API Reference

_load_config(config_path: str) -> Dict[str, Any]
Reads and parses the configuration JSON file.

Usage Example:

config = agent._load_config("agent_config.json")

_load_tools(tool_names: List[str]) -> Dict[str, Dict[str, Any]]
Loads tool configuration files from the tools/ directory based on the list of tool names.

_execute_tool(tool_name: str, parameters: Dict[str, Any]) -> Any
Dynamically imports and executes a tool function with the provided parameters.

_create_system_prompt() -> str
Generates a system prompt using configuration data and tool information.

_format_tools_for_llm() -> List[Dict[str, Any]]
Formats tool configurations into the structure required by the LLM provider.

process_query(query: str) -> str
Main method to process a user query:

• Builds message context (system prompt, conversation memory, and query).
• Sends messages to the LLM.
• Detects and executes tool calls.
• Returns the final response.

Usage Example:

from agent import create_agent, run_agent
agent = create_agent("agent_config.json", llm_provider="openai")
response = run_agent(agent, "Calculate 3 * 5")
print("Agent Response:", response)

create_agent(config_path: str, llm_provider: str = "openai") -> Agent
Factory function to instantiate an Agent.

run_agent(agent: Agent, query: str) -> str
Executes the agent's query processing and returns the response.

add(role: str, content: str) -> None
Adds a message with the given role and content to the memory.

Usage Example:

mem = Memory(max_items=10, enabled=True)
mem.add("user", "Hello!")

get_messages() -> List[Dict[str, str]]
Returns the list of stored messages.

clear() -> None
Clears all stored messages.

is_enabled() -> bool
Checks if memory is enabled.

set_enabled(enabled: bool) -> None
Enables or disables memory.

add(role: str, content: str) -> None
Overrides Memory.add to compute and index the embedding of the content.

retrieve_similar(query: str, top_k: int = 3) -> List[Dict[str, Any]]
Retrieves messages semantically similar to the query using FAISS.

Usage Example:

from semantic_memory import SemanticMemory
sem_mem = SemanticMemory(max_items=10, enabled=True)
sem_mem.add("user", "Tell me a joke")
similar = sem_mem.retrieve_similar("I need something funny", top_k=1)
print(similar)

format_tools(tools: Dict[str, Dict[str, Any]]) -> List[Dict[str, Any]]
Must be implemented by subclasses to format tools for the LLM.

get_response(messages: List[Dict[str, Any]], tools: Optional[List[Dict[str, Any]]]) -> str
Must be implemented to get a response from the LLM.

extract_tool_call(response_text: str) -> Optional[Dict[str, Any]]
Parses the LLM response for JSON formatted tool calls.

create_tool_config(tool_name: str, description: str, function_path: str, parameters: Dict[str, Any], required_params: List[str] = None, keywords: List[str] = None) -> None
Creates a tool configuration file (JSON) in the tools/ directory.

Usage Example:

from tool import ToolManager
ToolManager.create_tool_config(
    tool_name="sample_tool",
    description="A sample tool.",
    function_path="tools.sample_tool.run",
    parameters={"param": {"type": "string", "description": "A sample parameter"}},
    required_params=["param"]
)

register_basic_tools() -> None
Registers basic tools (e.g., a calculator) by creating configurations and setting up directories.

setup_basic_config() -> None
Registers basic tools and creates an initial agent configuration file (agent_config.json).

index_documents(documents: List[str]) -> None
Processes and indexes a list of documents.

retrieve(query: str, top_k: int = 3) -> List[str]
Retrieves the top-k documents relevant to the query.

send_message(agent_id: str, message: Dict[str, Any]) -> None
Sends a message to another agent.

receive_messages() -> List[Any]
Retrieves all messages from the communication queue.

log_metric(name: str, value: Any) -> None
Records a performance metric.

evaluate_response_time(start_time: float) -> float
Computes and logs the time taken for a process.

report() -> Dict[str, Any]
Returns a report of all logged metrics.

process_image(image_path: str) -> str
Processes an image and returns a description (e.g., dimensions).

process_audio(audio_path: str) -> str
Processes an audio file and returns a placeholder transcription or analysis.

validate_input(value: Any, expected_type: Type, param_name: str = "parameter") -> None
Validates that an input value is of the expected type; raises a TypeError if not.

validate_output(value: Any, expected_type: Type, param_name: str = "output") -> None
Validates that an output value is of the expected type; raises a TypeError if not.