Use AutoGen Studio and add a new skill to store notes in Obsidian

preview_player
Показать описание
AutoGen Studio is a UI for AutoGen maintained by the AutoGen Team. We can use AutoGen Studio to add new skills. In this Tutorial we add a new Skill to write notes to the Note-Taking App Obsidian.

02:31 Install AutoGen Studio
03:02 Introducing AutoGen Studio
03:48 Use the sample Markdown Prompt in AutoGen Studio
04:39 Use Obsidian help and interactive Graph
05:51 Write the function to be used in a new skill in AutoGen Studio
07:05 Add a new Skill to AutoGen Studio
07:20 Add a new Agent to AutoGen Studio
07:40 Add a new Workflow to AutoGen Studio
08:31 Test the Workflow in AutoGen Studio Playground

Gist: An AutoGen Studio skill to store markdown content to our second brain obsidian
Рекомендации по теме
Комментарии
Автор

Brilliant, thanks! Opens up possibilities to creating skills to other 3rd party apps

basilbrush
Автор

Great video...but those OAI fees are adding up, especially with these chatty agents! How about a followup to link this to an open source LLM with a local endpoint, ?: multi-model wrapper like text-generation-webui; ollama; liteLLM hugging face hosted model client interface.

Looking forward to the "memory autogen" features they just announced as coming :)

IdPreferNot
Автор

Great video! 👍 Do you happen to have the Obsidian skill code to share?

mikew
Автор

Great video. When you close the autogen studio, what happens to the new skills you loaded? Where are they stored? Are they available in another session?

paullopez_ai
Автор

Do you know if there is a way to link the obsidian notes to an agents "knowledge". Similar to how you can add pdf files or txt files to gpts in chatgpt? It would be interesting since you would be able to build a knowledge base for each agent giving each one its own specific domain of expertise.

jamescapernick
Автор

I followed the video and had everything working well for around 2 hours, the workflow, agent and skill ran and placed the test material in the obsidian vault as expected. Used gpt4 model ok

Then I started searching the web for various common info (states, population, etc) and those worked well for a while. Several documents with located material were placed in the vault.

Then suddenly the workflow would not complete, it would terminate and tell me to execute the code it made for each use case instead of running it itself. Would not interate and make and run the changes needed for each pass.

Anybody ever see this? any clues?

fkxfkx