> ## Documentation Index
> Fetch the complete documentation index at: https://uom-demo.vercel.app/docs/llms.txt
> Use this file to discover all available pages before exploring further.

# UOM Deep Agent

> The main entrypoint for the UOM Deep Agent (UOM Assistant), which provides general-purpose agent capabilities, including invoking UOM Translator as an `AsyncSubagent` and having complete access to local/sandbox shell and file system.[<span className="flex items-center gap-1"><Icon icon="external-link"/>View Source</span>](https://github.com/corovcam/Universal-Object-Mapping/tree/main/services/orchestrator/src/uom_deep_agent/uom_agent.py)

Demo coding agent using ACP.

#### uom\_deep\_agent.uom\_agent.get\_example\_input

```python theme={null}
def get_example_input(file_name: str) -> str
```

[<span className="flex items-center gap-1"><Icon icon="external-link" />View Source</span>](https://github.com/corovcam/Universal-Object-Mapping/blob/main/services/orchestrator/src/uom_deep_agent/uom_agent.py#L88)

Load input from a file, with caching.

#### uom\_deep\_agent.uom\_agent.configurable\_model

```python theme={null}
@wrap_model_call()
async def configurable_model(
    request: ModelRequest[Context],
    handler: Callable[[ModelRequest[Context]],
                      ModelResponse]) -> ModelResponse
```

[<span className="flex items-center gap-1"><Icon icon="external-link" />View Source</span>](https://github.com/corovcam/Universal-Object-Mapping/blob/main/services/orchestrator/src/uom_deep_agent/uom_agent.py#L118)

A wrapper middleware for deepagents model calls that dynamically configures the target ChatModel.

It extracts the provider/model string from the runtime context, configures OpenAI/Ollama settings,
loads the correct chat model, and forwards the overridden request.

**Arguments**:

* `request` - The dynamic model invocation request containing context.
* `handler` - The downstream handler callable to execute the model call.

**Returns**:

* `ModelResponse` - The generated response from the configured model.

#### uom\_deep\_agent.uom\_agent.custom\_system\_prompt

```python theme={null}
@dynamic_prompt
def custom_system_prompt(request: ModelRequest[Context]) -> str
```

[<span className="flex items-center gap-1"><Icon icon="external-link" />View Source</span>](https://github.com/corovcam/Universal-Object-Mapping/blob/main/services/orchestrator/src/uom_deep_agent/uom_agent.py#L149)

Always append the system prompt from the runtime context to the system prompt from the request.

#### uom\_deep\_agent.uom\_agent.build\_deep\_agent

```python theme={null}
def build_deep_agent(model: BaseChatModel,
                     dotnet_sandbox: DaytonaSandbox,
                     java_sandbox: DaytonaSandbox,
                     extra_middleware: list[Any] | None = None,
                     checkpointer: Checkpointer | None = None,
                     context: AgentSessionContext | None = None)
```

[<span className="flex items-center gap-1"><Icon icon="external-link" />View Source</span>](https://github.com/corovcam/Universal-Object-Mapping/blob/main/services/orchestrator/src/uom_deep_agent/uom_agent.py#L188)

Build and assemble the DeepAgent workspace graph, wrapping it with specialized sub-agents and tools.

**Arguments**:

* `model` - The base language model to use for core agent reasoning.
* `dotnet_sandbox` - Pre-allocated Daytona sandbox instance for C# compilation.
* `java_sandbox` - Pre-allocated Daytona sandbox instance for Java Maven compilation.
* `extra_middleware` - Optional list of additional middleware handlers to register.
* `checkpointer` - Optional persistence checkpointer for the graph's execution state.
* `context` - Optional agent session context containing custom rules and current mode.

**Returns**:

* `DeepAgent` - The constructed and compiled DeepAgent workspace graph instance.

#### uom\_deep\_agent.uom\_agent.build\_uom\_agent

```python theme={null}
async def build_uom_agent()
```

[<span className="flex items-center gap-1"><Icon icon="external-link" />View Source</span>](https://github.com/corovcam/Universal-Object-Mapping/blob/main/services/orchestrator/src/uom_deep_agent/uom_agent.py#L326)

Builds the Universal Object Mapping Assistant using LangChain's DeepAgents SDK.