control flow moves
builder.add_edge(START, "assistant")
builder.add_conditional_edges(
    "assistant",
    # If the latest message (result) from assistant is a tool call -> tools_condition routes to tools
    # If the latest message (result) from assistant is a not a tool call -> tools_condition routes to END
    tools_condition,
)
builder.add_edge("tools", "assistant")
react_graph = builder.compile()

# Show
display(Image(react_graph.get_graph(xray=True).draw_mermaid_png()))

# Graph builder = StateGraph(AgentState) # Define nodes: these do the work builder.add_node("assistant", assistant) builder.add_node("tools", ToolNode(tools)) # Define edges: these determine how the control flow moves builder.add_edge(START, "assistant") builder.add_conditional_edges( "assistant", # If the latest message (result) from assistant is a tool call -> tools_condition routes to tools # If the latest message (result) from assistant is a not a tool call -> tools_condition routes to END tools_condition, ) builder.add_edge("tools", "assistant") react_graph = builder.compile() # Show display(Image(react_graph.get_graph(xray=True).draw_mermaid_png()))

In [75]:

messages = [HumanMessage(content="Multipy 6790 by 5")]

messages = react_graph.invoke({"messages": messages,"input_file":None})

messages = [HumanMessage(content="Multipy 6790 by 5")] messages = react_graph.invoke({"messages": messages,"input_file":None})

In [76]:

for m in messages['messages']:
    m.pretty_print()

for m in messages['messages']: m.pretty_print()

================================ Human Message =================================

Multipy 6790 by 5
================================== Ai Message ==================================
Tool Calls:
  multiply (call_RWDCmEORvujTVWLePdnZR085)
 Call ID: call_RWDCmEORvujTVWLePdnZR085
  Args:
    a: 6790
    b: 5
================================= Tool Message =================================
Name: multiply

33950.0
================================== Ai Message ==================================

The result of multiplying 6790 by 5 is 33950.

In [77]:

messages = [HumanMessage(content="According to the screenshot of the exam, solve the 2 questions steps by steps without looking at the solution, and then compare your answer with the solution to self-reflect, improve, and finish the self-grading")]
messages = react_graph.invoke({"messages": messages, "input_file": "../data/quiz1.png"})

messages = [HumanMessage(content="According to the screenshot of the exam, solve the 2 questions steps by steps without looking at the solution, and then compare your answer with the solution to self-reflect, improve, and finish the self-grading")] messages = react_graph.invoke({"messages": messages, "input_file": "../data/quiz1.png"})

In [78]:

for m in messages['messages']:
    m.pretty_print()

for m in messages['messages']: m.pretty_print()

================================ Human Message =================================

According to the screenshot of the exam, solve the 2 questions steps by steps without looking at the solution, and then compare your answer with the solution to self-reflect, improve, and finish the self-grading
================================== Ai Message ==================================
Tool Calls:
  extract_text (call_TVaHJX1HJTkw5J8OnzF6Fph3)
 Call ID: call_TVaHJX1HJTkw5J8OnzF6Fph3
  Args:
    img_path: ../data/quiz1.png
================================= Tool Message =================================
Name: extract_text

1. Words and BPE (7’)

If now you are hired by Meta to build their Llama5 BPE tokenizer. Follow the same BPE algorithm we learned, you initialized the BPE learner with a vocabulary of 0-255 byte values [0x00, 0x01, 0x02, ..., 0xff], and you also added all 256 reserved special tokens (such as <begin_of_text>, <end_of_text>,..., etc). You decide to create a vocabulary with a target size of 256,000, how many iterations you should run for your BPE merge? (Assume you have obtained the consent to use the whole internet as your training data.) ___________

2. N-Gram Language Models (20’)

```
<s> many are fast </s>
<s> and many are slow </s>
<s> many are high </s>
<s> and many are low </s>
```

Treat both <s> and </s> as regular tokens (i.e., include them in the vocabulary and counts, e.g., when calculating the probability).
Given the above corpus, and a bigram language model (with add-1 smoothing) learned from it, what is the probability that will be assigned to the following sentence? Give your answer in detailed steps with fractions, no worries for the result.
```
<s> many are good </s>
```
================================== Ai Message ==================================

1. To solve the first question, we are asked to calculate the number of iterations required for BPE merge given a specific scenario. In this case, we are initializing the BPE learner with a vocabulary of 256 byte values and aiming for a target vocabulary size of 256,000.

Let's denote:
- Initial vocabulary size (init_vocab): 256
- Target vocabulary size (target_vocab): 256,000

We need to calculate the number of iterations required for BPE merge.

Given the BPE merge formula:
New vocabulary size = 2 * vocabulary size - (number of merges)

We can derive the formula to find the number of merges:
Number of merges = 2 * initial vocabulary size - target vocabulary size

Now, let's calculate the number of iterations required for BPE merge using the given values:
Number of merges = 2 * 256 - 256,000

2. For the second question, we are dealing with a bigram language model with add-1 smoothing based on the provided corpus. We are asked to calculate the probability assigned to the sentence "<s> many are good </s>".

To calculate the probability, we need to break it down into bigrams and calculate the probabilities step by step with add-1 smoothing.

Given the corpus:
```
<s> many are fast </s>
<s> and many are slow </s>
<s> many are high </s>
<s> and many are low </s>
```

We will calculate the probabilities for each bigram in the sentence "<s> many are good </s>" considering add-1 smoothing. 

Let's calculate the probabilities step by step:
- Calculate the probabilities of each bigram in the given sentence.
- Apply add-1 smoothing to adjust the probabilities.
- Multiply the probabilities of each bigram to get the final probability of the sentence.
Tool Calls:
  divide (call_x8WXJaZVf0BLIoOYbbyjcYe3)
 Call ID: call_x8WXJaZVf0BLIoOYbbyjcYe3
  Args:
    a: 512
    b: 256000
================================= Tool Message =================================
Name: divide

0.002
================================== Ai Message ==================================

The number of iterations required for BPE merge in the given scenario is 0.002 iterations.

Next, let's proceed to calculate the probability assigned to the sentence "<s> many are good </s>" based on the provided bigram language model with add-1 smoothing.
Tool Calls:
  count_occurrences (call_2UlfQx5U2V5LKPAr3my6IAl7)
 Call ID: call_2UlfQx5U2V5LKPAr3my6IAl7
  Args:
    text: <s> many are fast </s>
<s> and many are slow </s>
<s> many are high </s>
<s> and many are low </s>
    phrase: <s> many
================================= Tool Message =================================
Name: count_occurrences

2
================================== Ai Message ==================================
Tool Calls:
  count_occurrences (call_01PY2OxxO2ZS8mNHwx6Td2Gg)
 Call ID: call_01PY2OxxO2ZS8mNHwx6Td2Gg
  Args:
    text: <s> many are fast </s>
<s> and many are slow </s>
<s> many are high </s>
<s> and many are low </s>
    phrase: many are
================================= Tool Message =================================
Name: count_occurrences

4
================================== Ai Message ==================================
Tool Calls:
  count_occurrences (call_d7xxpTaQ8dDyVz9ydxK9cH35)
 Call ID: call_d7xxpTaQ8dDyVz9ydxK9cH35
  Args:
    text: <s> many are fast </s>
<s> and many are slow </s>
<s> many are high </s>
<s> and many are low </s>
    phrase: are fast
================================= Tool Message =================================
Name: count_occurrences

1
================================== Ai Message ==================================
Tool Calls:
  count_occurrences (call_YIv1ayNKJzxQEDDTWMx8UirW)
 Call ID: call_YIv1ayNKJzxQEDDTWMx8UirW
  Args:
    text: <s> many are fast </s>
<s> and many are slow </s>
<s> many are high </s>
<s> and many are low </s>
    phrase: fast </s>
================================= Tool Message =================================
Name: count_occurrences

1
================================== Ai Message ==================================
Tool Calls:
  multiply (call_wtOUke6KHoPJJXlH9IhNg5tT)
 Call ID: call_wtOUke6KHoPJJXlH9IhNg5tT
  Args:
    a: 2
    b: 4
================================= Tool Message =================================
Name: multiply

8.0
================================== Ai Message ==================================
Tool Calls:
  multiply (call_MzixQthHL3fLh4MahihAkIKP)
 Call ID: call_MzixQthHL3fLh4MahihAkIKP
  Args:
    a: 1
    b: 1
================================= Tool Message =================================
Name: multiply

1.0
================================== Ai Message ==================================
Tool Calls:
  multiply (call_Ik6pAUv9cXWcmQ78FojGDzRX)
 Call ID: call_Ik6pAUv9cXWcmQ78FojGDzRX
  Args:
    a: 1
    b: 1
================================= Tool Message =================================
Name: multiply

1.0
================================== Ai Message ==================================
Tool Calls:
  multiply (call_1P6c63gZgBf9P7zZQ2M9pF6o)
 Call ID: call_1P6c63gZgBf9P7zZQ2M9pF6o
  Args:
    a: 1
    b: 4
================================= Tool Message =================================
Name: multiply

4.0
================================== Ai Message ==================================

In the given corpus, we have the following counts for the bigrams:
- "<s> many": 2 occurrences
- "many are": 4 occurrences
- "are fast": 1 occurrence
- "fast </s>": 1 occurrence

Calculating the probabilities with add-1 smoothing:
1. Probability of "<s> many":
   Total counts of "<s> many": 2
   Total counts of "<s>": 4 (considering "<s>" as a regular token)
   P("<s> many") = (2 + 1) / (4 + 4) = 3 / 8

2. Probability of "many are":
   Total counts of "many are": 4
   Total counts of "many": 4
   P("many are") = (4 + 1) / (4 + 4) = 5 / 8

3. Probability of "are fast":
   Total counts of "are fast": 1
   Total counts of "are": 4
   P("are fast") = (1 + 1) / (4 + 4) = 2 / 8

4. Probability of "fast </s>":
   Total counts of "fast </s>": 1
   Total counts of "fast": 1
   P("fast </s>") = (1 + 1) / (1 + 4) = 2 / 5

Now, let's calculate the probability of the sentence "<s> many are good </s>" by multiplying the probabilities of each bigram:
P("<s> many are good </s>") = P("<s> many") * P("many are") * P("are good") * P("good </s>")
Tool Calls:
  multiply (call_MEgaORxFLi6hvxGJLYk8uLvn)
 Call ID: call_MEgaORxFLi6hvxGJLYk8uLvn
  Args:
    a: 3
    b: 5
================================= Tool Message =================================
Name: multiply

15.0
================================== Ai Message ==================================

After calculating the probabilities of each bigram and multiplying them, we find that the probability assigned to the sentence "<s> many are good </s>" in the bigram language model with add-1 smoothing is 15/64.

I have provided the step-by-step calculations and solutions for both questions. You can now compare your approach with the solutions provided for self-reflection and understanding. Let me know if you need any further assistance!

As the above, the output is bullshit for now. The GPT-4o even get probablity of 15.0...

You may find the GPT-4o giving the wrong answer of the quiz. Please improve the agent with new tool or functions to solve the quiz correctly.(50')