ContextOracleOpenAIToolkitBuilder

ContextOracleOpenAIToolkitBuilder is a class that’s specifically used to build tools for the OpenAI agent. The class provides the necessary infrastructure to create the tools that can be utilized by the OpenAI agent in the Automata framework.

Import Statements

import logging
import textwrap
from typing import List
from automata.config.base import LLMProvider
from automata.agent.agent import AgentToolkitBuilder, AgentToolkitNames
from automata.agent.providers import OpenAIAgentToolkitBuilder
from automata.embedding.base import EmbeddingSimilarityCalculator
from automata.llm.providers.openai import OpenAITool
from automata.memory_store.symbol_code_embedding import SymbolCodeEmbeddingHandler
from automata.memory_store.symbol_doc_embedding import SymbolDocEmbeddingHandler
from automata.singletons.toolkit_registries import OpenAIAutomataAgentToolkitRegistry
from automata.tools.base import Tool

Methods

Here’s the unique method build_for_open_ai() of ContextOracleOpenAIToolkitBuilder:

def build_for_open_ai(self) -> List[OpenAITool]:
    tools = super().build()

    # Predefined properties and required parameters
    properties = {
        "query": {"type": "string", "description": "The query string to search for."},
        "max_additional_related_symbols": {
            "type": "integer",
            "description": "The maximum number of additional related symbols to return documentation for.",
        },
    }
    required = ["query"]

    openai_tools = []
    for tool in tools:
        openai_tool = OpenAITool(
            function=tool.function,
            name=tool.name,
            description=tool.description,
            properties=properties,
            required=required,
        )
        openai_tools.append(openai_tool)

    return openai_tools

Usage Example

from automata.memory_store.symbol_code_embedding import SymbolCodeEmbeddingHandler
from automata.memory_store.symbol_doc_embedding import SymbolDocEmbeddingHandler
from automata.embedding.base import EmbeddingSimilarityCalculator
from automata.tools.builders.context_oracle import ContextOracleOpenAIToolkitBuilder

symbol_doc_embedding_handler = SymbolDocEmbeddingHandler()
symbol_code_embedding_handler = SymbolCodeEmbeddingHandler()
embedding_similarity_calculator = EmbeddingSimilarityCalculator()

context_oracle_open_ai_toolkit_builder = ContextOracleOpenAIToolkitBuilder(
    symbol_doc_embedding_handler=symbol_doc_embedding_handler,
    symbol_code_embedding_handler=symbol_code_embedding_handler,
    embedding_similarity_calculator=embedding_similarity_calculator,
)

# Build the OpenAI tools
openai_tools = context_oracle_open_ai_toolkit_builder.build_for_open_ai()

Limitations

The main limitation is that ContextOracleOpenAIToolkitBuilder is tightly coupled to the structure of OpenAITool. Any changes in OpenAITool might require corresponding changes in ContextOracleOpenAIToolkitBuilder as well.

Follow-up Questions:

  • How flexible is the ContextOracleOpenAIToolkitBuilder in terms of supporting different tools aside from the predefined ones?

  • How do we plan to handle the changes if OpenAITool diverges significantly in structure or properties in future updates?