Skip to content

Use Cases

Agentics is a versatile framework designed for a wide range of applications involving the manipulation of tabular data and, more broadly, JSON objects of arbitrary structure. Below are the core use cases that Agentics was originally designed to support. These capabilities are built-in and available natively within the system.

Native Capabilities

✅ Information Extraction from Documents

This foundational use case is modeled in Agentics by defining a Pydantic type for the output schema and executing a transduction from a list of input texts extracted from that document for to that type.

Advantages of using Agentics

  • 🚀 Asynchronous Execution: Enables >10× speedup through parallel LLM calls.

  • 🧠 No-Code Interface: Define output types using simple YAML or an interactive editor—no Python code required.

  • 🗂️ Seamless Ingestion from multiple document types: Agentics offers built in import and export capabilities to JSON, CSV, TXT and JSONL documents. Additionally, Agentics uses Docling to enable ingestion of multiple document formats incl. PDF, DOCX, XLSX, HTML, images, and more .

Application scenarios

  • Information Extraction from Financial Reports, Medical Records, Invoices, Technical Documentation.
  • Quality Evaluation of ETL workflows output, including Text2SQL
  • Automatic population of DBs Tables from texts

✅ Data Imputation in DB Tables

Agentics handles missing values in structured data by importing it into an Agentics object with column-based types. The system then applies self-transduction to each column with missing values, using the available (non-missing) data as few-shot examples.

Advantages of using Agentics

  • ⚡ Asynchronous Execution: Efficient batch processing of imputation tasks.

  • 🔁 Native Self-Transduction: Built-in support for learning from partial data and iteratively filling in missing values.

Application Scenarios

  • Automated Data Science: Inputation of missing values on table is a generalization of supervised learning from positive examples on a multiclass scenario.

  • Data Curation: Inputation of missing value enable data repair and augmentation in DBs

  • Data Enrichment: Dynamic extension of data types enables interactive definition of new dimension.

✅ Structured Retrieval-Augmented Generation (RAG)

Agentics includes a built-in memory component to support structured RAG, where both inputs and outputs are modeled as Pydantic types.

This approach generalizes RAG in two key ways:

  1. 🔣 The input can be any structured object, not just a single query.

  2. 🧩 The output is a structured object, capturing multiple dimensions or aspects of the answer.

Advantages of using Agentics:

  • 🧱 Structured Inputs and Outputs: Fully typed I/O using Pydantic.

  • 🧰 Built-in Memory Server: No additional setup required. Enable Ingestions of Multiple Data and Document types using Docling.

  • ⚙️ Async Execution: Executes RAG operations in parallel for significant performance gains.

Application Scenarios

  • Document QA: This is implemented natively by Agentics by ingesting the document corpus in a memory collection and transducing the question into an answer. Docling enable ingestion of a large variety of document sources. Low code (1 line) implementation in agentics.

  • Text2SQL: it is an excellent case of structured RAG, where the input is a question and additional data about the target source such as the DB schema, and the output is a SQL query which is further executed to return a DataFrame Object. All this is modelled by a single structured RAG operation in Agentics.

✅ Structured Data Workflows

Agentics integrates seamlessly with tools like LangGraph and can infer attributes of state graphs or structured workflows by modeling each step as a self-transducing unit.

This allows you to:

  • Represent states using typed objects

  • Apply self-transduction to infer unknown attributes from known ones

  • Compose steps algebraically to define low-code, multi-stage logic

Advantages of using Agentics

  • ✨ Streamlined Code: More readable and maintainable than typical Langchain-style graphs.

  • 🔗 Pydantic Compatibility: Fully aligned with agentic frameworks using Pydantic types.

  • 🧠 Composable Transduction: Enables advanced, multi-step agentic pipelines.

Application Scenario

  • visual IDEs for GenAI Workflows: Transduction operation among Agentics can be easily modeled by means of flow diagrams in a very intuitive and no code manner, extending the capabilities of frameworks like LangFlow.

  • NO Code ETL Workflows: Agentics enable representation of any type of structured data, providing helpful utilities to asynchronously apply transductions and/or ad hoc logics to modify data