sirji-agents 0.0.26

Orchestrator, Generic Agent, and Research Agent components of the Sirji AI agentic framework.

Sirji is an agentic AI framework for software development.

Built with ❤️ by True Sparrow

Sirji Agents

sirji-agents is a PyPI package that implements the following components of the Sirji AI agentic framework:

• Orchestrator: The Orchestrator is the central component in the Sirji framework, responsible for managing the flow and execution of tasks across different agents.
• Generic Agent: Run time composable class providing the agent functionality as per the pseudo code provided in the agent.yml file.
• Research Agent: Utilizes RAG (Retrieval-Augmented Generation) and gets trained on URLs and search terms.

By default, it utilizes:

• OpenAI Chat Completions API
• OpenAI Assistants API

Installation

Setup Virtual Environment

We recommend setting up a virtual environment to isolate Python dependencies, ensuring project-specific packages without conflicting with system-wide installations.

python3 -m venv venv
source venv/bin/activate

Install Package

Install the package from PyPi:

pip install sirji-agents

Run the following command to install playwright:

playwright install

Usage

Environment Variables

Ensure that the following environment variables are set:

export SIRJI_PROJECT="Absolute folder path for Sirji to use as its project folder."
export SIRJI_INSTALLATION_DIR='Absolute path of the Sirji installation directory.'
export SIRJI_RUN_PATH='Folder path containing run related logs, etc.'
export SIRJI_OPENAI_API_KEY='OpenAI API key for Chat Completions API and Assistants API'
export SIRJI_MODEL_PROVIDER='Model Provider to be used for LLM inference. Defaults to "openai".'
export SIRJI_MODEL='Model to be used for LLM inference. Defaults to "gpt-4o".'
export SIRJI_MODEL_PROVIDER_API_KEY='API key to be used for LLM inference.'

Orchestrator

The Orchestrator is the central component in the Sirji framework, responsible for managing the flow and execution of tasks across different agents.

# Following is a sample recipe
recipe = {
  "prescribed_tasks": [
    {
      "task": "Write epics and user stories.",
      "agent": "PRODUCT_MANAGER"
    },
    {
      "task": "Write architecture components.",
      "agent": "ARCHITECT"
    },
    {
      "task": "Implement the epic & user stories using the architecture components.",
      "agent": "CODER"
    }
  ],
  "tips": [
    {
      "deviation": "Architecture components are not in sync with the epics and user stories.",
      "agent": "PRODUCT_MANAGER",
      "task": "Make the epics and user stories consistent with the architecture components."
    }
  ]
}

from sirji_agents import Orchestrator

agent = Orchestrator(recipe)

# History is the array of LLM conversations till now
history = []

# Input message string
message_str = ""
response_message, history, prompt_tokens, completion_tokens = agent.message(message_str, history)

Generic Agent

Run time composable class providing the agent functionality as per the pseudo code provided in the agent.yml file.

# Convert the agent.yml file to a Python dictionary
config = {
  "id": "CODER",
  "name": "Coding Agent",
  "llm": {
    "provider": "openai",
    "model": "gpt-4o"
  },
  "skills": [
    {
      "skill": "Developing end-to-end working code for the epic & user stories, making use of the finalized architecture components.",
      "sub_tasks": [
        "Read problem statement, epics & user stories and architecture components from agent output folder.",
        "Write concrete code and not just conceptualize or outline or simulate it.",
        "Follow secure software development practices while generating code.",
        "Ensure that you don't create any file/folder outside of project folder, i.e. './'",
        "Install programming language-specific packages or libraries in local folders, utilizing tools such as venv for installing Python dependencies and package.json for managing Node.js dependencies.",
        "Verify whether a system-level command is already installed to avoid triggering the installation of packages that are already in place.",
        "Always execute the code and evaluate the response output. If the response has errors, solve them before moving ahead."
      ]
    }
  ]
}

agent_output_folder_index = {
  "SIRJI/problem.txt": {
    "description": "Problem statement from the user.",
    "created_by": "SIRJI"
  },
  "PRODUCT_MANAGER/finalized_epics_user_stories.txt": {
    "description": "Finalized Epics and User Stories for the Tic-Tac-Toe game with AI opponent.",
    "created_by": "PRODUCT_MANAGER"
  },
  "ARCHITECT/finalized_architecture_components.txt": {
    "description": "Finalized architecture components for the Tic-Tac-Toe game with AI opponent.",
    "created_by": "ARCHITECT"
  }
}

from sirji_agents import GenericAgent

agent = GenericAgent(config, agent_output_folder_index)

history = []
message_str = "***\nFROM: ORCHESTRATOR\nTO: CODER\nACTION: INVOKE_AGENT\nSUMMARY: Implement the epic & user stories using the architecture components.\nBODY:\nPImplement the epic & user stories using the architecture components.\n***"

response_message, history, prompt_tokens, completion_tokens = agent.message(message_str, history)

Research Agent

The Research Agent utilizes RAG (Retrieval-Augmented Generation) and gets trained on URLs and search terms.

Initialization

from sirji_agents import ResearchAgent

# Initialize Researcher without assistant ID
researcher = ResearchAgent('openai_assistant', 'openai_assistant')

# init_payload fetched from researcher object should be persisted
init_payload = researcher.init_payload

# Initialize Researcher with assistant ID
researcher = ResearchAgent('openai_assistant', 'openai_assistant', init_payload)

Some example message handling usages are given below.

Train using URL

from sirji_messages import MessageFactory, ActionEnum

message_class = MessageFactory[ActionEnum.TRAIN_USING_URL.name]

body = {
    "URL": "https://www.w3schools.com/python/python_json.asp"
}

message_str = message_class().generate({
            "from_agent_id": "Id of the agent, who is invoking the action",
            "summary": "{{Display a concise summary to the user, describing the action using the present continuous tense.}}",
            "body": body
        })

researcher.message(message_str)

Infer

from sirji_messages import MessageFactory, ActionEnum

message_class = MessageFactory[ActionEnum.INFER.name]
infer_query = "What is the capital of India?"
message_str = message_class().generate({
            "from_agent_id": "Id of the agent, who is invoking the action",
            "summary": "{{Display a concise summary to the user, describing the action using the present continuous tense.}}",
            "body": infer_query
        })  
        
response, total_tokens = researcher.message(message_str)

For Contributors

1. Fork and clone the repository.
2. Create and activate the virtual environment as described above.
3. Set the environment variables as described above.
4. Install the package in editable mode by running the following command from repository root:

pip install -e .

5. Run the following command to install playwright:

playwright install

Running Tests and Coverage Analysis

TODO - Introduce test cases.

License

Distributed under the MIT License. See LICENSE for more information.