{"id":28053366,"url":"https://github.com/jg8481/robot-framework-ai-agent-datadriver","last_synced_at":"2026-05-01T04:39:16.476Z","repository":{"id":290696197,"uuid":"971470145","full_name":"jg8481/Robot-Framework-AI-Agent-Datadriver","owner":"jg8481","description":"This is an evolving toolkit focused on AI agent experiments that combine the capabilities of Codename Goose, MCP tools, the Robot Framework ecosystem, decentralized LLMs on Ollama, and Docker.","archived":false,"fork":false,"pushed_at":"2025-05-08T04:35:17.000Z","size":2784,"stargazers_count":0,"open_issues_count":0,"forks_count":0,"subscribers_count":1,"default_branch":"main","last_synced_at":"2025-05-12T03:15:12.546Z","etag":null,"topics":["ai-agents","codename-goose","ollama","robotframework","rpa"],"latest_commit_sha":null,"homepage":"","language":"HTML","has_issues":true,"has_wiki":null,"has_pages":null,"mirror_url":null,"source_name":null,"license":null,"status":null,"scm":"git","pull_requests_enabled":true,"icon_url":"https://github.com/jg8481.png","metadata":{"files":{"readme":"README.md","changelog":null,"contributing":null,"funding":null,"license":null,"code_of_conduct":null,"threat_model":null,"audit":null,"citation":null,"codeowners":null,"security":null,"support":null,"governance":null,"roadmap":null,"authors":null,"dei":null,"publiccode":null,"codemeta":null,"zenodo":null}},"created_at":"2025-04-23T15:11:37.000Z","updated_at":"2025-05-08T04:50:36.000Z","dependencies_parsed_at":null,"dependency_job_id":"47e722e4-31f3-4bc7-bc3c-a7a48cd23cfc","html_url":"https://github.com/jg8481/Robot-Framework-AI-Agent-Datadriver","commit_stats":null,"previous_names":["jg8481/robot-framework-ai-agent-datadriver"],"tags_count":0,"template":false,"template_full_name":null,"repository_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories/jg8481%2FRobot-Framework-AI-Agent-Datadriver","tags_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories/jg8481%2FRobot-Framework-AI-Agent-Datadriver/tags","releases_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories/jg8481%2FRobot-Framework-AI-Agent-Datadriver/releases","manifests_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories/jg8481%2FRobot-Framework-AI-Agent-Datadriver/manifests","owner_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/owners/jg8481","download_url":"https://codeload.github.com/jg8481/Robot-Framework-AI-Agent-Datadriver/tar.gz/refs/heads/main","host":{"name":"GitHub","url":"https://github.com","kind":"github","repositories_count":253666170,"owners_count":21944625,"icon_url":"https://github.com/github.png","version":null,"created_at":"2022-05-30T11:31:42.601Z","updated_at":"2022-07-04T15:15:14.044Z","host_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub","repositories_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories","repository_names_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repository_names","owners_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/owners"}},"keywords":["ai-agents","codename-goose","ollama","robotframework","rpa"],"created_at":"2025-05-12T03:15:16.769Z","updated_at":"2026-05-01T04:39:16.470Z","avatar_url":"https://github.com/jg8481.png","language":"HTML","funding_links":[],"categories":[],"sub_categories":[],"readme":"# Robot-Framework-AI-Agent-Datadriver\n\n![Robot](./notes-images-and-demonstrations/robotframework-ai-agent.png)\n\n## LLMs, AI agents, MCP tools, combined with RPA\n\nThis experimental toolkit is intended to help explore new technologies related to LLMs, AI agents, and [MCP tools](https://www.pulsemcp.com/) combined with battle-tested Robot Framework RPA strategies. My intention is to demonstrate the usefulness of AI agents and LLMs [beyond the \"vibecode\" trend](https://x.com/karpathy/status/1886192184808149383). To get the gist of what I mean, scroll down to the `Toolkit Roadmap` section.\n\nI've taken some inspiration from the Codename Goose approach to using [instruction files](https://block.github.io/goose/docs/guides/running-tasks#using-an-instruction-file) to automate AI agents. Instead of Markdown files, I am using [robotframework-datadriver](https://github.com/Snooz82/robotframework-datadriver) and Robot Framework tags to control different categories of AI agent prompts. [Pabot](https://github.com/mkorpela/pabot) is also being used to control parallel running Codename Goose Docker containers.\n\n![Agent](./notes-images-and-demonstrations/demonstration1.png)\n\nThere are two primary options for using this project. [Both have their strengths and technical challenges.](https://block.github.io/goose/blog/2025/03/31/goose-benchmark/#technical-challenges-with-open-models) I will be writing up more documentation to explain that soon.\n\n- The first option is to use a fully decentralized LLM provider that runs privately and locally on your hardware. The screenshot above shows an example of an automation run using the qwen2.5:14b model on Ollama.\n    - In my case, I only use macOS machines for this work, but it is definitely possible to run all of this on Windows or Linux (scroll down to the `Technical Requirements` section for more information).\n    - This project was designed, created, and tested on an M3 Max MacBook Pro with the following specifications. **You may need to consider your hardware needs before attempting to run local LLMs.**\n        - https://support.apple.com/en-us/117736\n- The second option is available to anyone who does not have access to reliable hardware. This option requires the Codename Goose Docker container and an AI cloud provider's API key.\n    - *Examples:* Google Gemini, OpenAI, Anthropic, etc.\n\n### Toolkit Roadmap\n\nYears ago, I created another AI project called [robotframework-test-assistant](https://github.com/jg8481/leon/tree/develop/packages/robotframework-test-assistant) that focused mostly on software test activities. I consider this to be its successor and plan to expand this toolkit into the following areas that can assist in other types of activities (some may be useful to testers):\n\n- (April 29, 2024 - **Ready To Use**) An AI agent assistant that can research any topic.\n- (April 29, 2024 - **Ready To Use**) An AI agent assistant that can analyze a given text-based dataset using Retrieval Augmented Generation (RAG).\n    - *Examples:* Source code, logs, reports, e-books, PDFs, etc.\n    - For now, only the [Docker AI agent examples](https://github.com/jg8481/Robot-Framework-AI-Agent-Datadriver/blob/main/agent-instructions/docker-agent1-distributed-instructions.csv) in this toolkit can do this.\n- An AI agent assistant that can audit any given source code repo for vulnerabilities.\n    - For now, this partially works only in the [Docker AI agent examples.](https://github.com/jg8481/Robot-Framework-AI-Agent-Datadriver/blob/main/agent-instructions/docker-agent1-distributed-instructions.csv) I am working on a version that will use Ollama.\n- An AI agent assistant that will utilize relevant and secure MCP Servers from the [4,000+ projects in the Model Context Protocol (MCP) ecosystem](https://www.pulsemcp.com/servers).\n- Parallel running Docker AI agents running MCP tools using Anthropic, OpenAI, Groq, etc. at the same time.\n    - Basically multiple different cloud AI APIs running different MCP workflows simultaneously on the same hardware.\n- I have more ideas that I will be working on in the future.\n\n### Technical Requirements\n\nThe following are the basic technical requirements needed to run the AI agent RPA automation. Please note that **this entire toolkit was developed on a macOS machine**, but this should also work for Windows and Linux users.\n\n- **Python 3** -\u003e https://www.python.org/downloads/\n- **Robot Framework** -\u003e https://robotframework.org\n- **Docker** -\u003e https://docs.docker.com/\n- **Homebrew** -\u003e https://brew.sh/\n- **Codename Goose**, a framework for creating AI agents.\n    - macOS, Windows, and Linux users can learn how to install it properly here -\u003e https://block.github.io/goose/docs/getting-started/installation\n- **Ollama**, a framework for building and running language models on local machines.\n    - macOS, Windows, and Linux users can learn how to install it properly here -\u003e https://ollama.com/download\n    - macOS users can follow this guide to install it using Homebrew -\u003e https://formulae.brew.sh/formula/ollama\n    - After finishing the installation, you can run either of the following commands from your terminal. There are hundreds of models and I recommend you research them for your own use cases. Change the provided `config.yaml` file to match the model you installed. I will be experimenting with a lot of models in the future so the provided `config.yaml` file will change over time. \n        - `ollama run mistral`\n        - `ollama run qwen3`\n\n## Quick Start Guide\n\nThis is a fully functional RPA toolkit you can start using and customizing for your needs today. I recommend looking at all of the example report Markdown files stored in the `automation-resources/ai-agent-output-collector/ai-agent-output` [folder](https://github.com/jg8481/Robot-Framework-AI-Agent-Datadriver/tree/main/automation-resources/ai-agent-output-collector/ai-agent-output) to get an idea of what is currently possible with the AI agent RPA automation being maintained in this project. More documentation will be coming soon, along with a series of related Robot Framework [Medium.com](https://medium.com/@joshua.gorospe) articles I plan to write. Recent articles and walkthroughs will be listed below.\n\n- https://medium.com/@joshua.gorospe/robot-framework-rpa-and-ai-agents-a-powerful-combination-14be64d2fe79\n\n**Before running any of the automation, first make sure that all of the basic technical requirements are installed and working**, then you can run `bash ./start-ai-agent-workflow-experiments.sh Run-Datadriven-Sequential-Local-AI-Tasks ResearchORQuestions` to experiment with a basic workflow. To see more workflow options run `bash ./start-ai-agent-workflow-experiments.sh -h` in your terminal.\n\n***\n\n## Thank You!\n\nBig thank you to the following people and groups. Without their open source contributions, this RPA project would not exist.\n\n- Pekka Klarck\n- Mikko Korpela\n- René Rohner\n- Docker developers\n- Ollama developers\n- Codename Goose developers","project_url":"https://awesome.ecosyste.ms/api/v1/projects/github.com%2Fjg8481%2Frobot-framework-ai-agent-datadriver","html_url":"https://awesome.ecosyste.ms/projects/github.com%2Fjg8481%2Frobot-framework-ai-agent-datadriver","lists_url":"https://awesome.ecosyste.ms/api/v1/projects/github.com%2Fjg8481%2Frobot-framework-ai-agent-datadriver/lists"}