Update app.py

app.py

@@ -1,16 +1,21 @@
-import os, json, time
+# app.py  – handles images, txt/py, PDFs, any file…
+
+import os, json, time, io, mimetypes
 from functools import lru_cache
 
 import gradio as gr
 import requests
 import pandas as pd
 from openai import OpenAI, RateLimitError, APIError
-from duckduckgo_search import DDGS   # add to requirements.txt
+from duckduckgo_search import DDGS
+from PyPDF2 import PdfReader   # <- new dependency
 
-# ────────────────────────────────
 DEFAULT_API_URL = "https://agents-course-unit4-scoring.hf.space"
 OPENAI_MODEL    = "gpt-4o-mini"
-# ────────────────────────────────
+TEXT_CHAR_LIMIT = 8_000
+PDF_PAGE_LIMIT  = 3
+
+# ───────────────────────── helpers ─────────────────────────
 def duckduckgo_search(query: str, max_results: int = 5) -> str:
     bullets = []
     with DDGS() as ddgs:
@@ -24,62 +29,84 @@ DDG_SCHEMA = {
     "parameters": {
         "type": "object",
         "properties": {
-            "query":       {"type": "string"},
+            "query": {"type": "string"},
             "max_results": {"type": "integer", "default": 5},
         },
         "required": ["query"],
     },
 }
 
-# ╭──────────────────────────────────────────────────────────╮
-# │ AGENT (now supports optional image_url)                 │
-# ╰──────────────────────────────────────────────────────────╯
-class GPT4oMiniAgentWithDDG:
+def fetch_text_file(url: str) -> str:
+    try:
+        txt = requests.get(url, timeout=15).text
+        return txt[:TEXT_CHAR_LIMIT]
+    except Exception as e:
+        return f"[Could not download text file: {e}]"
+
+def fetch_pdf_text(url: str) -> str:
+    try:
+        resp = requests.get(url, timeout=20)
+        resp.raise_for_status()
+        reader = PdfReader(io.BytesIO(resp.content))
+        pages = []
+        for i, page in enumerate(reader.pages[:PDF_PAGE_LIMIT]):
+            pages.append(page.extract_text() or "")
+        return ("\n\n".join(pages))[:TEXT_CHAR_LIMIT]
+    except Exception as e:
+        return f"[Could not read PDF: {e}]"
+
+# ─────────────────────────── agent ─────────────────────────
+class GPT4oMiniAgentWithFiles:
     def __init__(self, retries:int = 3, backoff:float = 2.0):
-        api_key = os.getenv("OPENAI_API_KEY")
-        if not api_key:
-            raise EnvironmentError("Add OPENAI_API_KEY in your Space secrets!")
-        self.client   = OpenAI(api_key=api_key)
-        self.retries  = retries
-        self.backoff  = backoff
-        self.prompt   = (
+        key = os.getenv("OPENAI_API_KEY")
+        if not key:
+            raise EnvironmentError("OPENAI_API_KEY missing in Secrets.")
+        self.client, self.retries, self.backoff = OpenAI(api_key=key), retries, backoff
+        self.sys_prompt = (
             "You are a concise, accurate assistant. "
-            "If certain, answer immediately; otherwise call duckduckgo_search."
+            "If certain, answer directly; otherwise call duckduckgo_search."
         )
 
     @lru_cache(maxsize=512)
-    def __call__(self, question: str, image_url: str | None = None) -> str:
+    def __call__(self, question:str, file_url:str|None=None) -> str:
         user_content = [{"type": "text", "text": question}]
-        if image_url:
-            user_content.append(
-                {"type": "image_url", "image_url": {"url": image_url}}
-            )
+        if file_url:
+            kind = (file_url.split("?")[0].split("#")[0]).lower()
+            ext  = os.path.splitext(kind)[1]
+            if ext in {".png", ".jpg", ".jpeg", ".gif", ".webp"}:
+                user_content.append({"type":"image_url","image_url":{"url":file_url}})
+            elif ext == ".pdf":
+                text = fetch_pdf_text(file_url)
+                user_content.append({"type":"text","text": f"(PDF extract)\n{text}"})
+            elif ext in {".txt", ".py", ".md", ".json", ".csv", ".html"}:
+                text = fetch_text_file(file_url)
+                user_content.append({"type":"text","text": f"(File content)\n{text}"})
+            else:
+                user_content.append({"type":"text","text": f"[File available here] {file_url}"})
 
         msgs = [
-            {"role": "system", "content": self.prompt},
-            {"role": "user",   "content": user_content},
+            {"role":"system","content":self.sys_prompt},
+            {"role":"user","content":user_content},
         ]
 
-        # 1st pass – model may request the tool
         resp = self._chat(msgs, tools=[DDG_SCHEMA], tool_choice="auto")
 
-        # Run tool(s) if requested
         if resp.choices[0].message.tool_calls:
             for call in resp.choices[0].message.tool_calls:
                 args = json.loads(call.function.arguments or "{}")
-                tool_out = duckduckgo_search(**args) if call.function.name=="duckduckgo_search" else ""
+                tool_out = duckduckgo_search(**args)
                 msgs.append({
-                    "role": "tool",
-                    "tool_call_id": call.id,
-                    "name": call.function.name,
-                    "content": tool_out
+                    "role":"tool",
+                    "tool_call_id":call.id,
+                    "name":call.function.name,
+                    "content":tool_out,
                 })
             resp = self._chat(msgs)
 
         return resp.choices[0].message.content.strip()
 
     def _chat(self, messages, **kw):
-        for i in range(1, self.retries + 1):
+        for i in range(1, self.retries+1):
             try:
                 return self.client.chat.completions.create(
                     model=OPENAI_MODEL,
@@ -92,52 +119,39 @@ class GPT4oMiniAgentWithDDG:
                 time.sleep(self.backoff * i)
         raise RuntimeError("OpenAI API failed after retries.")
 
-# ╭──────────────────────────────────────────────────────────╮
-# │ RUN  +  SUBMIT                                          │
-# ╰──────────────────────────────────────────────────────────╯
-def run_and_submit_all(profile: gr.OAuthProfile | None):
+# ────────────────── run + submit ──────────────────
+def run_and_submit_all(profile: gr.OAuthProfile|None):
     if not profile:
         return "Please log in ↑", None
     username = profile.username
-    agent = GPT4oMiniAgentWithDDG()
-    space_id = os.getenv("SPACE_ID", "local")
+    agent = GPT4oMiniAgentWithFiles()
+    space_id = os.getenv("SPACE_ID","local")
     agent_code = f"https://huggingface.co/spaces/{space_id}/tree/main"
 
-    # ① Fetch
     qs = requests.get(f"{DEFAULT_API_URL}/questions", timeout=15).json()
 
-    # ② Answer
-    answers, rows = [], []
+    rows, answers = [], []
     for item in qs:
-        qid   = item["task_id"]
-        text  = item["question"]
-        img   = item.get("filename")          # <-- NEW
-        ans   = agent(text, img)
+        qid  = item["task_id"]
+        text = item["question"]
+        file_url = item.get("filename") or item.get("file_url")
+        ans  = agent(text, file_url)
         answers.append({"task_id": qid, "submitted_answer": ans})
-        rows.append({"Task ID": qid, "Question": text, "Image URL": img or "", "Answer": ans})
-
-    # ③ Submit
-    payload = {
-        "username": username,
-        "agent_code": agent_code,
-        "answers": answers
-    }
-    res = requests.post(f"{DEFAULT_API_URL}/submit", json=payload, timeout=60).json()
-    status = f"Score {res['score']} %  ({res['correct_count']}/{res['total_attempted']})"
+        rows.append({"Task ID": qid, "Question": text, "File": file_url or "", "Answer": ans})
 
+    payload = {"username": username, "agent_code": agent_code, "answers": answers}
+    res = requests.post(f"{DEFAULT_API_URL}/submit", json=payload, timeout=60).json()
+    status = f"Score {res['score']} % ({res['correct_count']}/{res['total_attempted']})"
     return status, pd.DataFrame(rows)
 
-# ╭──────────────────────────────────────────────────────────╮
-# │ GRADIO UI                                               │
-# ╰──────────────────────────────────────────────────────────╯
+# ────────────────────── UI ───────────────────────
 with gr.Blocks() as demo:
-    gr.Markdown("# Unit-4 Agent Runner – Image Ready")
+    gr.Markdown("# Unit-4 Agent – handles images, text/code files & PDFs")
     gr.LoginButton()
-    run_btn      = gr.Button("Run Evaluation & Submit All Answers")
-    status_box   = gr.Textbox(label="Status", interactive=False)
-    results_grid = gr.DataFrame(label="Log", wrap=True)
-
-    run_btn.click(run_and_submit_all, outputs=[status_box, results_grid])
+    btn = gr.Button("Run Evaluation & Submit All Answers")
+    status = gr.Textbox(label="Status", interactive=False)
+    table  = gr.DataFrame(label="Log", wrap=True)
+    btn.click(run_and_submit_all, outputs=[status, table])
 
 if __name__ == "__main__":
     demo.launch(debug=True, share=False)