phase2 / app.py
triflix's picture
Update app.py
1befb1d verified
# app.py
from fastapi import FastAPI, UploadFile, File, HTTPException, Query
from fastapi.middleware.cors import CORSMiddleware
from fastapi.responses import JSONResponse
import pandas as pd
import os
import json
import tempfile
import hashlib
import uuid
from datetime import datetime, timezone
from typing import Optional, List
from pydantic import BaseModel
from google import genai
from google.genai import types
import logging
import asyncio
from concurrent.futures import ThreadPoolExecutor
from functools import partial
import re
import traceback
import motor.motor_asyncio
# ----------------------------
# Configuration (deployment-ready)
# ----------------------------
MONGO_URI = os.getenv("MONGO_URI", "mongodb+srv://curseofwitcher:curseofwitcher@aianalyticsdata.btxby1j.mongodb.net/?retryWrites=true&w=majority&appName=aiAnalyticsData")
DB_NAME = os.getenv("DB_NAME", "data_analysis")
SNAPSHOT_BUCKET = os.getenv("SNAPSHOT_DIR", "/tmp/snapshots")
os.makedirs(SNAPSHOT_BUCKET, exist_ok=True)
MAX_UPLOAD_SIZE = int(os.getenv("MAX_UPLOAD_SIZE_BYTES", 200 * 1024 * 1024)) # 200MB default
METADATA_ONLY_FALLBACK = os.getenv("METADATA_ONLY_FALLBACK", "true").lower() == "true"
TTL_DAYS = int(os.getenv("SNAPSHOT_TTL_DAYS", "0")) # 0 = no TTL
EXECUTOR_WORKERS = int(os.getenv("EXECUTOR_WORKERS", "2"))
# Setup logging
logging.basicConfig(level=logging.INFO)
logger = logging.getLogger("app")
# FastAPI app
app = FastAPI(title="Data Analysis API", version="1.0.0")
app.add_middleware(
CORSMiddleware,
allow_origins=["*"], # keep as-is for deployment per user instruction
allow_credentials=True,
allow_methods=["*"],
allow_headers=["*"],
)
# Mongo client (async)
mongo_client = motor.motor_asyncio.AsyncIOMotorClient(MONGO_URI)
db = mongo_client[DB_NAME]
snapshots = db.snapshots # collection will be created automatically on insert
# Thread pool for blocking tasks
EXECUTOR = ThreadPoolExecutor(max_workers=EXECUTOR_WORKERS)
# Ensure /tmp exists (older app used /tmp)
os.makedirs("/tmp", exist_ok=True)
# ---------- Models ----------
class AnalysisResponse(BaseModel):
summary: dict
chart_data: dict
metadata: dict
class ErrorResponse(BaseModel):
error: str
details: Optional[str] = None
class DrillRequest(BaseModel):
snapshot_id: str
filter_column: str
filter_value: str
limit: Optional[int] = 100
offset: Optional[int] = 0
highlight_columns: Optional[List[str]] = None
# ---------- Utilities ----------
def sha256_bytes(data: bytes) -> str:
h = hashlib.sha256()
h.update(data)
return h.hexdigest()
def sha256_text(text: str) -> str:
return sha256_bytes(text.encode("utf-8"))
def sha256_obj(obj) -> str:
text = json.dumps(obj, sort_keys=True, default=str)
return sha256_text(text)
def canonical_types(df: pd.DataFrame) -> dict:
def map_type(dtype):
if pd.api.types.is_integer_dtype(dtype) or pd.api.types.is_float_dtype(dtype):
return "numeric"
if pd.api.types.is_datetime64_any_dtype(dtype):
return "datetime"
return "object"
return {col: map_type(dtype) for col, dtype in df.dtypes.items()}
def _find_balanced_json(s: str):
first = s.find('{')
if first == -1:
return None
stack = []
for i in range(first, len(s)):
ch = s[i]
if ch == '{':
stack.append('{')
elif ch == '}':
if not stack:
return None
stack.pop()
if not stack:
return s[first:i+1]
return None
def _escape_problematic_backslashes(s: str) -> str:
return re.sub(r'\\(?!["\\/bfnrtu])', r'\\\\', s)
def safe_json_loads(raw_text: str):
try:
return json.loads(raw_text)
except Exception as e1:
err1 = str(e1)
subset = _find_balanced_json(raw_text)
if subset:
try:
return json.loads(subset)
except Exception as e2:
err2 = str(e2)
else:
err2 = "no balanced braces found"
try:
fixed = _escape_problematic_backslashes(subset or raw_text)
return json.loads(fixed)
except Exception as e3:
err3 = str(e3)
diagnostic = {
"direct_error": err1,
"subset_error": err2,
"escaped_error": err3,
"raw_snippet": (raw_text[:4000] + '...') if len(raw_text) > 4000 else raw_text
}
raise ValueError("Unable to parse JSON from model output. Diagnostic: " + json.dumps(diagnostic))
def data_fingerprint(df: pd.DataFrame, n_sample_rows: int = 100) -> str:
df2 = df.copy()
df2 = df2.reindex(sorted(df2.columns), axis=1)
head = df2.head(n_sample_rows).to_json(orient="split", date_format="iso", force_ascii=False)
tail = df2.tail(n_sample_rows).to_json(orient="split", date_format="iso", force_ascii=False)
col_aggs = {c: {"nunique": int(df2[c].nunique()), "nulls": int(df2[c].isnull().sum())} for c in df2.columns}
text = head + tail + json.dumps(col_aggs, sort_keys=True, default=str)
return hashlib.sha256(text.encode("utf-8")).hexdigest()
def stream_save_and_hash(upload_file: UploadFile, tmp_path: str, size_limit: Optional[int] = None) -> str:
h = hashlib.sha256()
total = 0
upload_file.file.seek(0)
with open(tmp_path, "wb") as f:
while True:
chunk = upload_file.file.read(8192)
if not chunk:
break
f.write(chunk)
h.update(chunk)
total += len(chunk)
if size_limit and total > size_limit:
raise HTTPException(status_code=413, detail="Uploaded file exceeds maximum allowed size")
return h.hexdigest()
async def save_preprocessed_df(df: pd.DataFrame, snapshot_id: str) -> str:
path = os.path.join(SNAPSHOT_BUCKET, f"{snapshot_id}.csv")
loop = asyncio.get_running_loop()
# Use functools.partial to pass keyword args to to_csv (handles pandas 3.0+ keyword-only changes)
await loop.run_in_executor(EXECUTOR, partial(df.to_csv, path, index=False))
return path
def load_file_from_path(file_path: str, original_filename: str) -> pd.DataFrame:
ext = os.path.splitext(original_filename)[-1].lower()
if ext == ".csv":
return pd.read_csv(file_path)
elif ext in [".xls", ".xlsx"]:
return pd.read_excel(file_path, sheet_name=0)
else:
raise ValueError(f"Unsupported file type: {ext}")
def preprocess(df: pd.DataFrame, drop_thresh=0.5) -> pd.DataFrame:
df = df.copy()
df.columns = [str(c).strip().lower().replace(" ", "_") for c in df.columns]
df = df.loc[:, df.isnull().mean() < drop_thresh]
for col in df.columns:
if pd.api.types.is_numeric_dtype(df[col]):
df.loc[:, col] = df[col].fillna(df[col].median())
elif pd.api.types.is_datetime64_any_dtype(df[col]):
df.loc[:, col] = df[col].fillna(pd.Timestamp('1970-01-01'))
else:
df.loc[:, col] = df[col].fillna("Unknown")
for col in df.columns:
if df[col].dtype == 'object':
try:
df.loc[:, col] = pd.to_numeric(df[col])
except Exception:
pass
df = df.drop_duplicates()
return df
def get_metadata(df: pd.DataFrame) -> dict:
return {
"rows": int(df.shape[0]),
"columns": int(df.shape[1]),
"column_names": list(df.columns),
"column_types": {col: str(dtype) for col, dtype in df.dtypes.items()},
"unique_values": {col: int(df[col].nunique()) for col in df.columns}
}
# ---------- AI generation (blocking) ----------
def generate_summary_blocking(meta, fiverow, system_prompt_override: Optional[str] = None):
# using provided API key (kept as-is per deployment)
api_key = os.getenv("GEMINI_API_KEY") or "AIzaSyB1jgGCuzg7ELPwNEEwaluQZoZhxhgLmAs"
if not api_key:
raise RuntimeError("GEMINI_API_KEY environment variable not set")
client = genai.Client(api_key=api_key)
model = "gemini-2.5-flash-lite"
system_prompt = system_prompt_override or """
You are a strict JSON generator.
Input contains:
- meta: dataframe metadata
- fiverow: first 5 records of dataframe
You must output JSON with the following structure:
{
"summary": "<short natural language overview of dataset>",
"recommended_charts": [
{
"type": "<one of: bar, pie, timeseries, histogram, scatter, multiple_columns, stacked_bar, heatmap>",
"title": "<short title for chart>",
"columns": ["<col1>", "<col2>", "..."],
"python_code": "<full runnable Python code using seaborn/matplotlib that produces the chart>"
},
...
]
}
Mandatory rules:
- Always produce syntactically valid JSON ONLY. No text outside the JSON object.
- Provide at least these chart types somewhere in recommended_charts: bar, pie, timeseries, histogram, scatter, multiple_columns, stacked_bar, heatmap.
- Use only column names that appear in meta['column_names'].
- The python_code string must be self-contained and runnable assuming a variable `df` exists containing the full cleaned DataFrame. Start the code with imports:
import pandas as pd
import seaborn as sns
import matplotlib.pyplot as plt
and include any necessary preprocessing steps (e.g., parsing dates).
- For timeseries charts ensure the datetime column is parsed (`pd.to_datetime`) before plotting.
- For multiple_columns provide a pairplot or facetgrid example that uses up to 4 numeric columns or sensible categorical splits.
- For stacked_bar, show aggregation code (groupby + unstack) and plotting with df.plot(kind='bar', stacked=True).
- For heatmap, compute correlation matrix and plot sns.heatmap with annotations.
- For pie charts, ensure grouping/aggregation when there are >20 unique categories (group small categories into 'Other').
- For histogram and scatter include axis labels and tight_layout; include plt.show() at the end.
- Keep code minimal but complete so a user can copy-paste and run (assume seaborn, matplotlib, pandas installed).
- For each chart add a sensible "columns" list showing which columns the code uses.
- Do not include examples using columns not present in meta.
- Do not include more than 10 recommended_charts.
- Ensure strings inside the JSON are escaped properly so the JSON parses.
Produce concise natural-language one-line summary in "summary". Ensure JSON is parseable by json.loads in Python.
"""
user_prompt = {"meta": meta, "fiverow": fiverow}
contents = [
types.Content(
role="user",
parts=[types.Part.from_text(text=str(user_prompt))],
),
]
generate_content_config = types.GenerateContentConfig(
thinking_config=types.ThinkingConfig(thinking_budget=0),
response_mime_type="application/json",
system_instruction=[types.Part.from_text(text=system_prompt)],
)
raw = ""
try:
for chunk in client.models.generate_content_stream(
model=model,
contents=contents,
config=generate_content_config,
):
if chunk.text:
raw += chunk.text
except Exception as e:
logger.error("AI generation stream error: %s\n%s", str(e), traceback.format_exc())
raise RuntimeError("AI generation failed: " + str(e))
logger.debug("AI raw output (trimmed): %s", raw[:2000])
try:
parsed = safe_json_loads(raw)
except Exception as e:
logger.error("Failed to parse AI JSON. Raw (trimmed): %s", raw[:2000])
raise RuntimeError(f"AI JSON parse error: {e}")
if not isinstance(parsed, dict) or "summary" not in parsed or "recommended_charts" not in parsed:
logger.error("AI output missing required keys. Parsed keys: %s", list(parsed.keys()) if isinstance(parsed, dict) else type(parsed))
raise RuntimeError("AI output missing required keys: 'summary' and 'recommended_charts' required")
return parsed
async def generate_summary_async(meta, fiverow, system_prompt_override: Optional[str] = None):
loop = asyncio.get_running_loop()
return await loop.run_in_executor(EXECUTOR, generate_summary_blocking, meta, fiverow, system_prompt_override)
# ---------- Chart data extraction (accept parsed dict) ----------
def flatten_columns(df):
if isinstance(df.columns, pd.MultiIndex):
df.columns = ['_'.join(map(str, col)).strip() for col in df.columns.values]
return df
def extract_chart_data_json_by_type(parsed_summary: dict, df: pd.DataFrame):
try:
result = {}
for chart in parsed_summary.get("recommended_charts", []):
chart_type = chart.get("type")
columns = chart.get("columns", []) or []
title = chart.get("title", "unnamed_chart")
if chart_type not in result:
result[chart_type] = []
try:
if chart_type == "bar":
df_agg = df[columns].groupby(columns[0]).sum(numeric_only=True).reset_index()
chart_data = df_agg.to_dict(orient="records")
elif chart_type == "stacked_bar":
df_agg = df.groupby(columns).sum(numeric_only=True).unstack()
df_agg = flatten_columns(df_agg)
chart_data = df_agg.fillna(0).to_dict(orient="records")
elif chart_type == "pie":
col = columns[0]
counts = df[col].value_counts()
if len(counts) > 20:
top = counts.nlargest(19)
others = counts.iloc[19:].sum()
counts = pd.concat([top, pd.Series({'Other': others})])
chart_data = counts.reset_index().rename(columns={'index': col, col: 'value'}).to_dict(orient="records")
elif chart_type == "histogram":
chart_data = df[columns[0]].dropna().tolist()
elif chart_type == "scatter":
chart_data = df[columns].to_dict(orient="records")
elif chart_type == "timeseries":
df_copy = df[columns].copy()
for c in columns:
if not pd.api.types.is_datetime64_any_dtype(df_copy[c]):
df_copy[c] = pd.to_datetime(df_copy[c], errors='coerce')
chart_data = df_copy.astype(str).to_dict(orient="records")
elif chart_type == "multiple_columns":
chart_data = df[columns].to_dict(orient="records")
elif chart_type == "heatmap":
corr_df = df[columns].corr().fillna(0)
chart_data = flatten_columns(corr_df).to_dict()
else:
chart_data = []
except Exception as e:
chart_data = {"error": str(e)}
result[chart_type].append({"title": title, "data": chart_data})
return result
except Exception as e:
logger.error("Error extracting chart data: %s\n%s", str(e), traceback.format_exc())
raise RuntimeError(f"Error extracting chart data: {e}")
# ---------- Startup indexes ----------
@app.on_event("startup")
async def create_indexes():
try:
await snapshots.create_index("file_hash")
await snapshots.create_index("data_hash")
await snapshots.create_index("meta_hash")
await snapshots.create_index("snapshot_id", unique=True)
if TTL_DAYS > 0:
await snapshots.create_index("created_at_dt", expireAfterSeconds=TTL_DAYS * 24 * 3600)
logger.info("Mongo indexes ensured.")
except Exception:
logger.exception("Error creating indexes")
# ---------- Routes ----------
@app.get("/")
async def root():
return {"message": "Data Analysis API is running"}
@app.get("/health")
async def health_check():
return {"status": "healthy"}
@app.post("/analyze", response_model=AnalysisResponse)
async def analyze_data(file: UploadFile = File(...)):
if not file.filename:
raise HTTPException(status_code=400, detail="No file provided")
allowed_extensions = ['.csv', '.xls', '.xlsx']
file_ext = os.path.splitext(file.filename)[-1].lower()
if file_ext not in allowed_extensions:
raise HTTPException(status_code=400, detail=f"Unsupported file type. Allowed: {', '.join(allowed_extensions)}")
# Temp file path for streaming upload
with tempfile.NamedTemporaryFile(delete=False, suffix=file_ext) as tmp_file:
tmp_path = tmp_file.name
try:
# stream save with size guard and compute file hash
try:
file_hash = stream_save_and_hash(file, tmp_path, size_limit=MAX_UPLOAD_SIZE)
except HTTPException:
try:
os.unlink(tmp_path)
except Exception:
pass
raise
except Exception as e:
try:
os.unlink(tmp_path)
except Exception:
pass
logger.exception("Error saving uploaded file")
raise HTTPException(status_code=500, detail=str(e))
# Load and preprocess
try:
df = load_file_from_path(tmp_path, file.filename)
except Exception as e:
logger.exception("Error loading file")
raise HTTPException(status_code=400, detail=str(e))
try:
df_clean = preprocess(df)
except Exception as e:
logger.exception("Error preprocessing file")
raise HTTPException(status_code=500, detail=str(e))
meta = get_metadata(df_clean)
fiverow = df_clean.head(5).to_dict(orient="records")
# compute hashes
data_hash = data_fingerprint(df_clean)
meta_hash = sha256_obj({
"rows": meta["rows"],
"columns": meta["columns"],
"column_names": meta["column_names"],
"column_types": canonical_types(df_clean),
})
# search order: file_hash -> data_hash -> meta_hash
existing = await snapshots.find_one({"file_hash": file_hash})
cache_hit = None
if not existing:
existing = await snapshots.find_one({"data_hash": data_hash})
if existing:
cache_hit = "data"
if not existing and METADATA_ONLY_FALLBACK:
existing = await snapshots.find_one({"meta_hash": meta_hash})
if existing:
cache_hit = "meta"
if existing:
snapshot_id_return = existing.get("snapshot_id") or str(existing.get("_id"))
summary = existing.get("summary") or {}
chart_data = existing.get("chart_data") or {}
metadata = existing.get("metadata") or meta
return AnalysisResponse(summary=summary, chart_data=chart_data, metadata=metadata)
# create snapshot doc in processing state
snapshot_id = uuid.uuid4().hex
created_at_iso = datetime.now(timezone.utc).isoformat()
created_at_dt = datetime.now(timezone.utc)
doc = {
"snapshot_id": snapshot_id,
"filename": file.filename,
"file_hash": file_hash,
"data_hash": data_hash,
"meta_hash": meta_hash,
"metadata": meta,
"summary": None,
"chart_data": None,
"preprocessed_path": None,
"status": "processing",
"created_at": created_at_iso,
"created_at_dt": created_at_dt,
}
await snapshots.insert_one(doc)
# generate AI summary (offloaded)
try:
summary_obj = await generate_summary_async(meta, fiverow)
except Exception as e:
await snapshots.update_one({"snapshot_id": snapshot_id}, {"$set": {"status": "failed", "error": str(e)}})
logger.exception("AI generation failed")
raise HTTPException(status_code=500, detail=f"AI generation failed: {e}")
# extract chart_data from parsed summary
try:
chart_data = extract_chart_data_json_by_type(summary_obj, df_clean)
except Exception as e:
await snapshots.update_one({"snapshot_id": snapshot_id}, {"$set": {"status": "failed", "error": str(e)}})
logger.exception("Chart extraction failed")
raise HTTPException(status_code=500, detail=f"Chart extraction failed: {e}")
# save preprocessed CSV (async via executor)
try:
preprocessed_path = await save_preprocessed_df(df_clean, snapshot_id)
except Exception as e:
await snapshots.update_one({"snapshot_id": snapshot_id}, {"$set": {"status": "failed", "error": str(e)}})
logger.exception("Saving preprocessed failed")
raise HTTPException(status_code=500, detail=f"Saving preprocessed failed: {e}")
# finalize snapshot
await snapshots.update_one(
{"snapshot_id": snapshot_id},
{"$set": {
"summary": summary_obj,
"chart_data": chart_data,
"preprocessed_path": preprocessed_path,
"status": "done",
"completed_at": datetime.now(timezone.utc).isoformat()
}}
)
return AnalysisResponse(summary=summary_obj, chart_data=chart_data, metadata=meta)
finally:
try:
os.unlink(tmp_path)
except Exception:
pass
@app.get("/snapshots")
async def list_snapshots(limit: int = Query(20, ge=1, le=100), offset: int = Query(0, ge=0)):
cursor = snapshots.find({}, {"preprocessed_path": 0, "summary": 0, "chart_data": 0}).sort("created_at_dt", -1).skip(offset).limit(limit)
items = []
async for doc in cursor:
items.append({
"id": doc.get("snapshot_id") or str(doc.get("_id")),
"filename": doc.get("filename"),
"metadata": doc.get("metadata"),
"status": doc.get("status"),
"created_at": doc.get("created_at"),
})
return {"count": len(items), "items": items}
@app.get("/snapshot/{snapshot_id}")
async def get_snapshot(snapshot_id: str):
doc = await snapshots.find_one({"snapshot_id": snapshot_id})
if not doc:
raise HTTPException(status_code=404, detail="Snapshot not found")
return {
"id": doc["snapshot_id"],
"filename": doc.get("filename"),
"metadata": doc.get("metadata"),
"summary": doc.get("summary"),
"chart_data": doc.get("chart_data"),
"status": doc.get("status"),
"created_at": doc.get("created_at"),
}
@app.get("/preprocessed/{snapshot_id}")
async def get_preprocessed(snapshot_id: str, limit: int = 100, offset: int = 0):
doc = await snapshots.find_one({"snapshot_id": snapshot_id})
if not doc:
raise HTTPException(status_code=404, detail="Snapshot not found")
path = doc.get("preprocessed_path")
if not path or not os.path.exists(path):
raise HTTPException(status_code=404, detail="Preprocessed data not available")
df = pd.read_csv(path)
total = len(df)
rows = df.iloc[offset: offset + limit].to_dict(orient="records")
return {"total": total, "offset": offset, "limit": limit, "rows": rows}
@app.post("/drill")
async def drill(req: DrillRequest):
doc = await snapshots.find_one({"snapshot_id": req.snapshot_id})
if not doc:
raise HTTPException(status_code=404, detail="Snapshot not found")
path = doc.get("preprocessed_path")
if not path or not os.path.exists(path):
raise HTTPException(status_code=404, detail="Preprocessed data not available")
df = pd.read_csv(path)
if req.filter_column not in df.columns:
raise HTTPException(status_code=400, detail=f"Column {req.filter_column} not found in preprocessed data")
try:
filtered = df[df[req.filter_column] == req.filter_value]
if filtered.empty:
filtered = df[df[req.filter_column].astype(str) == str(req.filter_value)]
except Exception:
filtered = df[df[req.filter_column].astype(str) == str(req.filter_value)]
total = len(filtered)
rows = filtered.iloc[req.offset: req.offset + req.limit].to_dict(orient="records")
highlights = req.highlight_columns or [req.filter_column]
highlights = [c for c in highlights if c in df.columns]
return {
"snapshot_id": req.snapshot_id,
"filter_column": req.filter_column,
"filter_value": req.filter_value,
"total_matches": total,
"offset": req.offset,
"limit": req.limit,
"rows": rows,
"highlight_columns": highlights,
}
# Exception handlers
@app.exception_handler(HTTPException)
async def http_exception_handler(request, exc):
return JSONResponse(status_code=exc.status_code, content={"error": exc.detail})
@app.exception_handler(Exception)
async def general_exception_handler(request, exc):
logger.exception("Unhandled exception")
return JSONResponse(status_code=500, content={"error": "Internal server error", "details": str(exc)})
if __name__ == "__main__":
import uvicorn
uvicorn.run(app, host="0.0.0.0", port=int(os.getenv("PORT", "7860")))