# paradigma-inc/limite-violetto architecture

> Architecture diagram of the GitHub repository paradigma-inc/limite-violetto, 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/paradigma-inc/limite-violetto
- Repository: https://github.com/paradigma-inc/limite-violetto

## Overview

Limite Violetto is a vLLM plugin that serves Limite mathematical-language-model checkpoints. The principal workflow is a client request to vLLM, plugin registration of the checkpoint configuration and model architecture, validated model construction and weight loading, then autoregressive inference and a response. The graph preserves the model’s attention, feed-forward, and residual-mixing stages. Source excerpts are partial; relationships not shown in excerpts or documented in the README are omitted.

Main entry points: [`src/limite_vllm/modeling.py`](https://github.com/paradigma-inc/limite-violetto/blob/main/src/limite_vllm/modeling.py) (Causal model), [`src/limite_vllm/register.py`](https://github.com/paradigma-inc/limite-violetto/blob/main/src/limite_vllm/register.py) (Plugin registration), [`src/limite_vllm/configuration.py`](https://github.com/paradigma-inc/limite-violetto/blob/main/src/limite_vllm/configuration.py) (Checkpoint configuration), [`src/limite_vllm/validation.py`](https://github.com/paradigma-inc/limite-violetto/blob/main/src/limite_vllm/validation.py) (Runtime validation)

## Diagram (Mermaid)

```mermaid
flowchart TD

subgraph group_plugin["Plugin integration"]
  node_register["Plugin registration<br/>[register.py]"]
end

subgraph group_contract["Checkpoint contract"]
  node_identities["Model identities<br/>[contract.py]"]
  node_config["Checkpoint configuration<br/>[configuration.py]"]
  node_validation["Runtime validation<br/>[validation.py]"]
end

subgraph group_execution["Model execution"]
  node_causal_model["Causal model<br/>[modeling.py]"]
  node_decoder["Decoder layers<br/>[modeling.py]"]
  node_token_embeddings["Token embeddings<br/>[modeling.py]"]
  node_logit_head["Logit head<br/>[modeling.py]"]
  node_weight_loading["Weight loading<br/>[modeling.py]"]
end

subgraph group_features["Model features"]
  node_attention["Attention<br/>[_components.py]"]
  node_mlp["Feed-forward MLP<br/>[_components.py]"]
  node_mudd["Residual mixing<br/>[_components.py]"]
  node_rotary["Rotary embeddings<br/>[_components.py]"]
end

node_client(("Serving client"))
node_server["vLLM serving"]
node_checkpoint[("HF checkpoint")]

node_client -->|"sends request"| node_server
node_server -->|"loads plugin"| node_register
node_register -->|"uses identities"| node_identities
node_register -->|"registers config"| node_config
node_register -->|"registers model"| node_causal_model
node_checkpoint -->|"provides model"| node_server
node_server -->|"parses config"| node_config
node_causal_model -->|"validates config"| node_validation
node_validation -->|"checks architecture"| node_config
node_causal_model -->|"embeds tokens"| node_token_embeddings
node_causal_model -->|"runs layers"| node_decoder
node_decoder -->|"uses attention"| node_attention
node_decoder -->|"uses MLP"| node_mlp
node_decoder -->|"mixes residuals"| node_mudd
node_decoder -->|"applies rotary"| node_rotary
node_causal_model -->|"computes logits"| node_logit_head
node_checkpoint -->|"supplies weights"| node_weight_loading
node_weight_loading -->|"loads parameters"| node_causal_model
node_server -->|"returns generation"| node_client

click node_register "https://github.com/paradigma-inc/limite-violetto/blob/main/src/limite_vllm/register.py"
click node_identities "https://github.com/paradigma-inc/limite-violetto/blob/main/src/limite_vllm/contract.py"
click node_config "https://github.com/paradigma-inc/limite-violetto/blob/main/src/limite_vllm/configuration.py"
click node_validation "https://github.com/paradigma-inc/limite-violetto/blob/main/src/limite_vllm/validation.py"
click node_causal_model "https://github.com/paradigma-inc/limite-violetto/blob/main/src/limite_vllm/modeling.py"
click node_decoder "https://github.com/paradigma-inc/limite-violetto/blob/main/src/limite_vllm/modeling.py"
click node_token_embeddings "https://github.com/paradigma-inc/limite-violetto/blob/main/src/limite_vllm/modeling.py"
click node_logit_head "https://github.com/paradigma-inc/limite-violetto/blob/main/src/limite_vllm/modeling.py"
click node_weight_loading "https://github.com/paradigma-inc/limite-violetto/blob/main/src/limite_vllm/modeling.py"
click node_attention "https://github.com/paradigma-inc/limite-violetto/blob/main/src/limite_vllm/_components.py"
click node_mlp "https://github.com/paradigma-inc/limite-violetto/blob/main/src/limite_vllm/_components.py"
click node_mudd "https://github.com/paradigma-inc/limite-violetto/blob/main/src/limite_vllm/_components.py"
click node_rotary "https://github.com/paradigma-inc/limite-violetto/blob/main/src/limite_vllm/_components.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_register,node_client toneBlue
class node_identities,node_config,node_validation,node_checkpoint toneAmber
class node_causal_model,node_decoder,node_token_embeddings,node_logit_head,node_weight_loading toneMint
class node_attention,node_mlp,node_mudd,node_rotary toneRose
class node_server toneIndigo
```

## Components

### Plugin integration

- [Plugin registration](https://github.com/paradigma-inc/limite-violetto/blob/main/src/limite_vllm/register.py): `src/limite_vllm/register.py`

### Checkpoint contract

- [Model identities](https://github.com/paradigma-inc/limite-violetto/blob/main/src/limite_vllm/contract.py): `src/limite_vllm/contract.py`
- [Checkpoint configuration](https://github.com/paradigma-inc/limite-violetto/blob/main/src/limite_vllm/configuration.py): `src/limite_vllm/configuration.py`
- [Runtime validation](https://github.com/paradigma-inc/limite-violetto/blob/main/src/limite_vllm/validation.py): `src/limite_vllm/validation.py`

### Model execution

- [Causal model](https://github.com/paradigma-inc/limite-violetto/blob/main/src/limite_vllm/modeling.py): `src/limite_vllm/modeling.py`
- [Decoder layers](https://github.com/paradigma-inc/limite-violetto/blob/main/src/limite_vllm/modeling.py): `src/limite_vllm/modeling.py`
- [Token embeddings](https://github.com/paradigma-inc/limite-violetto/blob/main/src/limite_vllm/modeling.py): `src/limite_vllm/modeling.py`
- [Logit head](https://github.com/paradigma-inc/limite-violetto/blob/main/src/limite_vllm/modeling.py): `src/limite_vllm/modeling.py`
- [Weight loading](https://github.com/paradigma-inc/limite-violetto/blob/main/src/limite_vllm/modeling.py): `src/limite_vllm/modeling.py`

### Model features

- [Attention](https://github.com/paradigma-inc/limite-violetto/blob/main/src/limite_vllm/_components.py): `src/limite_vllm/_components.py`
- [Feed-forward MLP](https://github.com/paradigma-inc/limite-violetto/blob/main/src/limite_vllm/_components.py): `src/limite_vllm/_components.py`
- [Residual mixing](https://github.com/paradigma-inc/limite-violetto/blob/main/src/limite_vllm/_components.py): `src/limite_vllm/_components.py`
- [Rotary embeddings](https://github.com/paradigma-inc/limite-violetto/blob/main/src/limite_vllm/_components.py): `src/limite_vllm/_components.py`

### Other components

- **Serving client**
- **vLLM serving**
- **HF checkpoint**

## Connections

- Serving client → vLLM serving: sends request (evidence: `README.md`)
- vLLM serving → Plugin registration: loads plugin (evidence: `README.md`)
- Plugin registration → Model identities: uses identities (evidence: `src/limite_vllm/register.py`)
- Plugin registration → Checkpoint configuration: registers config (evidence: `src/limite_vllm/register.py`)
- Plugin registration → Causal model: registers model (evidence: `src/limite_vllm/register.py`)
- HF checkpoint → vLLM serving: provides model (evidence: `README.md`)
- vLLM serving → Checkpoint configuration: parses config (evidence: `README.md`)
- Causal model → Runtime validation: validates config (evidence: `src/limite_vllm/modeling.py`)
- Runtime validation → Checkpoint configuration: checks architecture (evidence: `src/limite_vllm/validation.py`)
- Causal model → Token embeddings: embeds tokens (evidence: `src/limite_vllm/modeling.py`)
- Causal model → Decoder layers: runs layers (evidence: `src/limite_vllm/modeling.py`)
- Decoder layers → Attention: uses attention (evidence: `src/limite_vllm/modeling.py`)
- Decoder layers → Feed-forward MLP: uses MLP (evidence: `src/limite_vllm/modeling.py`)
- Decoder layers → Residual mixing: mixes residuals (evidence: `src/limite_vllm/modeling.py`)
- Decoder layers → Rotary embeddings: applies rotary (evidence: `src/limite_vllm/modeling.py`)
- Causal model → Logit head: computes logits (evidence: `src/limite_vllm/modeling.py`)
- HF checkpoint → Weight loading: supplies weights (evidence: `src/limite_vllm/modeling.py`)
- Weight loading → Causal model: loads parameters (evidence: `src/limite_vllm/modeling.py`)
- vLLM serving → Serving client: returns generation (evidence: `README.md`)
