> ## 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.

# prompts

> [<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/react_agent/prompts.py)

System Prompts for LangGraph Nodes.

These multi-line strings define the core persona, instructions, and few-shot
examples for the LLM agents operating within the orchestrator's state machine.
They are dynamically interpolated at runtime using str.format() to inject
the chosen frameworks and context variables.

#### react\_agent.prompts.SYSTEM\_PROMPT\_EXTRACTION

Information Extraction prompt used at the very beginning of the graph execution.
Its purpose is strictly to classify intent (translation type) and extract the raw code.

#### react\_agent.prompts.SYSTEM\_PROMPT\_SCHEMA\_INSPECTOR

Schema inspection prompt used prior to code translation.
Guides the agent to use the MCP tools to interact with live relational and NoSQL databases.

#### react\_agent.prompts.build\_system\_prompt

```python theme={null}
async def build_system_prompt(state: State) -> 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/react_agent/prompts.py#L723)

Dynamically build the system prompt based on the specific translation pair.

This function assembles a comprehensive system prompt by combining the base persona,
rules, and dynamic framework configurations retrieved from the file system snippets.
It injects the exact C# and Java dependencies (like pom.xml or .csproj values) directly
into the prompt context so the LLM understands the exact versions it is translating for.