Pipelines that remember their path
Find the needle, remember the haystack
The Problem
Haystack is production AI done right. Modular pipelines. Pluggable components. Built for scale.
Your pipelines process documents, answer questions, and generate content. Beautifully engineered.
Then each run ends, and the pipeline forgets. What documents were most useful? What queries failed? What patterns emerged? Gone.
Your pipeline finds needles in haystacks. But it can't remember where it looked.
How Stompy Helps
Stompy gives your Haystack pipelines persistent memory.
Your pipelines remember: - Which documents answered which types of questions - Query patterns that work (and don't) - Processing decisions and their outcomes - User feedback and corrections over time
Pipelines that learn from every run.
Integration Walkthrough
Add Stompy to your Haystack pipeline
Use Haystack's component system to add Stompy memory tools.
from haystack import Pipelinefrom haystack.components.generators import OpenAIGeneratorfrom haystack_integrations.tools.mcp import MCPToolProviderimport os# Connect to Stompystompy = MCPToolProvider(url="https://mcp.stompy.ai/sse",headers={"Authorization": f"Bearer {os.environ['STOMPY_TOKEN']}"})pipeline = Pipeline()pipeline.add_component("generator", OpenAIGenerator())pipeline.add_component("memory", stompy)# Pipeline now has persistent memory
What You Get
- Pipelines learn from every run
- Automatic session handovers between executions
- Semantic search finds similar past queries
- Delta evaluation prevents redundant storage
- Version history tracks pipeline evolution
Ready to give Haystack a memory?
Join the waitlist and be the first to know when Stompy is ready. Your Haystack projects will never forget again.