Spaces:
Sleeping
Sleeping
Update app.py
Browse files
app.py
CHANGED
@@ -192,21 +192,45 @@ COLUMN_SYNONYMS = {
|
|
192 |
|
193 |
|
194 |
# Fuzzy matcher for mapping query terms to dataset columns
|
195 |
-
def fuzzy_match_columns(query
|
196 |
query = query.lower()
|
197 |
all_synonyms = {synonym: col for col, synonyms in COLUMN_SYNONYMS.items() for synonym in synonyms}
|
198 |
-
|
199 |
words = query.replace("and", "").replace("vs", "").replace("by", "").split()
|
200 |
-
|
201 |
matched_columns = []
|
202 |
for word in words:
|
203 |
-
matches = get_close_matches(word, all_synonyms.keys(), n=
|
204 |
-
for match in matches
|
205 |
-
matched_columns.append(all_synonyms[match])
|
206 |
|
207 |
return list(dict.fromkeys(matched_columns))
|
208 |
|
209 |
-
#
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
210 |
def add_stats_to_figure(fig, df, y_axis):
|
211 |
min_salary = df[y_axis].min()
|
212 |
max_salary = df[y_axis].max()
|
@@ -222,18 +246,25 @@ def add_stats_to_figure(fig, df, y_axis):
|
|
222 |
)
|
223 |
return fig
|
224 |
|
225 |
-
#
|
226 |
-
def generate_visual_from_query(query, df):
|
227 |
try:
|
228 |
matched_columns = fuzzy_match_columns(query)
|
229 |
|
230 |
-
#
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
231 |
if len(matched_columns) >= 2:
|
232 |
x_axis, group_by = matched_columns[0], matched_columns[1]
|
233 |
elif len(matched_columns) == 1:
|
234 |
x_axis, group_by = matched_columns[0], None
|
235 |
else:
|
236 |
-
st.warning("β No matching columns found.
|
237 |
return None
|
238 |
|
239 |
# Handle distribution queries
|
@@ -258,16 +289,15 @@ def generate_visual_from_query(query, df):
|
|
258 |
title=f"Salary Trend Over Years by {x_axis.replace('_', ' ').title()}")
|
259 |
return add_stats_to_figure(fig, df, "salary_in_usd")
|
260 |
|
261 |
-
# Handle remote work
|
262 |
elif "remote" in query:
|
263 |
grouped_df = df.groupby(["remote_ratio"] + ([group_by] if group_by else []))["salary_in_usd"].mean().reset_index()
|
264 |
fig = px.bar(grouped_df, x="remote_ratio", y="salary_in_usd", color=group_by,
|
265 |
title="Remote Work Impact on Salary")
|
266 |
return add_stats_to_figure(fig, df, "salary_in_usd")
|
267 |
|
268 |
-
# Default behavior if query doesn't match anything specific
|
269 |
else:
|
270 |
-
st.warning("
|
271 |
return None
|
272 |
|
273 |
except Exception as e:
|
@@ -275,6 +305,7 @@ def generate_visual_from_query(query, df):
|
|
275 |
return None
|
276 |
|
277 |
|
|
|
278 |
# SQL-RAG Analysis
|
279 |
if st.session_state.df is not None:
|
280 |
temp_dir = tempfile.TemporaryDirectory()
|
|
|
192 |
|
193 |
|
194 |
# Fuzzy matcher for mapping query terms to dataset columns
|
195 |
+
def fuzzy_match_columns(query):
|
196 |
query = query.lower()
|
197 |
all_synonyms = {synonym: col for col, synonyms in COLUMN_SYNONYMS.items() for synonym in synonyms}
|
|
|
198 |
words = query.replace("and", "").replace("vs", "").replace("by", "").split()
|
199 |
+
|
200 |
matched_columns = []
|
201 |
for word in words:
|
202 |
+
matches = get_close_matches(word, all_synonyms.keys(), n=1, cutoff=0.6)
|
203 |
+
matched_columns.extend([all_synonyms[match] for match in matches])
|
|
|
204 |
|
205 |
return list(dict.fromkeys(matched_columns))
|
206 |
|
207 |
+
# Ask LLM to suggest relevant columns if fuzzy matching fails
|
208 |
+
def ask_llm_for_columns(query, llm, df):
|
209 |
+
columns = ', '.join(df.columns)
|
210 |
+
prompt = f"""
|
211 |
+
Analyze this user query and suggest the most relevant dataset columns for visualization.
|
212 |
+
|
213 |
+
Query: "{query}"
|
214 |
+
|
215 |
+
Available Columns: {columns}
|
216 |
+
|
217 |
+
Respond in this JSON format:
|
218 |
+
{{
|
219 |
+
"x_axis": "column_name",
|
220 |
+
"y_axis": "column_name",
|
221 |
+
"group_by": "optional_column_name"
|
222 |
+
}}
|
223 |
+
"""
|
224 |
+
|
225 |
+
response = llm.generate(prompt)
|
226 |
+
try:
|
227 |
+
suggestion = json.loads(response)
|
228 |
+
return suggestion
|
229 |
+
except json.JSONDecodeError:
|
230 |
+
st.error("β οΈ Failed to interpret AI response. Please refine your query.")
|
231 |
+
return None
|
232 |
+
|
233 |
+
# Add min, max, and average salary annotations to the chart
|
234 |
def add_stats_to_figure(fig, df, y_axis):
|
235 |
min_salary = df[y_axis].min()
|
236 |
max_salary = df[y_axis].max()
|
|
|
246 |
)
|
247 |
return fig
|
248 |
|
249 |
+
# Unified visualization function with LLM fallback
|
250 |
+
def generate_visual_from_query(query, df, llm=None):
|
251 |
try:
|
252 |
matched_columns = fuzzy_match_columns(query)
|
253 |
|
254 |
+
# Fallback to LLM if fuzzy matching fails
|
255 |
+
if not matched_columns and llm:
|
256 |
+
st.info("π€ No match found. Asking AI for suggestions...")
|
257 |
+
suggestion = ask_llm_for_columns(query, llm, df)
|
258 |
+
if suggestion:
|
259 |
+
matched_columns = [suggestion.get("x_axis"), suggestion.get("group_by")]
|
260 |
+
|
261 |
+
# Handle cases when we have columns to plot
|
262 |
if len(matched_columns) >= 2:
|
263 |
x_axis, group_by = matched_columns[0], matched_columns[1]
|
264 |
elif len(matched_columns) == 1:
|
265 |
x_axis, group_by = matched_columns[0], None
|
266 |
else:
|
267 |
+
st.warning("β No matching columns found. Please refine your query.")
|
268 |
return None
|
269 |
|
270 |
# Handle distribution queries
|
|
|
289 |
title=f"Salary Trend Over Years by {x_axis.replace('_', ' ').title()}")
|
290 |
return add_stats_to_figure(fig, df, "salary_in_usd")
|
291 |
|
292 |
+
# Handle remote work impact
|
293 |
elif "remote" in query:
|
294 |
grouped_df = df.groupby(["remote_ratio"] + ([group_by] if group_by else []))["salary_in_usd"].mean().reset_index()
|
295 |
fig = px.bar(grouped_df, x="remote_ratio", y="salary_in_usd", color=group_by,
|
296 |
title="Remote Work Impact on Salary")
|
297 |
return add_stats_to_figure(fig, df, "salary_in_usd")
|
298 |
|
|
|
299 |
else:
|
300 |
+
st.warning("β οΈ No suitable visualization generated. Please refine your query.")
|
301 |
return None
|
302 |
|
303 |
except Exception as e:
|
|
|
305 |
return None
|
306 |
|
307 |
|
308 |
+
|
309 |
# SQL-RAG Analysis
|
310 |
if st.session_state.df is not None:
|
311 |
temp_dir = tempfile.TemporaryDirectory()
|