Core Concepts
How Neo interacts with your code, data, and AI/ML workflows
How Neo Approaches Work
Neo is an execution-focused AI agent designed to operate inside your development environment. Instead of responding with isolated suggestions, Neo treats each request as a task that may involve multiple steps, files, and iterations.
When you give Neo a task, it plans the work, runs the required steps, and adjusts based on real results such as metrics, logs, and errors.
Agent Characteristics
Neo operates with a core set of behaviors that define how it works inside your project.
-
Context aware
Neo reads your project structure, code, and data during a session so it can make informed decisions instead of starting from scratch each time. -
Self correcting
Neo detects errors in code execution, data processing, or model training and attempts reasonable fixes before continuing, while surfacing issues for review. -
Iterative
Neo improves results based on feedback from validation metrics, evaluation plots, and runtime behavior. This allows it to refine feature sets, model parameters, or pipeline steps across iterations. -
Transparent
Neo records what it does as it works. Plans, logs, generated code, metrics, and artifacts are written to your project and visible in the VS Code output panel so you can review or intervene at any time.
How Tasks Flow Through Neo
This section describes what happens internally when you send a task to Neo.
Submit a Task
Describe your goal in natural language using the Neo panel in VS Code.
Build a classification model to predict customer churnProject Analysis and Planning
Neo inspects your repository and data to understand the environment before writing or running code.
- Scans project structure and existing files
- Identifies datasets, schemas, and formats
- Creates a step-by-step execution plan
Pipeline Execution
Neo executes the planned steps inside your VS Code environment using your local setup and dependencies.
Results and Iteration
Neo generates concrete outputs and uses them to decide what to do next.
- Model files such as Pickle or ONNX
- Metrics, plots, and evaluation reports
- Logs explaining decisions and adjustments
What Neo Can Access
Neo only works with what you explicitly allow inside your environment.
- Local project files and folders opened in VS Code
- Datasets referenced in your instructions or configuration
- Existing code and scripts in the repository
- External resources such as databases or cloud storage when credentials are provided
Neo does not silently modify files. All changes are written to disk and visible for review.
You can stop or adjust a task at any time during execution.
Supported Data Types
Neo can work with the following data sources:
| Category | Formats |
|---|---|
| Tabular Data | CSV, Excel, Parquet |
| Images | JPG, PNG |
| Text | TXT, JSON |
| SQL Databases | PostgreSQL, MySQL |
| Cloud Storage | S3, GCS, Azure Blob |
| Audio | WAV, MP3, FLAC |
AI and Machine Learning Capabilities
Neo supports both modern AI workloads and traditional machine learning tasks. The exact workflow and level of automation depend on your project structure, data quality, and available compute.
-
Tabular ML
Classification, regression, clustering -
Computer Vision
Image classification, object detection, segmentation -
Natural Language Processing
Text classification, named entity recognition, summarization -
LLM fine tuning
LoRA based fine tuning and instruction tuning -
Generative and agentic AI
Retrieval augmented generation (RAG) pipelines and autonomous agents
What Neo Is Not
- Neo is not a conversational chatbot that only suggests code snippets.
- Neo is not a black box that hides execution or decisions.
- Neo does not replace your judgment or take control of your repository.
Neo is designed to assist with execution while keeping you in control of code, data, and outcomes.