Bringing LLM Power to Every Pipeline – The Multi-Model Prompt Processor

Need a quick sentiment check, a rolling summary, or a safety rule that’s just a bit too fuzzy for SQL?
With the Multi-Model Prompt Processor you can drop the full might of OpenAI, Anthropic, or your own Llama into any Apache StreamPipes pipeline—no extra code required.

Below you’ll find:

• A quick tour of the options
• Five every day use-cases
• A hands-on GIF tutorial
• Tips on writing prompts that return only what you need
• A full list of the models that work out of the box

---

What does the processor do?

Option	What it controls
Model provider / name	Choose any OpenAI or Anthropic model, or point to a Llama on Ollama.
Temperature	0 = repeatable answers, 1 = creative answers.
History strategy	Stateless, Windowed (N turns), or Full history.
Window size	How many past turns to keep in Windowed mode.
Multiple input fields	Combine several event attributes into one user message.

The processor sends your prompt + current event data to the model, grabs the reply, and writes it to a new field llmResponse; ideal when the answer must flow straight into the next processor.

---

Five ways to use it

#	Scenario	Prompt idea (few-shot style)	History
1	Data-quality gate – flag numbers > 15	Return WARNING if value > 15 else OK. Example → 12 → OK, 18 → WARNING.	Stateless
2	Rolling minimum of a sensor	Pass previousMin + current. “Return the smallest of the two numbers only.”	Windowed (1)
3	Five-tweet sentiment	Concatenate last 5 tweets. “Return POSITIVE / NEGATIVE / NEUTRAL only.”	Windowed (5)
4	Chat summariser for support tickets	System prompt: “Write one-line summary of the whole conversation.”	Full
5	Dynamic threshold helper	Feed last 10 load values. “Return max(values)*1.1 rounded 0 decimals.”	Windowed (10)

(Replace the $placeholders$ with the actual stream attributes in the mapping dialog.)

---

Tutorial – from stream to JSON warning

Below is a short clip showing the processor in action in a temperature pipeline.
The prompt forces the model to output a two-field JSON—perfect for machine parsing.

Model: claude-3-5-sonnet-20241022
Prompt Used:

You are an assistant that ALWAYS replies with a single JSON object
containing two string fields: "status" and "message".

Rules
1. If the temperature is strictly greater than 45 °C:
     • "status" must be "WARNING"
     • "message" must explain that the temperature is above the safe limit
2. Otherwise:
     • "status" must be "OK"
     • "message" must note that the temperature is within the safe range
3. Do not output anything except the JSON object.

### Examples

Temperature: 50
{
  "status": "WARNING",
  "message": "Temperature exceeds the 45 °C threshold"
}

Temperature: 30
{
  "status": "OK",
  "message": "Temperature is within the safe range"
}

Temperature: 45
{
  "status": "OK",
  "message": "Temperature is within the safe range"
}

Temperature: 60
{
  "status": "WARNING",
  "message": "Temperature exceeds the 45 °C threshold"
}

The model now returns either:

{
  "status": "WARNING",
  "message": "Temperature exceeds the 45 °C threshold"
}

or

{
  "status": "OK",
  "message": "Temperature is within the safe range"
}

Nothing else—exactly what we want for downstream processors.

---

Prompt writing tips for event streams

1. State the output format first: “Reply with OK or WARNING only.”
2. Give one or two concrete examples – LLMs copy patterns.
3. End with a clear cue: “Answer:” or “Output:”.
4. Keep temperature ≤ 0.3 when strict structure matters.
5. Short replies reduce context size in Windowed / Full modes.

---

Supported models

Anthropic models

claude-3-7-sonnet-20250219
claude-3-5-sonnet-20241022
claude-3-5-haiku-20241022
claude-3-5-sonnet-20240620
claude-3-opus-20240229
claude-3-sonnet-20240229
claude-3-haiku-20240307
claude-2.1
claude-2.0

OpenAI models

gpt-3.5-turbo              gpt-3.5-turbo-1106     gpt-3.5-turbo-0125
gpt-3.5-turbo-16k          gpt-4                  gpt-4-0613
gpt-4-turbo-preview        gpt-4-1106-preview     gpt-4-0125-preview
gpt-4-turbo                gpt-4-turbo-2024-04-09 gpt-4-32k
gpt-4-32k-0613             gpt-4o                 gpt-4o-2024-05-13
gpt-4o-2024-08-06          gpt-4o-2024-11-20      gpt-4o-mini
gpt-4o-mini-2024-07-18     o1                     o1-mini
o1-2024-12-17              o1-mini-2024-09-12     o1-preview
o1-preview-2024-09-12      o3-mini-2025-01-31

Llama suggestions (Ollama)

llama3, llama3:8b, llama2, mistral-7b-instruct, or any custom model you have pulled.

---

Final thoughts

Whether you need a lightweight rule engine, live summaries, or just a smarter filter, the Multi-Model Prompt Processor drops straight into your pipeline and starts working within minutes.

Give it a try, and let us know what you build!