Setup | DATASPHERES AI - Dataspheres AI

Overview You will install three things and connect them: your AI coding tool, the ari-dai-skills kit, and your Dataspheres API key. This is the only techni...

Overview You will install three things and connect them: your AI coding tool, the ari-dai-skills kit, and your Dataspheres API key. This is the only technical module. Take it slowly. Once it works, you rarely touch it again. Lesson 1.1: Make your accounts Create a free GitHub account at https://github.com if you do not have one. Sign up for a Dataspheres starter plan at https://dataspheres.ai. Create a datasphere for your show. Give it a name, a logo and a banner. Your logo appears on the cover, the sign-off card and every slide that names your show, so set it now. Lesson 1.2: Get your API key Go to https://dataspheres.ai/app/developers?tab=keys. Create a key. It starts with dsk_ . Treat it like a password. Never paste it into a post, a screenshot or a shared document. If you ever see "401 Invalid API key", stop. Go back to the same page, make a new key and replace the old one. Lesson 1.3: Connect your editor and install the skills Open the Dataspheres developer portal and follow its current connection instructions for your editor. Choose production when you intend to publish on dataspheres.ai. Install the workflow skills from ari-dai-skills . Its README and installer list supported clients and their options. MCP supplies tools; skills supply the workflow instructions. For the repository installer, clone the repository and run the supported installer in your project. On Windows, use Git Bash. Replace the project path with your own. git clone https://github.com/geekdreamzz/ari-dai-skills cd ari-dai-skills bash install.sh --all --project /path/to/your/project The installer supports Claude Code, Cursor and GitHub Copilot. Consult the README for your client’s current setup and the Windows copy option. For other clients, use the developer portal’s supported MCP connection instructions. Store an API key only in your private local configuration, outside version control. For the REST-based skills, the documented settings are DATASPHERES_API_KEY , DATASPHERES_BASE_URL=https://dataspheres.ai and DATASPHERES_PUBLIC_URL=https://dataspheres.ai . Do not paste your real key into course exercises. Ask your tool to list your dataspheres, then read the intended datasphere. Check the account, environment and permissions before the first draft write. Do this now Finish setup and ask "List my dataspheres." Then ask your tool to check your datasphere's brand: "Does my datasphere have a logo, a banner and a show name?" Fix anything missing. Quiz Where do you create your API key? a) In GitHub settings b) https://dataspheres.ai/app/developers?tab=keys c) Inside your AI tool Answer: b What do you do if you see "401 Invalid API key"? a) Try other keys until one works b) Make a new key on the developers page and replace the old one c) Ignore it Answer: b Why set your logo before your first show? a) It appears on the cover, sign-off card and slides that name your show b) It is required to log in c) It changes your API key Answer: a Lesson 1.4: Verify the MCP connection before generating The skills supply instructions; Dataspheres AI MCP exposes the platform tools. Connect to the intended environment, sign in, and verify your account and datasphere before starting work. Use get_content_governance , list_content_templates and get_content_template to read the current rules and required variables. If a tool is missing, check your connection and available tools before starting a paid run. Start one job with start_content_pipeline , save its runId , and inspect it with get_content_pipeline_run . Keep each returned researchId distinct from its conversation ID. Search completion can precede the written report; wait for the report instead of creating another paid run. The owner reviews the deck, script, rendered video and kit in that order. An agent may record a decision only after the person gives it. Practice: identify the connected environment, retrieve the rulebook and template without generating, then write the audience, topic, deadline and spend ceiling for a single pilot. Save the run ID and evidence for each review. Record the returned run ID, the artifact links and the actual result of each review.