Add Postgres vector store using pgvector

[fabiovalencio]

Add a component to vectorstore to allow saving vectors in a Postgres database with pgvector

[CLAassistant]

All committers have signed the CLA.

[HenryHengZJ]

you might have accidentally removed the .env.example files

[fabiovalencio]

you might have accidentally removed the .env.example files

Sorry about that, I just renamed them to use, putting back.

[HenryHengZJ]

@fabiovalencio were you abel to get this running? I have a docker compose file running like this:

services:
  db:
    image: ankane/pgvector
    ports:
      - 5432:5432
    volumes:
      - ./data:/var/lib/postgresql/data
    environment:
      - POSTGRES_PASSWORD=mypassword
      - POSTGRES_USER=myuser
      - POSTGRES_DB=api

And I created postgres credential in Flowise with the above user and password. However, when running the chatflow got this error: password authentication failed for user "myuser"

[fabiovalencio]

Hi @HenryHengZJ this probably happened because I hadn't set the default table name, I've just fixed it.

[HenryHengZJ]

Hi @HenryHengZJ this probably happened because I hadn't set the default table name, I've just fixed it.

still the same:

were you able to get it working on Flowise following the guide here

[fabiovalencio]

I redid it from the beginning and didn't find any errors.

I started by creating Postgres with Docker

Then I tried accessing the database

I didn't use Flowise from Docker because it doesn't have the Postgres vector store yet. So I started another application with Flowise and I saved the Postgres access credentials.

I tried to create a chat just like yours

I haven't really had any problems. Just wondering, can you access the database via psql?

[HenryHengZJ]

hmm pretty sure its my connection to postgres..

In the docker compose file I have set

- POSTGRES_USER=myuser
- POSTGRES_PASSWORD=mypassword
- POSTGRES_DB=api

Then using cli:

psql postgresql://abc:123@localhost:5432/api

I was still able to get into psql, even though the user and password are clearly not the one in docker compose.
I'm using Windows, mayb thats a Windows thingy

[HenryHengZJ]

thank you @fabiovalencio !

[rkeshwani]

This is returning a Error: Error: Credentials could not be decrypted on a windows machine running docker version of the database.
Note, I'm starting this based on the basic instructions without any configurations.
I updated my version of flowise and was able to see the new vectore store.

Console errors show:
Error: Malformed UTF-8 data

This is a different issue. Readding credentials for the language model solved this error.

[raymatos]

would it be possible to have flowise create the database and enable the vector extension when adding a database name that doesnt exist