feat: Add OpenGradient Action Provider #475
Open
Conversation
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Overwriting all commits in order to make them verified. Here's a sum of the list of changes that lead up to this commit: - Add OpenGradient action provider - Add hard-coded OpenGradient tool functions for ML workflow - Add prompt OpenGradient tools for Dobby and Qwen - Add unit tests for OpenGradient tools - Add block explorer links to each tool output - Update CHANGELOG - Add OpenGradient logo asset (might need to fix) - Add OpenGradient to poetry.lock
Still need to update their function descriptions, but the unit tests are added and passing.
🟡 Heimdall Review Status
|
github-actions
bot
added
documentation
|
Thanks for contributing! Taking a look through this over the coming days and will provide any feedback |
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.
This suggestion is invalid because no changes were made to the code.
Suggestions cannot be applied while the pull request is closed.
Suggestions cannot be applied while viewing a subset of changes.
Only one suggestion per line can be applied in a batch.
Add this suggestion to a batch that can be applied as a single commit.
Applying suggestions on deleted lines is not supported.
You must change the existing code in this line in order to create a valid suggestion.
Outdated suggestions cannot be applied.
This suggestion has been applied or marked resolved.
Suggestions cannot be applied from pending reviews.
Suggestions cannot be applied on multi-line comments.
Suggestions cannot be applied while the pull request is queued to merge.
Suggestion cannot be applied right now. Please check back later.
What changed?
Why was this change implemented?
OpenGradient is a Decentralized AI platform that enables end-to-end model storage and secure inference on the OpenGradient blockchain (currently in testnet).
We hope to add a few tools to the coinbase action provider repository:
BTC,ETH,SOL, etc.)These can be used in financial based AI agents with goals like optimizing a defi portfolio, using volatility to automatically adjust buy / sell stops, lending USDC, yield farming, etc.
SentientAGI/Dobby-Mini-Unhinged-Llama-3.1-8BandQwen/Qwen2.5-72B-Instruct)Allow users to access different on-chain LLMs from the current LLM used by the agent itself. This can eventually include custom system prompts, but currently set as Dobby (who is great for creative-writing and general humor) and Qwen. These independent LLMs can even interact with one another!
All of these tools will be run through the OpenGradient blockchain, allowing for transaction tracking through our block explorer.
We are hoping this PR can act as a stepping stone to add OpenGradient agent toolkit functions (see
Notesat the end for more details).Network support
Wallet support
Checklist
OpenGradienttopoetry.lockas it's necessary to use the OpenGradient SDK. Our SDK is open-sourced here, PyPI link here).How has it been tested?
Unit tests added under
tests/action_providers/opengradientExample usecase of quantitative model:
Example usecase of LLM prompting (From Agent test):
Developer note: The Dobby model is pretty funny
Notes to reviewers
Using any of these OpenGradient tools will require an OpenGradient private key (this can be generated through our SDK, generally users should have their own).
Question
For the
README Updateunder the checklist, we may need some guidance on how you want our logos formatted under theproviderslist? Is there other updates needed for this file?Future features
We're hoping to eventually include AlphaSense into the coinbase agent kit.
Alphasense is OpenGradient's langchain compatible toolkit for developing OG based tools. The reasoning behind providing this developer tool is that model inferences can end up being huge. For example, if your input is a 1000 int array that needs to be copied into the AI agent, it can easily pollute agent context window. By adding these tools in the coinbase agent kit we hope to allow users to write custom (essentially hard-coded) tools similar to the ones added in this PR.
Let us know what you think of this!