README.md

---
license: cc-by-4.0
dataset_info:
  features:
  - name: system
    dtype: string
  - name: in
    dtype: string
  - name: out
    dtype: string
  splits:
  - name: train
    num_bytes: 302707
    num_examples: 210
  - name: validation
    num_bytes: 12666
    num_examples: 9
  - name: test
    num_bytes: 38455
    num_examples: 26
  download_size: 39743
  dataset_size: 353828
configs:
- config_name: default
  data_files:
  - split: train
    path: data/train-*
  - split: validation
    path: data/validation-*
  - split: test
    path: data/test-*
task_categories:
- text-generation
language:
- en
tags:
- code
- dsl
- robotics
- robot-arm
- interactive-agents
- structured-output
- tool-calling
pretty_name: DIA Intent Sequencer for robot arm
size_categories:
- n<1K
---

# dia-intent-sequencer-robot-arm-dataset

This dataset is used to prototype models for the [DIA DSL module](https://github.com/gh9869827/fifo-dev-dsl/tree/main/fifo_dev_dsl/dia). It serves as a **demonstration and testbed** to evaluate, within the context of the DIA DSL, the model's capability  to engage with users in an attempt to resolve incomplete or ambiguous inputs, and to recover from runtime errors during task execution when possible.

It is provided for demonstration and experimentation purposes only.

It pairs natural language instructions, with symbolic DSL function calls compatible with the module **DIA intent sequencer** format. The natural language instructions refer to commands supported by a custom robotic arm I built for experimentation, which manages an inventory of screws.

Unlike synthetic datasets, this one was **authored by hand** to reflect grounded use cases. It illustrates:

- 🛠 Mapping user intent to tool calls
- 🔄 Attempting to resolve missing parameters using runtime-accessible resources
- 🤖 Attempting to clarify ambiguous input and to handle runtime errors through dialogue

This dataset is part of a set of three **`DIA`** datasets, each demonstrating different capabilities of intent sequencing and tool-based reasoning:

- [`a6188466/dia-intent-sequencer-calculator-dataset`](https://huggingface.co/datasets/a6188466/dia-intent-sequencer-calculator-dataset)  
- [`a6188466/dia-intent-sequencer-calendar-dataset`](https://huggingface.co/datasets/a6188466/dia-intent-sequencer-calendar-dataset)  

---

## 📦 Format

The dataset uses a **wide format** with the following columns:

- `system` — the system prompt that defines the model's task
- `in` — a natural language instruction corresponding to a command supported by my custom robotic arm (e.g., `"retrieve 3 screws that are 10mm long"`)
- `out` — a structured DSL expression encoding tool calls with function and argument mappings (e.g., `retrieve_screw(count=3, length=10)`)

Each entry follows the structure:

```
system → in → out
```

---

## 📋 System Prompt Format

The `system` prompt is a structured instruction that tells the model to translate the user input into one or more **atomic DIA DSL function calls** based on a list of available functions (tools) that my custom robotic arm can perform, such as `retrieve_screw`, `initialize`, and others.

The [DIA LLM invocation strategy](https://github.com/gh9869827/fifo-dev-dsl/tree/main/fifo_dev_dsl/dia#-llm-invocation-strategy) documents the prompts contained in this demo dataset.

Some system prompts contain the following section:

```
You have access to the following sources that can be queried to fill in missing information using QUERY_FILL:
...
```

This allows a model to handle missing or incomplete information, for example, to:

- Query the inventory for relevant values
- Attempt to clarify incomplete or conflicting user input through interaction

> 💡 Example: A user might say, "Give me one of the longest screws." The model will attempt to use `QUERY_FILL(...)` to retrieve the appropriate length at runtime.

---

## 🧠 DIA DSL Project

The DSL used in this dataset is defined in the [DIA module](https://github.com/gh9869827/fifo-dev-dsl/tree/main/fifo_dev_dsl/dia).

For an example using a demo model built with this **demo dataset**, see [`robot_arm.py`](https://github.com/gh9869827/fifo-dev-dsl/blob/main/fifo_dev_dsl/dia/demo/robot_arm.py).

---

## 🚀 Using the Dataset

To load the dataset:

```python
from fifo_tool_datasets.sdk.hf_dataset_adapters.dsl import DSLAdapter
dataset_dict = DSLAdapter().from_hub_to_dataset_dict("a6188466/dia-intent-sequencer-robot-arm-dataset")
print(dataset_dict["train"])
print(dataset_dict["validation"])
print(dataset_dict["test"])
```

---

## 🔧 Upload & Editing Tools

Uploaded via [`fifo-tool-datasets`](https://github.com/gh9869827/fifo-tool-datasets) using the `dsl` adapter.

You can edit or extend the dataset using its `.dat` format and CLI tools.

---

## ⚠️ Disclaimer & Limitations

This dataset was authored by hand for demonstration and experimentation purposes. It is intended solely for prototyping models for the [DIA DSL module](https://github.com/gh9869827/fifo-dev-dsl/tree/main/fifo_dev_dsl/dia).

It focuses on a limited set of custom functions specific to my custom robotic arm and serves as a simple, controlled example. It's not meant to cover the full range of possible user inputs, tasks, or edge cases. The dataset should be independently evaluated and extended to meet the specific requirements of any use case.

This dataset is provided *as is*, without any warranties, express or implied. The authors and contributors assume no responsibility for its accuracy, completeness, or suitability for any purpose.

---

## 🪪 License

This dataset is licensed under **CC BY 4.0**. See [LICENSE](LICENSE) for details.