xadrez

analyse_chess_position_tool.py

@@ -2,7 +2,7 @@ from langchain_core.tools import tool
 import requests
 
 @tool
-def get_best_move(fen: str) -> str:
+def get_chess_best_move(fen: str) -> str:
     """
     Given the description of a chess board using FEN notation, returns the next best move. 
     
@@ -16,8 +16,7 @@ def get_best_move(fen: str) -> str:
     CHESS_MOVE_API = "https://chess-api.com/v1"
     url = CHESS_MOVE_API
     payload = {
-       "fen": fen,
-       "depth": 1
+       "fen": fen
     }
 
     print(f"Buscando melhor jogada em {CHESS_MOVE_API} - {payload}")

app.py

@@ -84,8 +84,7 @@ def run_and_submit_all( profile: gr.OAuthProfile | None):
                 BASE_DIR = os.path.dirname(os.path.abspath(__file__))
                 if file_name.endswith(('.mp3', '.xlsx', '.png')):
                     file_path = os.path.join(BASE_DIR, 'files', f'{file_name}.b64')
-                    base64_attatched_file = get_text_file_contents(file_path)
-                    question_text_for_agent += f'. The content in base64 of the attatched file mentioned in the question is the following: {base64_attatched_file}'
+                    question_text_for_agent += f'. The path to the base64 contents of the attatched file mentioned in the question is the following: {file_path}'
                 else:
                     file_path = os.path.join(BASE_DIR, 'files', file_name)
                     plain_txt_file = get_text_file_contents(file_path)

constants.py

@@ -1,7 +1,7 @@
 import os
 
-# MODEL = 'gemini-2.5-pro-preview-05-06'
-MODEL = 'gemini-2.0-flash'
+MODEL = 'gemini-2.5-pro-preview-05-06'
+# MODEL = 'gemini-2.0-flash'
 # MODEL = 'gemini-2.0-flash-exp'
 # MODEL = 'gemini-2.5-pro-exp-03-25'
 
@@ -13,6 +13,8 @@ TAVILY_KEY = os.getenv("TAVILY_KEY")
 
 PROMPT_LIMITADOR_LLM = """
   You are a general AI assistant. I will ask you a question. Report your thoughts, and finish your answer with the following template: FINAL ANSWER: [YOUR FINAL ANSWER]. YOUR FINAL ANSWER should be a number OR as few words as possible OR a comma separated list of numbers and/or strings. If you are asked for a number, don't use comma to write your number neither use units such as $ or percent sign unless specified otherwise. If you are asked for a string, don't use articles, neither abbreviations (e.g. for cities), and write the digits in plain text unless specified otherwise. If you are asked for a comma separated list, apply the above rules depending of whether the element to be put in the list is a number or a string.
+  
+  You are INCAPABLE of analysing images. You MUST use tools for that.
 
   However, if I give you tools that access the internet, you may freely use them. In particular, you have a tool that allows you to search the internet. Use it when necessary.
 

convert_chessboard_image_to_fen_tool.py

@@ -0,0 +1,31 @@
+from langchain_core.tools import tool
+from image_to_text_tool import image_to_text
+from utils import get_base64
+from typing import Literal
+
+@tool
+def convert_chessboard_image_to_fen(chessboard_image_base64_path: str, current_player: Literal["black", "white"]) -> str:
+    """
+    Given the path to a file with a chessboard image in base64, returns the FEN description of the chessboard.
+
+    Args:
+        image_path: Path to the image file.
+        current_player: Whose turn it is to play. Must be either 'black' or 'white'.
+    Returns:
+        JSON with FEN (Forsyth-Edwards Notation) string representing the current board position.
+    """
+
+    chessboard_image_base64_str = get_base64(chessboard_image_base64_path)
+
+    print("Calling chessboard description")
+
+    fen = image_to_text.invoke(
+        {
+        "image_base64_str": chessboard_image_base64_str, 
+        "instructions": f'Return the fen notation of this chessboard. Assume {current_player} will play.'
+        }
+    )
+
+    print("FEN returned: " + fen)
+
+    #gabarito do FEN: 3r2k1/pp3pp1/4b2p/7Q/3n4/PqBBR2P/5PP1/6K1 b - - 0 1

excel_parser_tool.py

@@ -7,9 +7,11 @@ from utils import get_base64
 
 
 @tool
-def parse_excel(base64_str: str) -> str:
+def parse_excel(base64_filepath: str) -> str:
     """Parses a base64-encoded Excel (.xlsx) file and returns a markdown summary."""
     try:
+        base64_str = get_text_file_contents(file_path)
+        
         file_bytes = base64.b64decode(base64_str)
         excel_file = pd.ExcelFile(io.BytesIO(file_bytes))
 

langchain_agent.py

@@ -2,7 +2,7 @@ import re
 import constants
 import time
 from langchain_google_genai import ChatGoogleGenerativeAI
-from langchain.agents import AgentExecutor, create_tool_calling_agent
+from langchain.agents import AgentExecutor, create_tool_calling_agent, create_openai_functions_agent
 from langchain_core.prompts import ChatPromptTemplate, MessagesPlaceholder
 from langchain_core.messages import SystemMessage
 
@@ -13,14 +13,16 @@ from image_to_text_tool import image_to_text
 from internet_search_tool import internet_search
 from botanical_classification_tool import get_botanical_classification
 from excel_parser_tool import parse_excel
-from analyse_chess_position_tool import get_best_move
+from analyse_chess_position_tool import get_chess_best_move
+from convert_chessboard_image_to_fen_tool import convert_chessboard_image_to_fen
+
 
 class LangChainAgent:
     def __init__(self):
         llm = ChatGoogleGenerativeAI(
             model=constants.MODEL,
             api_key=constants.API_KEY, 
-            temperature=0.5,
+            temperature=0.0,
             timeout=20)
 
         tools = [
@@ -30,7 +32,7 @@ class LangChainAgent:
             internet_search,
             get_botanical_classification,
             parse_excel,
-            get_best_move
+            get_chess_best_move
         ]
 
         prompt = ChatPromptTemplate.from_messages([
@@ -40,7 +42,7 @@ class LangChainAgent:
             MessagesPlaceholder(variable_name="agent_scratchpad"),
         ])
 
-        agent = create_tool_calling_agent(llm, tools, prompt=prompt)
+        agent = create_openai_functions_agent(llm, tools, prompt=prompt)
         self.executor = AgentExecutor(agent=agent, tools=tools, verbose=True)
 
     def __call__(self, question: str) -> str: