Update app.py
Browse files
app.py
CHANGED
@@ -1,3 +1,5 @@
|
|
|
|
|
|
1 |
import sys
|
2 |
import os
|
3 |
import pandas as pd
|
@@ -12,16 +14,18 @@ import logging
|
|
12 |
|
13 |
# Logging
|
14 |
logging.basicConfig(level=logging.INFO, format='%(asctime)s - %(levelname)s - %(message)s')
|
15 |
-
logger = logging.getLogger(
|
16 |
|
17 |
# Cleanup
|
|
|
18 |
def cleanup():
|
19 |
if dist.is_initialized():
|
20 |
logger.info("Cleaning up PyTorch distributed process group")
|
21 |
dist.destroy_process_group()
|
|
|
22 |
atexit.register(cleanup)
|
23 |
|
24 |
-
#
|
25 |
persistent_dir = "/data/hf_cache"
|
26 |
os.makedirs(persistent_dir, exist_ok=True)
|
27 |
model_cache_dir = os.path.join(persistent_dir, "txagent_models")
|
@@ -33,6 +37,7 @@ for d in [model_cache_dir, tool_cache_dir, file_cache_dir, report_dir]:
|
|
33 |
os.environ["HF_HOME"] = model_cache_dir
|
34 |
os.environ["TRANSFORMERS_CACHE"] = model_cache_dir
|
35 |
|
|
|
36 |
sys.path.insert(0, os.path.abspath(os.path.join(os.path.dirname(__file__), "src")))
|
37 |
from txagent.txagent import TxAgent
|
38 |
|
@@ -41,15 +46,18 @@ MAX_CHUNK_TOKENS = 8192
|
|
41 |
MAX_NEW_TOKENS = 2048
|
42 |
PROMPT_OVERHEAD = 500
|
43 |
|
|
|
44 |
def clean_response(text: str) -> str:
|
45 |
text = re.sub(r"\[.*?\]|\bNone\b", "", text, flags=re.DOTALL)
|
46 |
text = re.sub(r"\n{3,}", "\n\n", text)
|
47 |
text = re.sub(r"[^\n#\-\*\w\s\.,:\(\)]+", "", text)
|
48 |
return text.strip()
|
49 |
|
|
|
50 |
def estimate_tokens(text: str) -> int:
|
51 |
return len(text) // 3.5 + 1
|
52 |
|
|
|
53 |
def extract_text_from_excel(file_obj: Union[str, Dict[str, Any]]) -> str:
|
54 |
if isinstance(file_obj, dict) and 'name' in file_obj:
|
55 |
file_path = file_obj['name']
|
@@ -71,6 +79,7 @@ def extract_text_from_excel(file_obj: Union[str, Dict[str, Any]]) -> str:
|
|
71 |
logger.warning(f"Failed to parse {sheet}: {e}")
|
72 |
return "\n".join(all_text)
|
73 |
|
|
|
74 |
def split_text_into_chunks(text: str) -> List[str]:
|
75 |
lines = text.split("\n")
|
76 |
chunks, current, current_tokens = [], [], 0
|
@@ -87,6 +96,7 @@ def split_text_into_chunks(text: str) -> List[str]:
|
|
87 |
chunks.append("\n".join(current))
|
88 |
return chunks
|
89 |
|
|
|
90 |
def build_prompt_from_text(chunk: str) -> str:
|
91 |
return f"""
|
92 |
### Clinical Records Analysis
|
@@ -105,17 +115,36 @@ Please analyze these clinical notes and provide:
|
|
105 |
Provide a structured response with clear medical reasoning.
|
106 |
"""
|
107 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
108 |
def init_agent() -> TxAgent:
|
109 |
new_tool_path = os.path.join(tool_cache_dir, "new_tool.json")
|
110 |
if not os.path.exists(new_tool_path):
|
111 |
with open(new_tool_path, 'w') as f:
|
112 |
-
json.dump({
|
113 |
-
"name": "new_tool",
|
114 |
-
"description": "Default tool",
|
115 |
-
"tools": [{"name": "dummy_tool", "description": "test", "version": "1.0"}]
|
116 |
-
}, f)
|
117 |
|
118 |
-
|
119 |
'opentarget': '/home/user/.pyenv/versions/3.10.17/lib/python3.10/site-packages/tooluniverse/data/opentarget_tools.json',
|
120 |
'fda_drug_label': '/home/user/.pyenv/versions/3.10.17/lib/python3.10/site-packages/tooluniverse/data/fda_drug_labeling_tools.json',
|
121 |
'special_tools': '/home/user/.pyenv/versions/3.10.17/lib/python3.10/site-packages/tooluniverse/data/special_tools.json',
|
@@ -123,34 +152,19 @@ def init_agent() -> TxAgent:
|
|
123 |
'new_tool': new_tool_path
|
124 |
}
|
125 |
|
126 |
-
|
127 |
-
for name,
|
128 |
-
|
129 |
-
|
130 |
-
|
131 |
-
if isinstance(data, dict) and 'tools' in data:
|
132 |
-
tools = data['tools']
|
133 |
-
elif isinstance(data, list):
|
134 |
-
tools = data
|
135 |
-
elif isinstance(data, dict) and 'name' in data:
|
136 |
-
tools = [data]
|
137 |
-
else:
|
138 |
-
logger.warning(f"Skipping {name}: bad structure")
|
139 |
-
continue
|
140 |
-
if all(isinstance(t, dict) and 'name' in t for t in tools):
|
141 |
-
validated[name] = path
|
142 |
-
else:
|
143 |
-
logger.warning(f"Skipping {name}: items malformed")
|
144 |
-
except Exception as e:
|
145 |
-
logger.error(f"Invalid tool {name}: {e}")
|
146 |
|
147 |
-
if not
|
148 |
-
raise ValueError("No valid tools
|
149 |
|
150 |
agent = TxAgent(
|
151 |
model_name="mims-harvard/TxAgent-T1-Llama-3.1-8B",
|
152 |
rag_model_name="mims-harvard/ToolRAG-T1-GTE-Qwen2-1.5B",
|
153 |
-
tool_files_dict=
|
154 |
force_finish=True,
|
155 |
enable_checker=True,
|
156 |
step_rag_num=4,
|
@@ -159,42 +173,42 @@ def init_agent() -> TxAgent:
|
|
159 |
agent.init_model()
|
160 |
return agent
|
161 |
|
|
|
162 |
def stream_report(agent: TxAgent, input_file: Union[str, Dict[str, Any]], full_output: str) -> Generator[Tuple[str, Union[str, None], str], None, None]:
|
163 |
accumulated = ""
|
164 |
-
if input_file is None:
|
165 |
-
yield "❌ Upload an Excel file.", None, ""
|
166 |
-
return
|
167 |
try:
|
|
|
|
|
|
|
168 |
text = extract_text_from_excel(input_file)
|
169 |
chunks = split_text_into_chunks(text)
|
170 |
-
|
171 |
-
|
172 |
-
|
173 |
-
|
174 |
-
|
175 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
176 |
for out in agent.run_gradio_chat(
|
177 |
-
message=
|
178 |
max_new_tokens=MAX_NEW_TOKENS, max_token=MAX_MODEL_TOKENS,
|
179 |
-
call_agent=False, conversation=[]
|
180 |
-
|
181 |
-
|
182 |
-
|
183 |
-
|
184 |
-
|
185 |
-
|
186 |
-
|
187 |
-
|
188 |
-
|
189 |
-
|
190 |
-
call_agent=False, conversation=[]
|
191 |
-
):
|
192 |
-
summary += out if isinstance(out, str) else out.content
|
193 |
-
final = clean_response(summary)
|
194 |
-
report_path = os.path.join(report_dir, f"report_{datetime.now().strftime('%Y%m%d_%H%M%S')}.md")
|
195 |
-
with open(report_path, 'w') as f:
|
196 |
-
f.write(f"# Clinical Report\n\n{final}")
|
197 |
-
yield f"{accumulated}\n\n📊 Final Summary:\n{final}", report_path, final
|
198 |
|
199 |
def create_ui(agent: TxAgent) -> gr.Blocks:
|
200 |
with gr.Blocks(theme=gr.themes.Soft()) as demo:
|
@@ -211,6 +225,7 @@ def create_ui(agent: TxAgent) -> gr.Blocks:
|
|
211 |
analyze_btn.click(fn=stream_report, inputs=[file_upload, full_output], outputs=[report_output, report_file, full_output])
|
212 |
return demo
|
213 |
|
|
|
214 |
if __name__ == "__main__":
|
215 |
try:
|
216 |
agent = init_agent()
|
|
|
1 |
+
# ✅ Fully updated app.py for TxAgent with strict tool validation to prevent runtime errors
|
2 |
+
|
3 |
import sys
|
4 |
import os
|
5 |
import pandas as pd
|
|
|
14 |
|
15 |
# Logging
|
16 |
logging.basicConfig(level=logging.INFO, format='%(asctime)s - %(levelname)s - %(message)s')
|
17 |
+
logger = logging.getLogger("app")
|
18 |
|
19 |
# Cleanup
|
20 |
+
|
21 |
def cleanup():
|
22 |
if dist.is_initialized():
|
23 |
logger.info("Cleaning up PyTorch distributed process group")
|
24 |
dist.destroy_process_group()
|
25 |
+
|
26 |
atexit.register(cleanup)
|
27 |
|
28 |
+
# Directories
|
29 |
persistent_dir = "/data/hf_cache"
|
30 |
os.makedirs(persistent_dir, exist_ok=True)
|
31 |
model_cache_dir = os.path.join(persistent_dir, "txagent_models")
|
|
|
37 |
os.environ["HF_HOME"] = model_cache_dir
|
38 |
os.environ["TRANSFORMERS_CACHE"] = model_cache_dir
|
39 |
|
40 |
+
# Import TxAgent
|
41 |
sys.path.insert(0, os.path.abspath(os.path.join(os.path.dirname(__file__), "src")))
|
42 |
from txagent.txagent import TxAgent
|
43 |
|
|
|
46 |
MAX_NEW_TOKENS = 2048
|
47 |
PROMPT_OVERHEAD = 500
|
48 |
|
49 |
+
|
50 |
def clean_response(text: str) -> str:
|
51 |
text = re.sub(r"\[.*?\]|\bNone\b", "", text, flags=re.DOTALL)
|
52 |
text = re.sub(r"\n{3,}", "\n\n", text)
|
53 |
text = re.sub(r"[^\n#\-\*\w\s\.,:\(\)]+", "", text)
|
54 |
return text.strip()
|
55 |
|
56 |
+
|
57 |
def estimate_tokens(text: str) -> int:
|
58 |
return len(text) // 3.5 + 1
|
59 |
|
60 |
+
|
61 |
def extract_text_from_excel(file_obj: Union[str, Dict[str, Any]]) -> str:
|
62 |
if isinstance(file_obj, dict) and 'name' in file_obj:
|
63 |
file_path = file_obj['name']
|
|
|
79 |
logger.warning(f"Failed to parse {sheet}: {e}")
|
80 |
return "\n".join(all_text)
|
81 |
|
82 |
+
|
83 |
def split_text_into_chunks(text: str) -> List[str]:
|
84 |
lines = text.split("\n")
|
85 |
chunks, current, current_tokens = [], [], 0
|
|
|
96 |
chunks.append("\n".join(current))
|
97 |
return chunks
|
98 |
|
99 |
+
|
100 |
def build_prompt_from_text(chunk: str) -> str:
|
101 |
return f"""
|
102 |
### Clinical Records Analysis
|
|
|
115 |
Provide a structured response with clear medical reasoning.
|
116 |
"""
|
117 |
|
118 |
+
|
119 |
+
def clean_and_rewrite_tool_file(original_path: str, cleaned_path: str) -> bool:
|
120 |
+
try:
|
121 |
+
with open(original_path, "r") as f:
|
122 |
+
data = json.load(f)
|
123 |
+
if isinstance(data, dict) and "tools" in data:
|
124 |
+
tools = data["tools"]
|
125 |
+
elif isinstance(data, list):
|
126 |
+
tools = data
|
127 |
+
elif isinstance(data, dict) and "name" in data:
|
128 |
+
tools = [data]
|
129 |
+
else:
|
130 |
+
return False
|
131 |
+
if not all(isinstance(t, dict) and "name" in t for t in tools):
|
132 |
+
return False
|
133 |
+
with open(cleaned_path, "w") as out:
|
134 |
+
json.dump(tools, out)
|
135 |
+
return True
|
136 |
+
except Exception as e:
|
137 |
+
logger.error(f"Failed to clean tool {original_path}: {e}")
|
138 |
+
return False
|
139 |
+
|
140 |
+
|
141 |
def init_agent() -> TxAgent:
|
142 |
new_tool_path = os.path.join(tool_cache_dir, "new_tool.json")
|
143 |
if not os.path.exists(new_tool_path):
|
144 |
with open(new_tool_path, 'w') as f:
|
145 |
+
json.dump([{"name": "dummy_tool", "description": "test", "version": "1.0"}], f)
|
|
|
|
|
|
|
|
|
146 |
|
147 |
+
raw_tool_files = {
|
148 |
'opentarget': '/home/user/.pyenv/versions/3.10.17/lib/python3.10/site-packages/tooluniverse/data/opentarget_tools.json',
|
149 |
'fda_drug_label': '/home/user/.pyenv/versions/3.10.17/lib/python3.10/site-packages/tooluniverse/data/fda_drug_labeling_tools.json',
|
150 |
'special_tools': '/home/user/.pyenv/versions/3.10.17/lib/python3.10/site-packages/tooluniverse/data/special_tools.json',
|
|
|
152 |
'new_tool': new_tool_path
|
153 |
}
|
154 |
|
155 |
+
validated_paths = {}
|
156 |
+
for name, original_path in raw_tool_files.items():
|
157 |
+
cleaned_path = os.path.join(tool_cache_dir, f"{name}_cleaned.json")
|
158 |
+
if clean_and_rewrite_tool_file(original_path, cleaned_path):
|
159 |
+
validated_paths[name] = cleaned_path
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
160 |
|
161 |
+
if not validated_paths:
|
162 |
+
raise ValueError("No valid tools found after sanitizing.")
|
163 |
|
164 |
agent = TxAgent(
|
165 |
model_name="mims-harvard/TxAgent-T1-Llama-3.1-8B",
|
166 |
rag_model_name="mims-harvard/ToolRAG-T1-GTE-Qwen2-1.5B",
|
167 |
+
tool_files_dict=validated_paths,
|
168 |
force_finish=True,
|
169 |
enable_checker=True,
|
170 |
step_rag_num=4,
|
|
|
173 |
agent.init_model()
|
174 |
return agent
|
175 |
|
176 |
+
|
177 |
def stream_report(agent: TxAgent, input_file: Union[str, Dict[str, Any]], full_output: str) -> Generator[Tuple[str, Union[str, None], str], None, None]:
|
178 |
accumulated = ""
|
|
|
|
|
|
|
179 |
try:
|
180 |
+
if input_file is None:
|
181 |
+
yield "❌ Upload a valid Excel file.", None, ""
|
182 |
+
return
|
183 |
text = extract_text_from_excel(input_file)
|
184 |
chunks = split_text_into_chunks(text)
|
185 |
+
for i, chunk in enumerate(chunks):
|
186 |
+
prompt = build_prompt_from_text(chunk)
|
187 |
+
result = ""
|
188 |
+
for out in agent.run_gradio_chat(
|
189 |
+
message=prompt, history=[], temperature=0.2,
|
190 |
+
max_new_tokens=MAX_NEW_TOKENS, max_token=MAX_MODEL_TOKENS,
|
191 |
+
call_agent=False, conversation=[]):
|
192 |
+
result += out if isinstance(out, str) else out.content
|
193 |
+
cleaned = clean_response(result)
|
194 |
+
accumulated += f"\n\n📄 Part {i+1}:\n{cleaned}"
|
195 |
+
yield accumulated, None, ""
|
196 |
+
summary_prompt = f"Summarize this analysis:\n\n{accumulated}"
|
197 |
+
summary = ""
|
198 |
for out in agent.run_gradio_chat(
|
199 |
+
message=summary_prompt, history=[], temperature=0.2,
|
200 |
max_new_tokens=MAX_NEW_TOKENS, max_token=MAX_MODEL_TOKENS,
|
201 |
+
call_agent=False, conversation=[]):
|
202 |
+
summary += out if isinstance(out, str) else out.content
|
203 |
+
final = clean_response(summary)
|
204 |
+
report_path = os.path.join(report_dir, f"report_{datetime.now().strftime('%Y%m%d_%H%M%S')}.md")
|
205 |
+
with open(report_path, 'w') as f:
|
206 |
+
f.write(f"# Clinical Report\n\n{final}")
|
207 |
+
yield f"{accumulated}\n\n📊 Final Summary:\n{final}", report_path, final
|
208 |
+
except Exception as e:
|
209 |
+
logger.error(f"Stream error: {e}", exc_info=True)
|
210 |
+
yield f"❌ Error: {str(e)}", None, ""
|
211 |
+
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
212 |
|
213 |
def create_ui(agent: TxAgent) -> gr.Blocks:
|
214 |
with gr.Blocks(theme=gr.themes.Soft()) as demo:
|
|
|
225 |
analyze_btn.click(fn=stream_report, inputs=[file_upload, full_output], outputs=[report_output, report_file, full_output])
|
226 |
return demo
|
227 |
|
228 |
+
|
229 |
if __name__ == "__main__":
|
230 |
try:
|
231 |
agent = init_agent()
|