dosubot[bot]
commented
1 month ago To resolve the issue where the placeholder [insert output from running the Pandas instruction] is not being replaced with the actual output when querying the Pandas Query Engine via a FastAPI endpoint, you need to ensure that the PandasInstructionParser function is correctly processing the output and that the PandasQueryEngine is being used properly.
Here is a code snippet that demonstrates how to use the PandasQueryEngine to parse and execute Pandas instructions:
from fastapi import FastAPI, HTTPException
from pydantic import BaseModel
import pandas as pd
from llama_index.experimental.query_engine.pandas.output_parser import PandasInstructionParser
from llama_index.experimental.query_engine.pandas.pandas_query_engine import PandasQueryEngine
app = FastAPI()
# Load the CSV file
df = pd.read_csv("sample.csv")
# Initialize the PandasQueryEngine with the DataFrame
query_engine = PandasQueryEngine(df)
instruction_parser = PandasInstructionParser(df)
class Query(BaseModel):
query: str
@app.post("/query")
async def process_query(query: Query):
try:
# Create a QueryBundle with the instruction
query_bundle = QueryBundle(query_str=query.query)
# Execute the query using the PandasQueryEngine
response = query_engine._query(query_bundle)
# Parse the output using PandasInstructionParser
parsed_output = instruction_parser.parse(response.response)
return {"response": parsed_output}
except Exception as e:
raise HTTPException(status_code=500, detail=str(e))
if __name__ == "__main__":
import uvicorn
uvicorn.run(app, host="0.0.0.0", port=7000)In this example, the FastAPI endpoint /query accepts a POST request with a Pandas instruction as input. The PandasQueryEngine is used to parse and execute the instruction, and the result is processed by the PandasInstructionParser to ensure the placeholder is replaced with the actual output. Make sure that the query string contains valid Pandas code that can be executed on the provided DataFrame df [1][2][3].
To continue talking to Dosu, mention @dosu.
Sridevi17j
Question Validation
Question
Below is my code
from fastapi import FastAPI, HTTPException from pydantic import BaseModel import pandas as pd from IPython.display import Markdown, display import time
import pandas as pd from llama_index.llms.azure_openai import AzureOpenAI from llama_index.experimental.query_engine import PandasQueryEngin
Initialize the PandasQueryEngine
query_engine = PandasQueryEngine(df=df, llm=llm, verbose=True, synthesize_response=True)
class Query(BaseModel): query: str
app = FastAPI()
Load the CSV file
df = pd.read_csv("sample.csv")
@app.post("/query") async def process_query(query: Query): try: response = query_engine.query(query.query) return {"response": response.response} except Exception as e: raise HTTPException(status_code=500, detail=str(e))
if name == "main": import uvicorn uvicorn.run(app, host="0.0.0.0", port=7000)
"query": "Give me the number of users in the document" } { "response": "The number of unique users in the document is [insert output from running the Pandas instruction]." }