Quick start

Get up and running in 1 minute

To start trying Percolate, clone the repo and from the root

git clone https://github.com/Percolation-Labs/percolate.git
docker compose up -d

You now have a postgres instance on port 5438 hat you can log into with postgres:postgres

We manage environment variables that are needed for interacting with LLMs/APIs in different ways but using the percolate cli can be a generally useful way to bootstrap your environment.

You can install percolate-db with pip but lets use the codebase for now...

cd clients/python/percolate
#if you have API keys like OPEN_AI_KEY these are synced into your local instance
python percolate/cli.py add env --sync

Another thing you can do is index Percolate files so you can ask questions about Percolate. This will use your Open AI key to generate embeddings.

python percolate/cli.py index 

Now you can ask questions from the cli

python percolate/cli.py ask 'are there SQL functions in Percolate for interacting with models like Claude?'

---

Percolate is a database - it wraps Postgres and adds extensions for vector and graph data. It also pushes agentic AI down into the data tier. Using your favourite Postgres client,

select * from percolate('What is the capital of ireland?')
--try different models
--select * from percolate('how can percolate help me with creating agentic systems',
--  'deepseek-chat')
--see what Models are in Percolate by default
--select * from p8."LangaugeModelApi"

This trivial example tests that we are connected to a langauge model(s) without using tools or data

If we want to use an Agent we can try the built in ones as an example

To understand creating agents see Add agents or Percolating Python-first

 select * from percolate_with_agent('give a brief summary of percolate', 
                                    'p8.PercolateAgent')

If you have created an agent using the example with the sample pets store tools

select * from percolate_with_agent('list some pets that are sold', 'MyFirstAgent')