CPEE-LLM

This gem realizes the 3 CPEE Conversational Agents

• CM-A for creating models
• EP-A for selecting endpoints
• DF-A for creating and validating dataflow

Installation

gem install --user cpee-llm
mkdir ~/run
cd ~/run
cpee-llm new cllm # scaffold; you maybe need to add installed bins to PATH
cd cllm
vim cpee-llm.conf # only add the connect adapters you need
vim connect_gemini # add your key
vim connect_gpt # add your key
vim connect_morpheus # add your key, and change url of you morpheus hosting
# connect adapters can be added or deleted as you wish, see rubyllm for
# possible config options. The myllm variable always contains the requested
# model.
./cpee-llm start

These commands install and scaffold a sample server. In the same directory you see a set of sample connectors.

Supported LLMs

We support all LLMs that https://rubyllm.org supports. Check it out.

---

1. Text → Process Model

POST https://cpee.org/llm/

Converts a textual process description into a process model or adapts an existing model.

Parameters

Parameter	Type	Description
rpst_xml	text/xml	Existing CPEE XML model
user_input	text/plain	Natural language process description
llm	text/plain	LLM identifier
prompt_type	text/plain	generate_noendpoints or adapt_noendpoints
temperature	text/plain (optional)	Defaults to 0

Important: Parameter order matters.

rpst_ xml contains the content of the <description> tag from the CPEE testset

Empty RPST Example

<description xmlns="http://cpee.org/ns/description/1.0"/>

Response

{
  "user_input": "...",
  "used_llm": "...",
  "input_cpee": "...",
  "input_intermediate": "...",
  "output_intermediate": "...",
  "output_cpee": "...",
  "status": "..."
}

Key Outputs

• output_intermediate → generated process model in Mermaid .js format
• output_cpee → CPEE XML process model

cURL Example

curl -X POST https://cpee.org/llm/ \
  -H 'Content-Type:multipart/form-data' \
  -F "rpst_xml=@cpee_empty_example;type=text/xml" \
  -F "user_input=Create task A after dismissal review.;type=text/plain" \
  -F "llm=gemini-2.5-flash-lite;type=text/plain" \
  -F "prompt_type=generate_noendpoints;type=text/plain"

---

2. Process Model → Text

POST https://cpee.org/llm/text/llm/

Generates a textual process description from a process model.

Parameters

Parameter	Type
rpst_xml	text/xml
llm	string

Response

{
  "input_cpee": "...",
  "input_intermediate": "...",
  "output_text": "...",
  "status": "..."
}

Key Output

• output_text → Generated process description

cURL Example

curl -X POST https://cpee.org/llm/text/llm/ \
  -H 'Content-Type:multipart/form-data' \
  -F "rpst_xml=@cpee_example;type=text/xml" \
  -F "llm=gemini-2.5-flash-lite"

---

3. Generic Functionality

POST https://cpee.org/llm/generic/

Performs arbitrary LLM tasks using a system prompt and user input.

Parameters

Parameter	Type
llm	text/plain
user_input	text/plain
system_prompt	text/plain
format	text/plain (true/false)
temperature	text/plain (optional)

Response

{
  "user_input": "...",
  "used_llm": "...",
  "system_prompt": "...",
  "llm_response": "...",
  "status": "..."
}

Notes

• format=true requests JSON output (guarantees valid JSON but not any specific structure).

List of supporteb providers: https://rubyllm.com/chat/#getting-structured-output

cURL Example

curl -X POST https://cpee.org/llm/generic/ \
  -H 'Content-Type:multipart/form-data' \
  -F "llm=mistralai/Ministral-3-14B-Reasoning-2512;type=text/plain" \
  -F "user_input=The MPON sends the dismissal to the MPOO.;type=text/plain" \
  -F "system_prompt=Return the list of tasks.;type=text/plain" \
  -F "format=false;type=text/plain"

---