superagent-ai / superagent Goto Github PK
View Code? Open in Web Editor NEWπ₯· Run AI-agents with an API
Home Page: https://docs.superagent.sh
License: Other
π₯· Run AI-agents with an API
Home Page: https://docs.superagent.sh
License: Other
Add a replicate Tool that can work together with an agent.
No response
To get started with SuperAgent, follow these steps:
Clone the SuperAgent repository into a public GitHub repository or fork it from https://github.com/homanp/superagent/fork. If you plan to distribute the code, keep the source code public.
git clone https://github.com/homanp/superagent.git
Set up your .env
file.
cp .env.example .env
To run the script, simply execute it using:
bash setup.sh
You can save the script in a file, e.g., setup.sh
,
#!/bin/bash
# Create and activate the virtual environment
virtualenv venv
source venv/bin/activate
# Install dependencies using Poetry
poetry install
# Run Prisma migration
poetry run prisma migrate dev
# Start the application with auto-reload
uvicorn app.main:app --reload
This script streamlines the setup process by creating the virtual environment, installing dependencies, performing the Prisma migration, and starting the application with auto-reload. It helps ensure a consistent and efficient development environment for Python web applications.
In order to do more granular logging and debugging for an agent we should handle all callbacks from the Agent
or enable a more verbose logging.
This PR is connected to #64
No response
Create a scheduler that can either:
No response
Wolfram Alpha makes all systematic knowledge immediately computable and accessible to everyone.
https://www.wolframalpha.com/
Add endpoints for retrieving a user.
Use prisma
as ORM and add authentication.
Use black
, flake8
for formatting and linting
A user should be able to attach documents/vectorstores for QA
An API token should be present when running an Agent
.
The display button in digital ocean does not work.
An error is thrown when it is observed that the example file does not exist.
A user should be able to invocate Agents using a JWT token. These tokens need a separate DB with the following fields:
description
, token
Would be great if we could add unit tests to Superagent
.
Use pytest
test suite to add unit tests and integration tests to Superagent.
No response
No response
Add the llm field to Agent model. Should be of type JSONB.
We should introduce versioning of the prompts in order for users to track/evaluate different versions of a specific prompt. Similar to what Replicate
does with their models.
No response
A user should be able to update a prompt
No response
Create a DB schema for Agents
.
name
type
react, plan-solve
Sometimes queries take a long time to execute. When that happens it would be nice to offer webhooks for more flexible DX.
No response
Hey there, is there a possibility, to add:
https://supabase.com/blog/openai-embeddings-postgres-vector
instead of going with pinecone? So in this way we could store the embeddings also locally.
There is an existing project "SuperAgent" which is a popular NPM package for performing requests.
In order to avoid impacting the searchability for each project changing the name would be an effective solution.
Add retrieval chain if an Agent
Β has a Document
connected to it.
The Agent lib is getting bigger for each feature we release. Instead of running everything in the same class we should create a AgentBase
and implement common features there.
Each new Agent
should inherit from the AgentBase
Β class.
No response
Create a prompt model containing the following:
name
template
-> Large text field
inputs
-> JSON
type
-> QA prompt etc.
Autonomous Agents such as AutoGPT or BabyAGI would be a great Agent type to bring into Superagent.
No response
Add memory to agents
Add a new type of Agent that specializes on content aggreation.
This would be an ideal issue to use Llama Index
Β for.
No response
To keep the code consistent we should move the models from the api
Β folder to the lib
folder.
lib
--models
--agents
The invocation endpoint should take a separate API token and has a single key in the payload called message
.
Sometimes, users will upsert documents that need to be more "real-time". Possibility to update the indexes with new content would be a useful feature for this.
No response
A user should be able to add multiple documents to an Agent.
No response
In order to introduce spending limits we would want to save the token usage for an agent in the DB
Create a new DB table for saving tokens and cost estimations.
Each run should be saved as a run entry.
One to many relationship between usage and agent.
Add a new type of tool that enables an Agent to do real-time search via Bing or Google.
No response
Users want to be able to control settings such as max_tokens
, temperature
etc.
No response
Implement the OpenAI Agent type.
No response
One of the great things about @huggingface face is the race to build what I call "mini" models suited for specific use cases. Wrap those models around a specific Agent and you build amazing apps.
Superagent should support these mini models wrapped so that they can be used with Agents.
No response
All Agent callbacks and intermediate steps should be stored in a log.
TBD
No response
No response
Create a browser agent that can search the web and aggregate content.
This could be followed up with controlling the browser in future PRs.
No response
Hey, we I am currently working with @thesushilsharma and we are looking for ways to integrate the GPT4ALL LLM in order to use it as our native LLM instead of the one provided by OpenAI. Are there any methods existing at this stage which would allow us to achieve this?
Add support for pulling in Llamaindex into Superagent.
No response
Some tools require additional authentication params on runtime. Examples of such tools are Wolfram
. The API should support passing required authentication methods similar to YAML documents.
No response
A user should be able to set a prompt on a Document QA Agent.
No response
A declarative, efficient, and flexible JavaScript library for building user interfaces.
π Vue.js is a progressive, incrementally-adoptable JavaScript framework for building UI on the web.
TypeScript is a superset of JavaScript that compiles to clean JavaScript output.
An Open Source Machine Learning Framework for Everyone
The Web framework for perfectionists with deadlines.
A PHP framework for web artisans
Bring data to life with SVG, Canvas and HTML. πππ
JavaScript (JS) is a lightweight interpreted programming language with first-class functions.
Some thing interesting about web. New door for the world.
A server is a program made to process requests and deliver data to clients.
Machine learning is a way of modeling and interpreting data that allows a piece of software to respond intelligently.
Some thing interesting about visualization, use data art
Some thing interesting about game, make everyone happy.
We are working to build community through open source technology. NB: members must have two-factor auth.
Open source projects and samples from Microsoft.
Google β€οΈ Open Source for everyone.
Alibaba Open Source for everyone
Data-Driven Documents codes.
China tencent open source team.