import openai, json
from typing import List
from pydantic import BaseModel
client = openai.AsyncOpenAI(
# defaults to os.environ.get("OPENAI_API_KEY")
# or set/uncomment below
#api_key="<api key>"
)
[docs]class SuggestedQueries(BaseModel):
"""Suggested Queries model"""
suggestions: List[str]
[docs]async def get_genai_suggestions(partial_query: str, context: dict, fields_info: dict, n_queries: int = 10) -> SuggestedQueries:
"""Get genai suggestions for the given partial query, context and fields_info, Uses latest function call architecture.
:param partial_query: The partial query for which suggestions are needed
:param context: The context for the query
:param fields_info: The fields information
:param n_queries: Number of queries to suggest, defaults to 10
:type partial_query: str\
:type context: dict
:type fields_info: dict
:type n_queries: int, optional
:return: Suggested queries
:rtype: :class:`SuggestedQueries`
"""
response = await client.chat.completions.create(
model="gpt-3.5-turbo",
messages=[
{"role": "system", "content": "You are an AI assistant guiding user towards better ecommerce search. You are given a partial query and some context (might not be avaliable), Use it to generate query recommendations for the user, Optionally if exists then you can also use field info, that is mapping between a feature name to its possible values."},
{"role": "user", "content": f"Call function with {n_queries} recommendation, Query: {partial_query}, Supporting Data: {json.dumps(context)} and avaliable fields: {json.dumps(fields_info)}"}
],
functions=[
{
"name": "get_suggested_queries",
"description": "Get suggested queries for the user based on the partial query and context",
"parameters": SuggestedQueries.model_json_schema()
}
],
function_call={"name": "get_suggested_queries"},
temperature=0.5,
)
output = json.loads(response.choices[0].message.function_call.arguments)
return SuggestedQueries(**output)