AgentBase

AI Powered Backend Platform

Create your app algorithm in minutes.

Getting Started

AgentBase facilitates the management of agents with access to your databases and APIs, allowing you to extend, modify, or make decisions using natural language.

Overview

Welcome to the AgentBase documentation. AgentBase is a self-built and self-maintained natural language API powered by an agent orchestration system integrated with an IDE. With AgentBase, you can build natural language self-improving functions that connect to your database, run cloud functions, and interact with AI Agents.

AgentBase is an advanced platform designed to enhance computational efficiency by seamlessly integrating different methodologies. One of its most notable features is its ability to determine autonomously whether to employ probabilistic solutions by deploying specialized agents or to generate deterministic code aimed at optimizing performance and standardizing application programming interfaces (APIs). This dynamic adaptability ensures that each task receives the most suitable approach, balancing flexibility and precision.

By amalgamating these two strategies, AgentBase attains optimal solutions across a wide range of scenarios. Probabilistic methods are particularly effective in situations where outcomes are uncertain and need to account for randomness or probabilistic distribution. In contrast, deterministic code is preferable in scenarios demanding high performance, predictability, and replicability.

AgentBase, by leveraging its dual-approach framework, can provide robust and versatile solutions. This combined methodology not only maximizes performance but also ensures consistent and standardized results. Whether the challenge requires handling complex probabilistic models or executing highly optimized deterministic processes, AgentBase stands ready to deliver optimal outcomes.

Example

A request to create a new post and analyze its content sentiment might look like this:

{
    "function": "create_post",
    "prompt": "create a new post for the user, analyze the content and add the respective tags for the sentiment to the post",
    "data" : {
        "email": "[e-mail]",
        "subject": "post subject",
        "message": "post message"
    },
    "return" : {
        "post_id" : {
            "type" : "string",
            "description" : "the post id"
        }, 
        "tags" : {
            "type" : "array",
            "description" : "list of tags_ids",
            "object" : {
                "tag_id" : {
                    "type" : "string",
                    "description" : "the id of the analyzed sentiment for the post"
                }
            }
        }
    }
}

This prompt will end with the following steps to solve the query:

Analyze Sentiment Agent

Designing an agent to analyze the sentiment of posts and create tags allows for better tuning. AgentBase will create an agent to resolve the prompt.

Create_Tags Function

While not explicitly mentioned in the prompt, AgentBase will create a separate function to ensure better code quality, testing, and handling.

Create_Post Function

The original function represents the API intention. The create_post function is called once the sentiment is analyzed using AI and the tag_ids are available to pass as parameters.

Return

To process the tags array, AgentBase must first execute the Analyze Sentiment and Create_Tags functions before returning the results to your client. A time optimization technique involves creating an agent to analyze sentiment in the background.

Agents

Agents in AgentBase can perform a wide variety of functions beyond creating code for your queries. These agents are capable of testing code, generating comprehensive documentation, setting up firewalls to protect your database, and tracking events related to customer success initiatives. By default, AgentBase comes equipped with a set of highly functional agents designed to serve as Natural Language interpreters for your database. This ensures seamless interaction and efficient management of your database through intuitive language commands, enhancing overall productivity and security.

Prompt Chaining

The best scenario for AgentBase involves utilizing prompt chaining to streamline complex tasks, which are managed by different agents. This approach also includes multi-layer agent validation that employs a variety of models or tasks to ensure accuracy and efficiency.

AgentBase effectively handles data chaining by ensuring that all context is passed to the next agent. This process is crucial for the seamless completion of tasks, as each agent adds its expertise and validation to the workflow.

In more detail, prompt chaining allows AgentBase to break down intricate processes into manageable segments, with each agent specializing in specific areas. This contributes to a robust system where tasks are not only completed but also thoroughly validated.

The multi-layer agent validation process further strengthens this framework. By incorporating multiple models or tasks, AgentBase enhances the reliability and quality of the outcomes, making it a sophisticated tool for managing complex operations.

This structured approach of data chaining ensures that no critical information is lost in transition, fostering a cohesive and efficient working environment. Ultimately, AgentBase leverages this methodology to optimize task management, ensuring every task is handled with precision and care.

Using Agent Base

POST /base

Creates a new flow with one or more actions, installs dependencies, builds tests and executes the resulting code

Headers

Name	Value
Content-Type	application/json
Authorization	Bearer <token>

Body

Name	Type	Required	Description
name	string		Name of the flow to summarize the actions
prompt	string		The instructions for AgentBase to transform in to code
data	object		The data example or parameters for your function to work
schedule	string		The cron calendar for the function to run automatically
return	object		The data example of your expected return
model	string		The model to run for your prompt
errors	array		List or possible errors your function needs to catch
cached	boolean		It will use the last version of the function whenever is available to save time and tokens

Response

200

{
  "run_id": 1,
  "return": {
    "message" : "message example"
  },
  "status" : {
    "code" : 1,
    "status" : "waiting for user",
  },
  "error": null
}

400

{
  "error": "Invalid request"
}

Body Example

{
    "function": "create_lead",
    "prompt": "create a new lead and send an email to the sales team with the info",
    "data" : {
        "email": "[e-mail]",
        "subject": "subject example",
        "message": "message example"
    },
    "return" : {
        "agent_message" : {
            "type" : "string",
            "description" : "the customer success agent message"
        }
    }
}

Under the Hood

AgentBase converts your prompt through specialized agents into a JavaScript function. It installs npm dependencies, creates tests, and documents your code. The resulting code runs on a Node.js server and is accessible through a custom domain. You can always install the generated code on your server.

Calling base generates a new function each time. All versions of the functions are stored and accessible for comparison.

Database Management

AgentBase efficiently manages database integrity, structure, normalization, and field processing. It can automatically create collections, fields, new entries, and design roles for your database, enabling you to start using your API without any manual setup.