Update tools.py

tools.py

@@ -8,7 +8,7 @@ import regex as re
 import time
 import os
 from duckduckgo_search import DDGS
-
+from langchain_core.tools import tool
 DEFAULT_API_URL = "https://agents-course-unit4-scoring.hf.space"
 
 
@@ -38,20 +38,17 @@ def _download_file_for_task(task_id: str, ext: str) -> str:
     # If we get here, either 404 or download error
     return ""
 
-def ocr_image_tool(args: dict) -> str:
+@tool
+def image_tool(task_id: str) -> str:
     """
-    Expects: state["ocr_path"] is either:
-      • a local image path (e.g. "./hf_files/abc.png"), OR
-      • a Task ID (e.g. "abc123"), in which case we try downloading
-        GET {DEFAULT_API_URL}/files/{task_id} with .png/.jpg/.jpeg extensions.
-
+    Expects: task_id is a string
     Returns: "OCR text + brief caption or an error message"
       
     """
-    print("reached ocr_image_tool")
+    print("reached image_tool")
     # path_or_id = state.get("ocr_path", "")
     for ext in ("png", "jpg", "jpeg"):
-            candidate = _download_file_for_task(args["task_id"], ext)
+            candidate = _download_file_for_task(task_id, ext)
             if candidate:
                 local_img = candidate
                 break
@@ -122,19 +119,18 @@ def ocr_image_tool(args: dict) -> str:
     print("combined: ")
     return combined
 
-
-def parse_excel_tool(args: dict) -> str:
+@tool
+def excel_tool(task_id: str) -> str:
     """
     Downloads <task_id>.xlsx (if any) and returns a stringified list of
     records from the specified sheet.  No fallback to user-supplied tables.
-    Expected keys in `args`:
+    Expected keys in `task_id`:
         • task_id            – required (used to download the file)
-        • excel_sheet_name   – optional sheet to load
+
     returns: stringified list of records from the specified sheet
     """
-    print("reached parse_excel_tool")
-    task_id = args.get("task_id", "")
-    sheet   = args.get("excel_sheet_name", "")
+    print("reached excel_tool")
+    sheet   = "Sheet1"
 
     local_xlsx = _download_file_for_task(task_id, "xlsx")
     if not local_xlsx or not os.path.exists(local_xlsx):
@@ -153,27 +149,22 @@ def parse_excel_tool(args: dict) -> str:
     
 
 import openai
-
-def audio_transcriber_tool(args: dict) -> str:
+@tool
+def audio_transcriber_tool(task_id: str) -> str:
     """
     LangGraph tool for transcribing audio via OpenAI's Whisper API.
-    Expects: state["audio_path"] to be either:
-      • A local file path (e.g. "./hf_files/abc.mp3"), OR
-      • A Task ID (e.g. "abc123"), in which case we try downloading
-        GET {DEFAULT_API_URL}/files/{task_id} with .mp3, .wav, .m4a extensions.
+    Expects: task_id is a string
     Returns:
     "<text or error message>"
     Always attempts to download the file for the given path or task ID.
     """
     print("reached audio_transcriber_tool")
-    # path_or_id = state.get("audio_path", "")
-    # if not path_or_id:
-    #     return {}
+
 
     # Always attempt to download the file, regardless of local existence
     local_audio = ""
     for ext in ("mp3", "wav", "m4a"):
-        candidate = _download_file_for_task(args["task_id"], ext)
+        candidate = _download_file_for_task(task_id, ext)
         if candidate:
             local_audio = candidate
             break
@@ -205,16 +196,17 @@ def audio_transcriber_tool(args: dict) -> str:
 import re
 import requests
 
-def wikipedia_search_tool(args: dict) -> str:
+@tool
+def wikipedia_search_tool(wiki_query: str) -> str:
     """
     LangGraph wrapper for searching Wikipedia.
-    Expects: state["wiki_query"] to be a non‐empty string.
+    Expects: wiki_query is a non‐empty string.
     Returns: text summary of first matching page or an error message>"
       
     If no valid wiki_query is provided, returns {}.
     """
     print("reached wikipedia search tool")
-    query = args["wiki_query"]
+    query = wiki_query
     if not query:
         return {}
 
@@ -268,9 +260,10 @@ from langchain_openai import ChatOpenAI
 from langchain.schema import SystemMessage, HumanMessage
 LLM = ChatOpenAI(model_name="gpt-4.1-mini", temperature=0.2)
 
-def analyze_code_tool(args: dict) -> str:
+@tool
+def analyze_code_tool(task_id: str) -> str:
     """
-    Either args['snippet']  OR  (args['file'] + args['task_id'])
+    Either task_id  OR  (file + task_id)
     Reads the code (max 400 lines / 10 kB) and asks the LLM for:
         • plain-language summary
         • list of key functions/classes
@@ -279,15 +272,15 @@ def analyze_code_tool(args: dict) -> str:
     """
     print("reached analyze_code_tool")
     code_txt = ""
-    if args.get("snippet"):
-        code_txt = args["snippet"]
-    elif args.get("file") and args.get("task_id"):
-        path = _download_file_for_task(args["task_id"], "py")
+    if not task_id:
+        code_txt = "No code provided."
+    else:
+        path = _download_file_for_task(task_id, "py")
         if not path:
             return "Error: .py file not found for this task."
         code_txt = Path(path).read_text(encoding="utf-8", errors="ignore")
-    else:
-        return "Error: neither snippet nor file provided."
+    # else:
+    #     return "Error: neither snippet nor file provided."
 
     # Truncate for safety
     lines = code_txt.splitlines()[:400]
@@ -300,6 +293,7 @@ def analyze_code_tool(args: dict) -> str:
             "Summarise what it does, list key functions/classes, "
             "and point out any obvious bugs, performance issues or style problems.\n\n"
             f"```python\n{code_sample}\n```"
+            "If you can then find the output of the code and return it in the output."
         ))
     ]
     return LLM.invoke(prompt).content.strip()