Configuration

We use Pydantic-Settings to manage configuration in Dingent. This allows you to define your configuration in a structured way, with support for environment variables, default values, and type validation. However, we recommend using a TOML file to manage your configuration, which is the default approach in our templates. For secret keys like OPENAI_API_KEY, you can set them as environment variables or use a .env file.

Plugin Configuration

Concept: Tool / Plugin

In the Dingent framework, a Tool or Plugin is the smallest functional unit the system can execute. You can think of it as a specific skill available to an Agent. For example:
- A tool that can check the weather.
- A tool that can send an email.
- A tool that can execute a database query.
- A tool that can say hello, like in our previous examples.

Each tool is created by a developer and encapsulates the logic for interacting with the external world or performing a specific computation. As a user, your main task is to enable and configure these tools within your assistant's configuration to meet your specific needs.

How to Configure Tools?

You do not need a separate configuration file for tools. A tool's configuration is defined directly inside the Assistant that will use it.

The workflow for managing and using plugins is as follows:
- Obtain the Plugin: Written by a developer or downloaded from the community, and placed in the project's assistants/plugins directory.
- View Available Plugins: You can use the command-line tool to see all currently loaded and available plugins.
```
dingent assistants plugin list
```
- Configure the Plugin in an Assistant: In the assistant's configuration file, fill in the specific settings for the plugin you want to use. You can see a detailed example in the "Assistant" documentation below.

Assistant Configuration

Concept: Assistant

An Assistant is a collection of tools oriented toward a specific domain task.

If a "tool" is a screwdriver or a wrench, then an "assistant" is a complete toolbox prepared for a specific job, like "repairing a computer" or "assembling furniture." It contains not only the tools needed to complete the task but also high-level instructions on how to perform it.

For example, you could create:
- A General Chat Assistant: Configured with tools for "greeting," "weather checking," "web search," etc.
- A Data Analysis Assistant: Directly integrates a series of professional tools in its configuration, such as "database connection" and "SQL execution."
- A Customer Service Assistant: Includes configurations for tools like "order lookup," "knowledge base retrieval," and "ticket creation."
  
  As a user, you can create multiple agents with different capabilities and personalities by defining different assistants.
How to Configure an Assistant?

You can define one or more assistants in a single TOML configuration file. Each assistant's configuration primarily includes its basic information (like name), behavioral instructions, and the tools it is authorized to use, along with their detailed configurations.

We recommend using the .toml format to manage your assistant configurations because it clearly represents complex nested structures.

Configuration File Example
- File Location: assistants/config.toml

# "[[assistants]]" defines a specific assistant instance.
# You can define multiple [[assistants]] blocks in the same file.

[[assistants]]
name = "sakila"
host = "127.0.0.1"
port = "8888"
description = """
This assistant can only answer questions about the operations of a DVD rental business stored in the database. The query should be about analyzing business performance, customer behavior, or film inventory.

Use this tool for questions about:
- Sales & Revenue: Find sales figures, total revenue for stores or films, and details about specific payments.
- Customer Analysis: Inquire about customer rental habits, find top customers, or analyze customer demographic data like their location.
- Film & Inventory: Find films by title, genre (category), or actor. Check inventory levels for a specific film at a particular store.
- Store & Staff Operations: Explore information about individual stores, their staff, and the rental transactions they process.
"""

# The "tools" list directly defines and configures the tools available to this assistant.
tools = [
  {
    type = "text2sql",
    name = "sakila_text2sql",
    description = "A tool to translate natural language questions into SQL queries for the Sakila database.",

    # --- Configuration specific to the text2sql type ---
    llm = { model="gpt-4.1", provider="openai" },
    database = {
      name = "sakila",
      uri = "sqlite:///./data/sakila.db",
      schemas_file = "schemas/sakila.py"
    }
  }
]

# [[assistants]]
# name = "another_assistant"
# ... configuration for more assistants ...

Configuration Fields Explained
- name (Required): The unique name you assign to this assistant.
- host (Optional): The IP address for exposing this assistant's service.
- port (Optional): The port for exposing this assistant's service.
- description (Required): A detailed description of this assistant's function. This content also serves as the core instruction (System Prompt) for the large language model. The model will strictly follow these instructions, which determines its personality, behavioral guidelines, and task objectives. You can use """ for multi-line text.
- tools (Required): A list that directly defines and configures all the tools this assistant is authorized to use.
  - This is a list of configuration objects, not just a list of IDs.
  - The fields within each tool object (e.g., type, name, llm, database) are determined by its plugin type.
  - You will need to consult the specific plugin's README.md or related documentation to understand its configurable parameters and how to fill them out.

Plugin Configuration​

Assistant Configuration​

Plugin Configuration

Assistant Configuration