conva-ai 0.1.3

Python SDK for using Conva AI co-pilots

Python Library for Conva AI

This is the python library for using Conva AI Co-pilots

Examples

1. A simple example for generating response using Conva Co-pilot

import asyncio
from conva_ai import AsyncConvaAI
client = AsyncConvaAI(
    assistant_id="<YOUR_ASSISTANT_ID>", 
    assistant_version="<YOUR_ASSISTANT_VERSION>", 
    api_key="<YOUR_API_KEY>"
)
async def generate_with_capability_group(client: AsyncConvaAI, query: str, capability_group: str = "default", stream: bool = "True"):
  if stream:
    response = await client.invoke_capability_group_stream(query, capability_group=capability_group)
    out = ""
    async for res in response:
        out = res.model_dump_json(indent=4)
    return out
  else:
    response = await client.invoke_capability_group(query, capability_group=capability_group)
    return response.model_dump_json(indent=4)

final_response = asyncio.run(generate_with_capability_group(client, "how are you", stream=True))
print(final_response)

The above snippet of code is used for invoking a capability group.

Similarly, a particular capability can be invoked by

import asyncio
from conva_ai import AsyncConvaAI
client = AsyncConvaAI(
    assistant_id="<YOUR_ASSISTANT_ID>", 
    assistant_version="<YOUR_ASSISTANT_VERSION>", 
    api_key="<YOUR_API_KEY>"
)
async def generate_with_capability_name(client: AsyncConvaAI, query: str, capability_name: str, stream: bool):
  if stream:
    response = await client.invoke_capability_stream(query, capability_name=capability_name)
    out = ""
    async for res in response:
        out = res.model_dump_json(indent=4)
    return out
  else:
    response = await client.invoke_capability(query, capability_name=capability_name)
    return response.model_dump_json(indent=4)

final_response = asyncio.run(generate_with_capability_name(client, "buy 10 shares", "order_management", True))
print(final_response)

You can try out the co-pilot on Google Colab

If you want to get the response as dictionary, then replace

out = res.model_dump_json(indent=4)

with

out = res.model_dump()

2. How to clear history

Conva AI client, by default keeps track of your conversation history and uses it as the context for responding intelligently

You can clear conversation history by executing the below code:

from conva_ai.client import AsyncConvaAI
client = AsyncConvaAI(
    assistant_id="<YOUR_ASSISTANT_ID>", 
    assistant_version="<YOUR_ASSISTANT_VERSION>", 
    api_key="<YOUR_API_KEY>"
)
client.clear_history()

In case you are buliding an application where you don't want to track conversation history, you can disable history tracking

client.use_history(False)

You can enable history by

client.use_history(True)

3. Debugging responses

Conva AI uses generative AI to give you the response to your query. In order for you to understand the reasoning behind the response. We also provide you with AI's reasoning

final_response_dict = json.loads(final_response)
print(final_response_dict["reason"])