BERAdapter
Summary
The BERAdapter
is a core module of BER
. The module connects the BER
server with popular project management platforms e.g. GitHub or Jira; company messaging platforms, like Slack. This makes BER
platform agnostic and easy to extend.
It is a tool that by implementing 3rd-party APIs, adds control of BERAgents
to a platform users are already familiar with. BERAdapters
make @ber
instantly available across your workspaces.
Features
The role of the module is to enable frictionless communication between the user and BER
. This component is used for task discussion and remote task execution. Configuration about how to select BERAgents
can also be communicated here.
All commands and instructions are in natural language. All of the input fields such as attachments can be used to enrich the text.
When remote tasks are executed, and external systems are affected the BERAdapter
can display and accept special forms. Every special form takes precedence over natural language input. This separate format gives control and oversight to the user, similar to having draft, preview or --dry-run
abilities.
Find practical, detailed examples in our tutorial about BERAdapter for GitHub
Inputs
To receive user inputs BERAdapters
have API endpoints that can be used as webhooks. In the general form, the input can be natural language and structured data. In the special form, the input is a formal command, such as approving or rejecting an external change or configuring settings.
Outputs
The output is always received on the same BERAdapter
where the input was sent.
Depending on where the response comes from, an external system or a BERAgent
the structure of the output is different.
The output event is skipped by the same BERAdapter
to prevent infinite reaction-cycles.
Comparative chart
User Intent | Input | Output |
---|---|---|
Action | { approval, rejection } <TASK_HASH> |
external system response |
Discussion | free-form natural language | templated, structured, Agent-generated natural language |
Selection | { label, similarity } |
templated, structured, Agent-generated natural language |