Autonomous AI: AI-Powered Workflow Automation Teams
A modular, Python-based workflow automation concept presented on March 21, 2025. Powered by local ollama AI models, it automates tasks like code generation & document writing.
Last updated
A modular, Python-based workflow automation concept presented on March 21, 2025. Powered by local ollama AI models, it automates tasks like code generation & document writing.
Last updated
Launch Date: March 21, 2025
Purpose: The Autonomous AI Teams Project is a modular, Python-based system designed to automate tasks such as code generation, content creation, and social media posting. Introduced via a video released on March 21, 2025, it leverages AI agents powered by the local ollama library to streamline workflows, offering a solution for enhancing productivity without external service dependencies.
Autonomous AI Teams operate as a collaborative network of AI agents, each assigned a specific role to transform prompts into polished outputs. The system processes requests—like generating a script or drafting a post—through a team of digital workers that refine results iteratively, ensuring organized file storage and detailed progress tracking.
What the Code Does: The code serves as the system’s foundation, orchestrating agent interactions via a central script. It establishes directories for outputs and logs, reads configuration settings, and executes a loop that handles prompts, produces results, and supports refinement based on additional input. It functions as the coordinator, keeping all components aligned.
Modular Agents: The system includes specialized agents for planning, coding, writing, auditing, and sharing, working together as a cohesive unit.
File Management: Outputs are saved with unique filenames, preventing overwrites by appending numerical suffixes (e.g., script_v1, script_v2).
Refinement Loop: Results can be iteratively improved with new inputs, enhancing flexibility.
Local AI: Powered by ollama, it operates entirely offline, ensuring privacy and speed.
Logging: Agents maintain individual diaries, while a system log provides an overview of operations.
What the Code Does:
The code defines each agent’s role in separate modules, implements file naming logic to avoid conflicts, and manages the refinement loop. It integrates ollama for AI processing and logs activities across the system for transparency.
The system requires:
Python 3.8 or higher.
The ollama library for local AI functionality.
The black library for code formatting.
Standard Python libraries: json, os, and datetime.
What the Code Does:
The code leverages json to parse settings, os to handle file and directory operations, and datetime to timestamp logs. It uses ollama to enable agent intelligence and black to standardize code appearance.
AutonomousAI is structured as follows:
A main script to drive the system.
Individual files for each agent (e.g., planner, coder, writer, auditor).
A settings.json file for configuration.
A “Work” directory for outputs and a “logs” directory for records.
What the Code Does:
The code initializes this structure, creating necessary folders, defining agent responsibilities in separate files, and linking them through the main script. It reads settings.json to apply custom configurations.
A settings.json file allows customization, specifying ollama models for each agent, loop iteration counts, and directories for files and logs.
What the Code Does: The code loads this file at startup, applying the settings to tailor agent behavior, control workflow iterations, and direct output storage.
The system comprises several key components:
Main Workflow: This core process accepts prompts, assigns tasks to appropriate agents, saves outputs, and offers refinement options. Code Explanation: The main script executes a loop, sequencing agent actions, saving results, and logging steps. It analyzes prompts to select coding or writing tasks and maintains workflow continuity.
File Manager: Ensures outputs are stored with unique names, while a “Code Manager” provides feedback on plans or code. Code Explanation: This component checks for existing files, appends counters as needed, and saves content. The Code Manager reviews work, offering concise suggestions without altering it.
Planner: Generates ideas or refines content based on prompts.
Code Explanation: It uses ollama to create plans or adjust previous outputs according to input.
Finances: Assesses plan feasibility using a “capabilities.txt” file listing available resources. Code Explanation: It evaluates plans against resource constraints, approving or rejecting them.
Code Engineer: Produces fully functional Python scripts. Code Explanation: It generates complete, executable code—including imports and functions—based on provided plans.
Author: Creates creative content, such as social media posts.
Code Explanation: It generates text, adhering to length limits (e.g., 260 characters for X) when specified, using ollama for creativity.
Auditor: Reviews code or content for errors or excessive length. Code Explanation: It checks for issues, corrects minor problems, and flags significant ones to ensure quality.
Code Formatter: Standardizes Python code appearance.
Code Explanation: It applies black to reformat code for consistency and readability.
Communications: Prepares final outputs for sharing, simulating a broadcast. Code Explanation: It formats or summarizes content for distribution, mimicking a post or announcement.
To operate AutonomousAI:
Configure settings.json and a capabilities.txt file.
Install required dependencies.
Run the main script, input a prompt, and observe the process.
Refine outputs as desired—results are stored in “Work,” logs in “logs.”
What the Code Does: The code initializes the system, processes prompts through the agent network, saves outputs, and logs activities. It supports refinement by looping again with new inputs.
Example: A prompt like “Create a social media poster for X” triggers the Planner to outline it, Finances to verify feasibility, Code Engineer to build it, Auditor to check it, Formatter to polish it, and saves the result for broadcast.
File Issues: Logs reveal if overwrites occur, typically prevented by the system’s naming logic, though permissions may interfere.
Agent Errors: Logs indicate if agents exceed their roles; strict rules maintain their boundaries.
Refinement: Logs confirm the last file is tracked correctly for refinement.
What the Code Does: The code records errors and progress in logs, aiding diagnosis by detailing each agent’s actions.
Potential upgrades include a graphical interface, real social media integration, parallel task processing, and improved error handling.
What the Code Could Do: Future iterations might implement a user-friendly window, connect to APIs, execute tasks concurrently, and enhance robustness with error-catching mechanisms.
AutonomousAI was introduced to the world through a video released on March 21, 2025, showcasing its capabilities as a versatile tool for coders, writers, and social media managers.
is the project site, it includes the Blueprint for Autonomous AI Expansion, and will soon include all of the documentation for the AI Console, AI Chat, AI Loops & Teams, and other open source AI projects.