Spaces:
Running
Running
Update app.py
Browse files
app.py
CHANGED
@@ -12,7 +12,7 @@ import pdfplumber
|
|
12 |
# ==== CONFIG ====
|
13 |
DEFAULT_API_URL = "https://agents-course-unit4-scoring.hf.space"
|
14 |
HF_TOKEN = os.getenv("HF_TOKEN")
|
15 |
-
GROK_API_KEY = os.getenv("GROK_API_KEY")
|
16 |
|
17 |
CONVERSATIONAL_MODELS = [
|
18 |
"deepseek-ai/DeepSeek-LLM",
|
@@ -25,7 +25,7 @@ wiki_api = wikipediaapi.Wikipedia(language="en", user_agent="SmartAgent/1.0 (cho
|
|
25 |
# ==== UTILITY: Link/file detection ====
|
26 |
def extract_links(text):
|
27 |
url_pattern = re.compile(r'(https?://[^\s\)\],]+)')
|
28 |
-
return url_pattern.findall(text)
|
29 |
|
30 |
def download_file(url, out_dir="tmp_files"):
|
31 |
os.makedirs(out_dir, exist_ok=True)
|
@@ -70,7 +70,7 @@ def analyze_file(file_path):
|
|
70 |
return f"TXT error: {e}"
|
71 |
else:
|
72 |
return f"Unsupported file type: {file_path}"
|
73 |
-
|
74 |
def analyze_webpage(url):
|
75 |
try:
|
76 |
r = requests.get(url, timeout=15)
|
@@ -108,36 +108,40 @@ def llm_conversational(query):
|
|
108 |
hf_client = InferenceClient(model_id, token=HF_TOKEN)
|
109 |
# Try conversational if available, else fallback to text_generation
|
110 |
if hasattr(hf_client, "conversational"):
|
111 |
-
|
112 |
-
|
113 |
-
|
114 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
115 |
if isinstance(result, dict) and "generated_text" in result:
|
116 |
return result["generated_text"]
|
117 |
-
elif hasattr(result, "generated_text"):
|
118 |
-
return result.generated_text
|
119 |
elif isinstance(result, str):
|
120 |
return result
|
121 |
-
|
122 |
-
|
123 |
-
result = hf_client.text_generation(query, max_new_tokens=384)
|
124 |
-
if isinstance(result, dict) and "generated_text" in result:
|
125 |
-
return result["generated_text"]
|
126 |
-
elif isinstance(result, str):
|
127 |
-
return result
|
128 |
except Exception as e:
|
129 |
last_error = f"{model_id}: {e}"
|
130 |
return None
|
131 |
|
132 |
def is_coding_question(text):
|
133 |
-
# Basic heuristic: mentions code, function, "python", code blocks, etc.
|
134 |
code_terms = [
|
135 |
"python", "java", "c++", "code", "function", "write a", "script", "algorithm",
|
136 |
"bug", "traceback", "error", "output", "compile", "debug"
|
137 |
]
|
138 |
-
if any(term in text.lower() for term in code_terms):
|
139 |
return True
|
140 |
-
if re.search(r"```.+```", text, re.DOTALL):
|
141 |
return True
|
142 |
return False
|
143 |
|
@@ -160,9 +164,8 @@ def grok_completion(question, system_prompt=None):
|
|
160 |
r = requests.post(url, headers=headers, json=payload, timeout=45)
|
161 |
r.raise_for_status()
|
162 |
data = r.json()
|
163 |
-
# Extract assistant's reply
|
164 |
return data['choices'][0]['message']['content']
|
165 |
-
except Exception
|
166 |
return None
|
167 |
|
168 |
# ==== SMART AGENT ====
|
@@ -181,6 +184,8 @@ class SmartAgent:
|
|
181 |
if local:
|
182 |
file_analysis = analyze_file(local)
|
183 |
results.append(f"File ({url}):\n{file_analysis}")
|
|
|
|
|
184 |
else:
|
185 |
results.append(analyze_webpage(url))
|
186 |
if results:
|
|
|
12 |
# ==== CONFIG ====
|
13 |
DEFAULT_API_URL = "https://agents-course-unit4-scoring.hf.space"
|
14 |
HF_TOKEN = os.getenv("HF_TOKEN")
|
15 |
+
GROK_API_KEY = os.getenv("GROK_API_KEY")
|
16 |
|
17 |
CONVERSATIONAL_MODELS = [
|
18 |
"deepseek-ai/DeepSeek-LLM",
|
|
|
25 |
# ==== UTILITY: Link/file detection ====
|
26 |
def extract_links(text):
|
27 |
url_pattern = re.compile(r'(https?://[^\s\)\],]+)')
|
28 |
+
return url_pattern.findall(text or "")
|
29 |
|
30 |
def download_file(url, out_dir="tmp_files"):
|
31 |
os.makedirs(out_dir, exist_ok=True)
|
|
|
70 |
return f"TXT error: {e}"
|
71 |
else:
|
72 |
return f"Unsupported file type: {file_path}"
|
73 |
+
|
74 |
def analyze_webpage(url):
|
75 |
try:
|
76 |
r = requests.get(url, timeout=15)
|
|
|
108 |
hf_client = InferenceClient(model_id, token=HF_TOKEN)
|
109 |
# Try conversational if available, else fallback to text_generation
|
110 |
if hasattr(hf_client, "conversational"):
|
111 |
+
try:
|
112 |
+
result = hf_client.conversational(
|
113 |
+
messages=[{"role": "user", "content": query}],
|
114 |
+
max_new_tokens=384,
|
115 |
+
)
|
116 |
+
if isinstance(result, dict) and "generated_text" in result:
|
117 |
+
return result["generated_text"]
|
118 |
+
elif hasattr(result, "generated_text"):
|
119 |
+
return result.generated_text
|
120 |
+
elif isinstance(result, str):
|
121 |
+
return result
|
122 |
+
except Exception:
|
123 |
+
pass
|
124 |
+
# Fallback to text_generation
|
125 |
+
try:
|
126 |
+
result = hf_client.text_generation(query, max_new_tokens=384)
|
127 |
if isinstance(result, dict) and "generated_text" in result:
|
128 |
return result["generated_text"]
|
|
|
|
|
129 |
elif isinstance(result, str):
|
130 |
return result
|
131 |
+
except Exception:
|
132 |
+
pass
|
|
|
|
|
|
|
|
|
|
|
133 |
except Exception as e:
|
134 |
last_error = f"{model_id}: {e}"
|
135 |
return None
|
136 |
|
137 |
def is_coding_question(text):
|
|
|
138 |
code_terms = [
|
139 |
"python", "java", "c++", "code", "function", "write a", "script", "algorithm",
|
140 |
"bug", "traceback", "error", "output", "compile", "debug"
|
141 |
]
|
142 |
+
if any(term in (text or "").lower() for term in code_terms):
|
143 |
return True
|
144 |
+
if re.search(r"```.+```", text or "", re.DOTALL):
|
145 |
return True
|
146 |
return False
|
147 |
|
|
|
164 |
r = requests.post(url, headers=headers, json=payload, timeout=45)
|
165 |
r.raise_for_status()
|
166 |
data = r.json()
|
|
|
167 |
return data['choices'][0]['message']['content']
|
168 |
+
except Exception:
|
169 |
return None
|
170 |
|
171 |
# ==== SMART AGENT ====
|
|
|
184 |
if local:
|
185 |
file_analysis = analyze_file(local)
|
186 |
results.append(f"File ({url}):\n{file_analysis}")
|
187 |
+
else:
|
188 |
+
results.append(f"Could not download file: {url}")
|
189 |
else:
|
190 |
results.append(analyze_webpage(url))
|
191 |
if results:
|