File size: 7,736 Bytes
ff30164
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
from dotenv import load_dotenv
from langchain_core.messages import (
    BaseMessage,
    HumanMessage,
    ToolMessage,
)
import base64
from langchain_core.prompts import ChatPromptTemplate, MessagesPlaceholder
from langgraph.graph import END, StateGraph, START
from typing import Annotated, List
from langchain_community.tools import DuckDuckGoSearchRun
from langchain_core.tools import tool
from langchain_experimental.utilities import PythonREPL
import operator
from typing import Annotated, Sequence, TypedDict
from langchain_groq import ChatGroq
import functools
from langchain_core.messages import AIMessage
from langchain_google_genai import ChatGoogleGenerativeAI
from langgraph.prebuilt import ToolNode
from typing import Literal
import gradio as gr
import io
import PIL

load_dotenv()
llm_coder = ChatGroq(temperature=0, model_name="llama-3.1-8b-instant")
llm_image = ChatGoogleGenerativeAI(
    model="gemini-1.5-flash",
    temperature=0,
    max_tokens=None,
    timeout=None,
    max_retries=2,
)

search_tool = DuckDuckGoSearchRun()
repl_tool = PythonREPL()

@tool
def python_repl(
    code: Annotated[str, "The python code to execute to answer the question."],
):
    """Use this to execute python code. If you want to see the output of a value,
    you should print it out with `print(...)`. This is visible to the user."""
    try:
        result = repl_tool.run(code)
    except BaseException as e:
        return f"Failed to execute. Error: {repr(e)}"
    result_str = f"Successfully executed:\n```python\n{code}\n```\nStdout: {result}"
    return (
        result_str + "\n\nIf you have completed all tasks, respond with FINAL ANSWER."
    )

def create_agent(llm, tools, system_message: str):
    """Create an agent."""
    prompt = ChatPromptTemplate.from_messages(
        [
            (
                "system",
                "You are a helpful AI assistant, collaborating with other assistants."
                " Use the provided tools to progress towards answering the question."
                " If you are unable to fully answer, that's OK, another assistant with different tools "
                " will help where you left off. Execute what you can to make progress."
                " If you or any of the other assistants have the final answer or deliverable,"
                " prefix your response with FINAL ANSWER so the team knows to stop."
                " You have access to the following tools: {tool_names}.\n{system_message}",
            ),
            MessagesPlaceholder(variable_name="messages"),
        ]
    )
    prompt = prompt.partial(system_message=system_message)
    prompt = prompt.partial(tool_names=", ".join([tool.name for tool in tools]))
    return prompt | llm.bind_tools(tools)

class AgentState(TypedDict):
    messages: Annotated[Sequence[BaseMessage], operator.add]
    sender: str

def agent_node(state, agent, name):
    result = agent.invoke(state)
    if isinstance(result, ToolMessage):
        pass
    else:
        result = AIMessage(**result.dict(exclude={"type", "name"}), name=name)
    return {
        "messages": [result],
        "sender": name,
    }    

problem_agent = create_agent(
    llm_image,
    [],
    system_message="You should understand the problem properly and provide a clear description with the edge cases, don't provide the solution, after completing all tasks."
)
problem_node = functools.partial(agent_node, agent=problem_agent, name="problem_agent")

solution_agent = create_agent(
    llm_image,
    [],
    system_message="after understanding the problem, you should provide a solution to the problem in python that is clear and concise and solves all edge cases, also provide intuition behind the solution."
)
solution_node = functools.partial(agent_node, agent=solution_agent, name="solution_agent")

checker_agent = create_agent(
    llm_coder,
    [],
    system_message="critically analyze the solution provided by the solution agent, check for correctness, efficiency, and edge cases, if the solution is correct, provide a message saying so, if not, provide a message with the error and suggest a fix."
)

def checker_node(state):
    text_only_messages = []
    for msg in state["messages"]:
        if isinstance(msg.content, list):
            text_content = [item["text"] for item in msg.content if item["type"] == "text"]
            new_msg = msg.copy()
            new_msg.content = " ".join(text_content)
            text_only_messages.append(new_msg)
        else:
            text_only_messages.append(msg)
    
    text_only_state = {
        "messages": text_only_messages,
        "sender": state["sender"]
    }
    
    result = checker_agent.invoke(text_only_state)
    if isinstance(result, ToolMessage):
        pass
    else:
        result = AIMessage(**result.dict(exclude={"type", "name"}), name="checker_agent")
    return {
        "messages": [result],
        "sender": "checker_agent",
    }

tools = [search_tool, python_repl]
tool_node = ToolNode(tools)

def router(state) -> Literal["call_tool", "__end__", "continue"]:
    messages = state["messages"]
    last_message = messages[-1]
    if last_message.tool_calls:
        return "call_tool"
    if "FINAL ANSWER" in last_message.content:
        return "__end__"
    return "continue"

workflow = StateGraph(AgentState)

workflow.add_node("problem_creator", problem_node)
workflow.add_node("solution_generator", solution_node)
workflow.add_node("checker_agent", checker_node)
workflow.add_node("call_tool", tool_node)

workflow.add_conditional_edges(
    "problem_creator",
    router,
    {"continue": "solution_generator", "call_tool": "call_tool", "__end__": END},
)
workflow.add_conditional_edges(
    "solution_generator",
    router,
    {"continue": "checker_agent", "call_tool": "call_tool", "__end__": END},
)
workflow.add_conditional_edges(
    "checker_agent",
    router,
    {"continue": "problem_creator", "call_tool": "call_tool", "__end__": END},
)
workflow.add_conditional_edges(
    "call_tool",
    lambda x: x["sender"],
    {
        "problem_creator": "problem_creator",
        "solution_generator": "solution_generator",
        "checker_agent": "checker_agent",
    },
)
workflow.add_edge(START, "problem_creator")

graph = workflow.compile()

def process_images(images: List[tuple[PIL.Image.Image, str | None]]):
    if not images:
        return "No images uploaded"
    
    # Convert all images to base64
    image_contents = []
    for (image, _) in images:
        buffered = io.BytesIO()
        image.save(buffered, format="PNG")
        img_str = base64.b64encode(buffered.getvalue()).decode()
        image_contents.append({
            "type": "image_url",
            "image_url": {"url": f"data:image/png;base64,{img_str}"}
        })
    
    # Create the input for the workflow
    input_data = {"messages": [HumanMessage(
        content = [
            {"type": "text", "text": "answer the question about the following images"},
            *image_contents
        ]
    )]}
    
    # Run the workflow
    output = []
    try:
        for chunk in graph.stream(input_data, {"recursion_limit": 10}, stream_mode="values"):
            message = chunk["messages"][-1]
            output.append(f"{message.name}: {message.content}")
    except Exception as e:
        output.append(f"Error: {repr(e)}")        
    
    return "\n\n".join(output)

# Create Gradio interface
iface = gr.Interface(
    fn=process_images,
    inputs=[gr.Gallery(label="Upload an image", type="pil")],
    outputs=[gr.Markdown(label="Output", show_copy_button=True)],
    title="Image Question Answering",
    description="Upload an image to get it processed and answered."
)

# Launch the interface
iface.launch()