By storing a final reflection at the end, and pulling the most relevant reflection to guide future runs, BabyAGI slowly generates better and better tasks lists.
You can chat w BabyAGI! We have an experimental UI where the chat is separate from the tasks/output panel, allowing you to request multiple tasks in parallel.
The Chat UI can use a single skill quickly, or chain multiple skills together using a tasklist.
The latter is what you formally know as BabyAGI.
I have not yet added streaming of BabyAGI progress to the UI (much needed).
Dev note:
What was previously in has moved to , which has an additional function to execute just one skill.
The new houses the back-end functions for the chat UI, and connections with BabyAGI.main.py babyagi.py main.py
🛠️ New skills (DALLE, Airtable, etc)
Some fun new skills:
🎨 DALLE skill with prompt assist
🎶 Music player w Deezer
📊 Airtable search (add your own table/base ID)
🔍 Startup Analyst (example of beefy function call as a skill)
📄 It’s own README
I know BabyAGI hasn’t been documented well, and the GitHub is not organized well. This is my attempt at helping a bit more than usual…
🐜 Bugs!
I fell behind on work, so I’m pushing this out a bit before I’m really proud of it. Just think of it like fun puzzle to solve if it breaks 🧩
⚙️ Replit
On @Replit, I push each mod as it’s own Repl, so just fork it and add your openai key as a secret called OPENAI_API_KEY, etc.
Quick Backstory: I’ve been modding the original BabyAGI instead of working off of the one people contributed to, so my mods are in the “classic” folder.
Our new approach to self improving auto agents, introduced in the next mod of @babyagi_.
Dig in shall we? 👇
How it works:
With FOXY (Final Output eXamination from "Yesterday"), we do a final reflection on the output of each run, and use this to guide future runs, pulling most relevant reflection using a similarity search, paired with a decay mechanism to prioritize recent reflections.
Here's a simplified version of what this looks like.
In this example, we run the same objective twice. The second task list is improved based on notes from the final reflection of the first run.
Biggest update to the OG @babyAGI_, finally broke out into multiple files:
🛠️ Skills Class allows for easy skill creation
📝 Dynamic Task List example w vector search
🔍 Beta Reflection Agent
🖇️ Can read and write own code (yes it can write new skills and task… https://t.co/mN6jHZydwqtwitter.com/i/web/status/1…
Skills Class
New “Skills” (prev called Mini Agents) are what the task execution agent uses to complete tasks. Enabled in main files handles APIs centrally.
Each skill can be as simple as a simple API call, or a long script w many functions.
New Skills Template
Create easily by simply adding a file in the “skills” folder following our template.
In the template, define the name, description, and required API keys. Description is provided to task creation agent to decide when to use the skill.
- Higher completion of objectives
- Lower errors
- Faster (GPT4 task creation runs once)
- Hand-crafted Mini agent as tool
✨~300 lines of code
Let’s dig in 👇
As a quick recap, OG @babyAGI_ is the original commit, combining three agents in a loop (task execution/creation/prioritization), only using GPT-3, and designed to never stop.