# h2oai/h2o-llmstudio architecture

> Architecture diagram of the GitHub repository h2oai/h2o-llmstudio, generated by GitDiagram from its file tree, README and a sample of source files (last updated 2026-10-01). It is an AI-made overview: check details against the source.

- Interactive diagram: https://gitdiagram.com/h2oai/h2o-llmstudio
- Repository: https://github.com/h2oai/h2o-llmstudio

## Overview

H2O LLM Studio is a GUI and CLI framework for fine-tuning LLMs. Users configure a project and dataset, create and launch an experiment, monitor and compare results, then chat with or export a trained model. The graph captures the GUI workflow and training/evaluation layers; the README also documents LoRA, DPO, multiple problem types, W&B tracking, and Hugging Face export. Some subsystem links use README-level evidence because their implementations were not sampled.

Main entry points: [`llm_studio/train.py`](https://github.com/h2oai/h2o-llmstudio/blob/main/llm_studio/train.py) (Training orchestrator), [`llm_studio/app_utils/handlers.py`](https://github.com/h2oai/h2o-llmstudio/blob/main/llm_studio/app_utils/handlers.py) (UI handlers), [`llm_studio/app_utils/sections/experiment.py`](https://github.com/h2oai/h2o-llmstudio/blob/main/llm_studio/app_utils/sections/experiment.py) (Experiment UI), [`llm_studio/app_utils/sections/dataset.py`](https://github.com/h2oai/h2o-llmstudio/blob/main/llm_studio/app_utils/sections/dataset.py) (Dataset UI)

## Diagram (Mermaid)

```mermaid
flowchart TD

subgraph group_gui["GUI and projects"]
  node_app["Wave app<br/>[app.py]"]
  node_routes["UI handlers<br/>[handlers.py]"]
  node_home["Home and projects<br/>[home.py]"]
  node_dataset_ui["Dataset UI<br/>[dataset.py]"]
  node_experiment_ui["Experiment UI<br/>[experiment.py]"]
  node_chat_ui["Model chat<br/>[chat.py]"]
  node_settings_ui["User settings<br/>[settings.py]"]
end

subgraph group_data["Data and configuration"]
  node_config["App configuration<br/>[config.py]"]
  node_db[("Experiment database<br/>[db.py]")]
  node_data_utils["Dataset preparation<br/>[data_utils.py]"]
  node_conversation["Conversation chains"]
  node_text_utils["Tokenization and text<br/>[text_utils.py]"]
  node_problem_configs["Problem configurations"]
  node_config_utils["Config conversion<br/>[config_utils.py]"]
end

subgraph group_training["Training and evaluation"]
  node_trainer["Training orchestrator<br/>[train.py]"]
  node_modeling["Modeling utilities<br/>[modeling_utils.py]"]
  node_models["Task models"]
  node_losses["Task losses"]
  node_metrics["Evaluation metrics"]
end

subgraph group_results["Results and integrations"]
  node_plots["Training plots"]
  node_logs["Logging and charts<br/>[logging_utils.py]"]
  node_export["Model export<br/>[export_utils.py]"]
  node_hf(("Hugging Face Hub"))
  node_wandb(("Weights &amp; Biases"))
end

node_user(("User"))

node_user -->|"uses"| node_app
node_app -->|"dispatches"| node_routes
node_routes -->|"routes to"| node_home
node_routes -->|"routes to"| node_dataset_ui
node_routes -->|"routes to"| node_experiment_ui
node_routes -->|"routes to"| node_chat_ui
node_routes -->|"routes to"| node_settings_ui
node_home -.->|"reads"| node_config
node_dataset_ui -->|"prepares data"| node_data_utils
node_dataset_ui -->|"uses"| node_conversation
node_conversation -->|"calls"| node_text_utils
node_experiment_ui -->|"configures"| node_problem_configs
node_experiment_ui -->|"uses"| node_config_utils
node_experiment_ui -.->|"starts"| node_trainer
node_trainer -->|"loads config"| node_config_utils
node_trainer -->|"loads data"| node_data_utils
node_trainer -->|"uses"| node_modeling
node_modeling -.->|"supports"| node_models
node_modeling -.->|"supports"| node_losses
node_trainer -.->|"evaluates with"| node_metrics
node_trainer -->|"logs"| node_logs
node_trainer -.->|"produces plots"| node_plots
node_experiment_ui -.->|"reads and writes"| node_db
node_chat_ui -.->|"chats with"| node_models
node_experiment_ui -.->|"exports model"| node_export
node_export -->|"publishes to"| node_hf
node_trainer -.->|"tracks runs"| node_wandb

click node_app "https://github.com/h2oai/h2o-llmstudio/blob/main/llm_studio/app.py"
click node_routes "https://github.com/h2oai/h2o-llmstudio/blob/main/llm_studio/app_utils/handlers.py"
click node_home "https://github.com/h2oai/h2o-llmstudio/blob/main/llm_studio/app_utils/sections/home.py"
click node_dataset_ui "https://github.com/h2oai/h2o-llmstudio/blob/main/llm_studio/app_utils/sections/dataset.py"
click node_experiment_ui "https://github.com/h2oai/h2o-llmstudio/blob/main/llm_studio/app_utils/sections/experiment.py"
click node_chat_ui "https://github.com/h2oai/h2o-llmstudio/blob/main/llm_studio/app_utils/sections/chat.py"
click node_settings_ui "https://github.com/h2oai/h2o-llmstudio/blob/main/llm_studio/app_utils/sections/settings.py"
click node_config "https://github.com/h2oai/h2o-llmstudio/blob/main/llm_studio/app_utils/config.py"
click node_db "https://github.com/h2oai/h2o-llmstudio/blob/main/llm_studio/app_utils/db.py"
click node_data_utils "https://github.com/h2oai/h2o-llmstudio/blob/main/llm_studio/src/utils/data_utils.py"
click node_conversation "https://github.com/h2oai/h2o-llmstudio/blob/main/llm_studio/src/datasets/conversation_chain_handler.py"
click node_text_utils "https://github.com/h2oai/h2o-llmstudio/blob/main/llm_studio/src/datasets/text_utils.py"
click node_problem_configs "https://github.com/h2oai/h2o-llmstudio/tree/main/llm_studio/python_configs"
click node_config_utils "https://github.com/h2oai/h2o-llmstudio/blob/main/llm_studio/src/utils/config_utils.py"
click node_trainer "https://github.com/h2oai/h2o-llmstudio/blob/main/llm_studio/train.py"
click node_modeling "https://github.com/h2oai/h2o-llmstudio/blob/main/llm_studio/src/utils/modeling_utils.py"
click node_models "https://github.com/h2oai/h2o-llmstudio/tree/main/llm_studio/src/models"
click node_losses "https://github.com/h2oai/h2o-llmstudio/tree/main/llm_studio/src/losses"
click node_metrics "https://github.com/h2oai/h2o-llmstudio/tree/main/llm_studio/src/metrics"
click node_plots "https://github.com/h2oai/h2o-llmstudio/tree/main/llm_studio/src/plots"
click node_logs "https://github.com/h2oai/h2o-llmstudio/blob/main/llm_studio/src/utils/logging_utils.py"
click node_export "https://github.com/h2oai/h2o-llmstudio/blob/main/llm_studio/src/utils/export_utils.py"

classDef toneNeutral fill:#f8fafc,stroke:#334155,stroke-width:1.5px,color:#0f172a
classDef toneBlue fill:#dbeafe,stroke:#2563eb,stroke-width:1.5px,color:#172554
classDef toneAmber fill:#fef3c7,stroke:#d97706,stroke-width:1.5px,color:#78350f
classDef toneMint fill:#dcfce7,stroke:#16a34a,stroke-width:1.5px,color:#14532d
classDef toneRose fill:#ffe4e6,stroke:#e11d48,stroke-width:1.5px,color:#881337
classDef toneIndigo fill:#e0e7ff,stroke:#4f46e5,stroke-width:1.5px,color:#312e81
classDef toneTeal fill:#ccfbf1,stroke:#0f766e,stroke-width:1.5px,color:#134e4a
class node_app,node_routes,node_home,node_dataset_ui,node_experiment_ui,node_chat_ui,node_settings_ui,node_user toneBlue
class node_config,node_db,node_data_utils,node_conversation,node_text_utils,node_problem_configs,node_config_utils toneAmber
class node_trainer,node_modeling,node_models,node_losses,node_metrics toneMint
class node_plots,node_logs,node_export,node_hf,node_wandb toneRose
```

## Components

### GUI and projects

- [Wave app](https://github.com/h2oai/h2o-llmstudio/blob/main/llm_studio/app.py): `llm_studio/app.py`
- [UI handlers](https://github.com/h2oai/h2o-llmstudio/blob/main/llm_studio/app_utils/handlers.py): `llm_studio/app_utils/handlers.py`
- [Home and projects](https://github.com/h2oai/h2o-llmstudio/blob/main/llm_studio/app_utils/sections/home.py): `llm_studio/app_utils/sections/home.py`
- [Dataset UI](https://github.com/h2oai/h2o-llmstudio/blob/main/llm_studio/app_utils/sections/dataset.py): `llm_studio/app_utils/sections/dataset.py`
- [Experiment UI](https://github.com/h2oai/h2o-llmstudio/blob/main/llm_studio/app_utils/sections/experiment.py): `llm_studio/app_utils/sections/experiment.py`
- [Model chat](https://github.com/h2oai/h2o-llmstudio/blob/main/llm_studio/app_utils/sections/chat.py): `llm_studio/app_utils/sections/chat.py`
- [User settings](https://github.com/h2oai/h2o-llmstudio/blob/main/llm_studio/app_utils/sections/settings.py): `llm_studio/app_utils/sections/settings.py`

### Data and configuration

- [App configuration](https://github.com/h2oai/h2o-llmstudio/blob/main/llm_studio/app_utils/config.py): `llm_studio/app_utils/config.py`
- [Experiment database](https://github.com/h2oai/h2o-llmstudio/blob/main/llm_studio/app_utils/db.py): `llm_studio/app_utils/db.py`
- [Dataset preparation](https://github.com/h2oai/h2o-llmstudio/blob/main/llm_studio/src/utils/data_utils.py): `llm_studio/src/utils/data_utils.py`
- [Conversation chains](https://github.com/h2oai/h2o-llmstudio/blob/main/llm_studio/src/datasets/conversation_chain_handler.py): `llm_studio/src/datasets/conversation_chain_handler.py`
- [Tokenization and text](https://github.com/h2oai/h2o-llmstudio/blob/main/llm_studio/src/datasets/text_utils.py): `llm_studio/src/datasets/text_utils.py`
- [Problem configurations](https://github.com/h2oai/h2o-llmstudio/tree/main/llm_studio/python_configs): `llm_studio/python_configs`
- [Config conversion](https://github.com/h2oai/h2o-llmstudio/blob/main/llm_studio/src/utils/config_utils.py): `llm_studio/src/utils/config_utils.py`

### Training and evaluation

- [Training orchestrator](https://github.com/h2oai/h2o-llmstudio/blob/main/llm_studio/train.py): `llm_studio/train.py`
- [Modeling utilities](https://github.com/h2oai/h2o-llmstudio/blob/main/llm_studio/src/utils/modeling_utils.py): `llm_studio/src/utils/modeling_utils.py`
- [Task models](https://github.com/h2oai/h2o-llmstudio/tree/main/llm_studio/src/models): `llm_studio/src/models`
- [Task losses](https://github.com/h2oai/h2o-llmstudio/tree/main/llm_studio/src/losses): `llm_studio/src/losses`
- [Evaluation metrics](https://github.com/h2oai/h2o-llmstudio/tree/main/llm_studio/src/metrics): `llm_studio/src/metrics`

### Results and integrations

- [Training plots](https://github.com/h2oai/h2o-llmstudio/tree/main/llm_studio/src/plots): `llm_studio/src/plots`
- [Logging and charts](https://github.com/h2oai/h2o-llmstudio/blob/main/llm_studio/src/utils/logging_utils.py): `llm_studio/src/utils/logging_utils.py`
- [Model export](https://github.com/h2oai/h2o-llmstudio/blob/main/llm_studio/src/utils/export_utils.py): `llm_studio/src/utils/export_utils.py`
- **Hugging Face Hub**
- **Weights & Biases**

### Other components

- **User**

## Connections

- User → Wave app: uses (evidence: `README.md`)
- Wave app → UI handlers: dispatches (evidence: `llm_studio/app.py`)
- UI handlers → Home and projects: routes to (evidence: `llm_studio/app_utils/handlers.py`)
- UI handlers → Dataset UI: routes to (evidence: `llm_studio/app_utils/handlers.py`)
- UI handlers → Experiment UI: routes to (evidence: `llm_studio/app_utils/handlers.py`)
- UI handlers → Model chat: routes to (evidence: `llm_studio/app_utils/handlers.py`)
- UI handlers → User settings: routes to (evidence: `llm_studio/app_utils/handlers.py`)
- Home and projects → App configuration: reads
- Dataset UI → Dataset preparation: prepares data (evidence: `llm_studio/app_utils/sections/dataset.py`)
- Dataset UI → Conversation chains: uses (evidence: `llm_studio/app_utils/sections/dataset.py`)
- Conversation chains → Tokenization and text: calls (evidence: `llm_studio/src/datasets/conversation_chain_handler.py`)
- Experiment UI → Problem configurations: configures (evidence: `llm_studio/app_utils/sections/experiment.py`)
- Experiment UI → Config conversion: uses (evidence: `llm_studio/app_utils/sections/experiment.py`)
- Experiment UI → Training orchestrator: starts
- Training orchestrator → Config conversion: loads config (evidence: `llm_studio/train.py`)
- Training orchestrator → Dataset preparation: loads data (evidence: `llm_studio/train.py`)
- Training orchestrator → Modeling utilities: uses (evidence: `llm_studio/train.py`)
- Modeling utilities → Task models: supports (evidence: `llm_studio/src/models/text_dpo_modeling_model.py`)
- Modeling utilities → Task losses: supports
- Training orchestrator → Evaluation metrics: evaluates with
- Training orchestrator → Logging and charts: logs (evidence: `llm_studio/train.py`)
- Training orchestrator → Training plots: produces plots
- Experiment UI → Experiment database: reads and writes
- Model chat → Task models: chats with
- Experiment UI → Model export: exports model (evidence: `llm_studio/app_utils/sections/experiment.py`)
- Model export → Hugging Face Hub: publishes to (evidence: `README.md`)
- Training orchestrator → Weights & Biases: tracks runs (evidence: `README.md`)
