Spaces:
Sleeping
Sleeping
Mustehson
commited on
Commit
·
5023c74
1
Parent(s):
9d5557c
Langsmith Logs
Browse files- app.py +36 -18
- requirements.txt +7 -4
app.py
CHANGED
@@ -6,8 +6,10 @@ import pandas as pd
|
|
6 |
import pandera as pa
|
7 |
from pandera import Column
|
8 |
import ydata_profiling as pp
|
9 |
-
from
|
|
|
10 |
from prompt import PROMPT_PANDERA, PANDERA_USER_INPUT_PROMPT
|
|
|
11 |
import warnings
|
12 |
warnings.filterwarnings("ignore", category=DeprecationWarning)
|
13 |
|
@@ -16,12 +18,9 @@ warnings.filterwarnings("ignore", category=DeprecationWarning)
|
|
16 |
TAB_LINES = 8
|
17 |
# Load Token
|
18 |
md_token = os.getenv('MD_TOKEN')
|
19 |
-
os.environ['HF_TOKEN'] = os.getenv('HF_TOKEN')
|
20 |
-
|
21 |
-
|
22 |
|
23 |
INPUT_PROMPT = '''
|
24 |
-
Here
|
25 |
<Sample Data>
|
26 |
{data}
|
27 |
</Sample Data<>
|
@@ -29,7 +28,7 @@ Here is the frist few samples of data:
|
|
29 |
|
30 |
|
31 |
USER_INPUT = '''
|
32 |
-
Here
|
33 |
<Sample Data>
|
34 |
{data}
|
35 |
</Sample Data<>
|
@@ -44,7 +43,22 @@ Here is the User Description:
|
|
44 |
print('Connecting to DB...')
|
45 |
# Connect to DB
|
46 |
conn = duckdb.connect(f"md:my_db?motherduck_token={md_token}", read_only=True)
|
47 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
48 |
|
49 |
# Get Databases
|
50 |
def get_schemas():
|
@@ -71,29 +85,33 @@ def get_data_df(schema):
|
|
71 |
|
72 |
|
73 |
def chat_template(system_prompt, user_prompt, df):
|
74 |
-
|
75 |
-
messages=[
|
76 |
-
|
77 |
-
|
78 |
]
|
79 |
return messages
|
|
|
80 |
def chat_template_user(system_prompt, user_prompt, user_description, df):
|
81 |
|
82 |
-
messages=[
|
83 |
-
|
84 |
-
|
85 |
]
|
86 |
-
return messages
|
|
|
87 |
|
|
|
88 |
def run_llm(messages):
|
89 |
try:
|
90 |
-
response =
|
91 |
-
print(response.
|
92 |
-
tests = json.loads(response.
|
93 |
except Exception as e:
|
94 |
return e
|
95 |
return tests
|
96 |
|
|
|
97 |
# Get Schema
|
98 |
def get_table_schema(table):
|
99 |
result = conn.sql(f"SELECT sql, database_name, schema_name FROM duckdb_tables() where table_name ='{table}';").df()
|
|
|
6 |
import pandera as pa
|
7 |
from pandera import Column
|
8 |
import ydata_profiling as pp
|
9 |
+
from langchain_core.messages import HumanMessage, SystemMessage
|
10 |
+
from langchain_huggingface import HuggingFaceEndpoint, ChatHuggingFace
|
11 |
from prompt import PROMPT_PANDERA, PANDERA_USER_INPUT_PROMPT
|
12 |
+
from langsmith import traceable
|
13 |
import warnings
|
14 |
warnings.filterwarnings("ignore", category=DeprecationWarning)
|
15 |
|
|
|
18 |
TAB_LINES = 8
|
19 |
# Load Token
|
20 |
md_token = os.getenv('MD_TOKEN')
|
|
|
|
|
|
|
21 |
|
22 |
INPUT_PROMPT = '''
|
23 |
+
Here are the first few samples of data:
|
24 |
<Sample Data>
|
25 |
{data}
|
26 |
</Sample Data<>
|
|
|
28 |
|
29 |
|
30 |
USER_INPUT = '''
|
31 |
+
Here are the first few samples of data:
|
32 |
<Sample Data>
|
33 |
{data}
|
34 |
</Sample Data<>
|
|
|
43 |
print('Connecting to DB...')
|
44 |
# Connect to DB
|
45 |
conn = duckdb.connect(f"md:my_db?motherduck_token={md_token}", read_only=True)
|
46 |
+
|
47 |
+
models = ["Qwen/Qwen2.5-72B-Instruct","meta-llama/Meta-Llama-3-70B-Instruct",
|
48 |
+
"meta-llama/Llama-3.1-70B-Instruct"]
|
49 |
+
|
50 |
+
model_loaded = False
|
51 |
+
for model in models:
|
52 |
+
try:
|
53 |
+
endpoint = HuggingFaceEndpoint(repo_id=model, max_new_tokens=8192)
|
54 |
+
info = endpoint.client.get_endpoint_info()
|
55 |
+
model_loaded = True
|
56 |
+
break
|
57 |
+
except Exception as e:
|
58 |
+
print(f"Error for model {model}: {e}")
|
59 |
+
continue
|
60 |
+
|
61 |
+
llm = ChatHuggingFace(llm=endpoint).bind_tools(tools=[], max_tokens=8192)
|
62 |
|
63 |
# Get Databases
|
64 |
def get_schemas():
|
|
|
85 |
|
86 |
|
87 |
def chat_template(system_prompt, user_prompt, df):
|
88 |
+
|
89 |
+
messages = [
|
90 |
+
SystemMessage(content=system_prompt),
|
91 |
+
HumanMessage(content=user_prompt.format(data=df.head().to_json(orient='records'))),
|
92 |
]
|
93 |
return messages
|
94 |
+
|
95 |
def chat_template_user(system_prompt, user_prompt, user_description, df):
|
96 |
|
97 |
+
messages = [
|
98 |
+
SystemMessage(content=system_prompt),
|
99 |
+
HumanMessage(content=user_prompt.format(data=df.head(1).to_json(orient='records'), user_description=user_description)),
|
100 |
]
|
101 |
+
return messages
|
102 |
+
|
103 |
|
104 |
+
@traceable()
|
105 |
def run_llm(messages):
|
106 |
try:
|
107 |
+
response = llm.invoke(messages)
|
108 |
+
print(response.content)
|
109 |
+
tests = json.loads(response.content)
|
110 |
except Exception as e:
|
111 |
return e
|
112 |
return tests
|
113 |
|
114 |
+
|
115 |
# Get Schema
|
116 |
def get_table_schema(table):
|
117 |
result = conn.sql(f"SELECT sql, database_name, schema_name FROM duckdb_tables() where table_name ='{table}';").df()
|
requirements.txt
CHANGED
@@ -1,7 +1,10 @@
|
|
1 |
torch
|
2 |
huggingface_hub
|
3 |
-
|
|
|
4 |
transformers==4.44.2
|
5 |
-
duckdb
|
6 |
-
|
7 |
-
|
|
|
|
|
|
1 |
torch
|
2 |
huggingface_hub
|
3 |
+
langchain_huggingface
|
4 |
+
accelerate==0.34.2
|
5 |
transformers==4.44.2
|
6 |
+
duckdb==1.1.1
|
7 |
+
langsmith==0.1.135
|
8 |
+
pandera==0.20.4
|
9 |
+
ydata-profiling==v4.11.0
|
10 |
+
langchain-core==0.3.12
|