How to make AI with self-hosted deployment of v0.20 work

[hnykda]

I wasn't sure if I can make this work, but voila, it works. I am grateful for Affine and the team behind it, thanks a lot!

What I needed to do was to add this to the affine.ts, set the COPILOT_OPENAI_API_KEY env variable to my OpenAI API Key, and voila, it just works:

/* Copilot Plugin */
AFFiNE.use('copilot', {
  openai: {
    apiKey: process.env.COPILOT_OPENAI_API_KEY
  },
})

in my specific case, I have a docker-compose deployment via coolify, so with this config:

services:
  affine:
    image: 'ghcr.io/toeverything/affine-graphql:stable'
    command:
      - sh
      - '-c'
      - 'node ./scripts/self-host-predeploy && node ./dist/index.js'
    depends_on:
      redis:
        condition: service_healthy
      postgres:
        condition: service_healthy
    volumes:
      - './data/config:/root/.affine/config'
      - './data/storage:/root/.affine/storage'
    logging:
      driver: json-file
      options:
        max-size: 1000m
    environment:
      - SERVICE_FQDN_AFFINE_3010
      - NODE_OPTIONS=--import=./scripts/register.js
      - AFFINE_CONFIG_PATH=/root/.affine/config
      - REDIS_SERVER_HOST=redis
      - 'DATABASE_URL=postgres://${SERVICE_USER_POSTGRES}:${SERVICE_PASSWORD_POSTGRES}@postgres:5432/${POSTGRES_DB:-affine}'
      - NODE_ENV=production
      - AFFINE_SERVER_HOST=$SERVICE_FQDN_AFFINE
      - AFFINE_SERVER_EXTERNAL_URL=$SERVICE_FQDN_AFFINE
      - 'MAILER_HOST=${MAILER_HOST}'
      - 'MAILER_PORT=${MAILER_PORT}'
      - 'MAILER_USER=${MAILER_USER}'
      - 'MAILER_PASSWORD=${MAILER_PASSWORD}'
      - 'MAILER_SENDER=${MAILER_SENDER}'
      - COPILOT_OPENAI_API_KEY=${COPILOT_OPENAI_API_KEY}
    healthcheck:
      test:
        - CMD-SHELL
        - "bash -c ':> /dev/tcp/127.0.0.1/3010' || exit 1"
      interval: 5s
      timeout: 20s
      retries: 3
  redis:
    image: redis
    volumes:
      - 'affine-redis-data:/data'
    healthcheck:
      test:
        - CMD
        - redis-cli
        - '--raw'
        - incr
        - ping
      interval: 10s
      timeout: 5s
      retries: 5
  postgres:
    image: 'postgres:16'
    volumes:
      - 'affine-postgres-data:/var/lib/postgresql/data'
    healthcheck:
      test:
        - CMD-SHELL
        - 'pg_isready -U $$POSTGRES_USER -d $$POSTGRES_DB'
      interval: 10s
      timeout: 5s
      retries: 5
    environment:
      - 'POSTGRES_USER=${SERVICE_USER_POSTGRES}'
      - 'POSTGRES_PASSWORD=${SERVICE_PASSWORD_POSTGRES}'
      - 'POSTGRES_DB=${POSTGRES_DB:-affine}'
      - PGDATA=/var/lib/postgresql/data/pgdata

and then the file where I added the above is under ./data/config/affine.js

[xtiang7c0]

Hey there, thanks for the description. I added it to the file and saved + restarted.

Still getting this error:

An error occurred, If this issue persists please let us know.
Show detail
INTERNAL_SERVER_ERROR: Internal server error Identify: INTERNAL_SERVER_ERROR

Any idea?

I added the OPEN AI API KEY to the env vars.

[hnykda]

No idea, sorry. Try to dig deeper into server logs to see what's wrong?

[sync-adm]

My affine.ts is located at src/config/affine.ts is it right? I just nano it, added the code and did not restart (restart would undo the change).

But I'm seeing the same error. I'm also hosting it on coolify.

[hnykda]

That IMHO won't work. Once it gets interpreted by the process (which is after starting the container), it won't change (this is a generic behavior of e.g. Python programs etc. as well).

[mcfedr]

AI was working great, but just updated to 0.21 and its disappeared, anyone know how to get it back?

[richie80]

They changed to schema.json with the .21 version.

[penleychan]

They changed to schema.json with the .21 version.

Where is this located? Is this something that we can change?

[richie80]

Yes its in the release notes. But I havent figured out yet how to do this on docker

[vladkor97]

@hnykda hey, have you figured out how to make it work on the latest ver?

[hnykda]

hey, nope, was scare of updating after you reporte dthe above guys

[penleychan]

Alright had sometime to dig around the code today, here's what working for me based on how all thing is wired up on self hosted.
Few things to get it working.
You need to create config.json in your .affine/config/ this is the same location as affine.js and you configure your user to allow UnlimitedCopilot

My AFFiNE app version: App Version: AFFiNE 0.21.2 in the /admin/about I believe 0.21.2 is required as there has been some changes on this canary version see PR: #11584

Your json should look like this:

{

  "server": {
    "name": "YOUR AFFINE SERVER NAME"
  },
  "copilot": {
    "enabled": true,
    "providers.openai": {
       "apiKey": "sk-proj-****"
    }
  }
}

In your admin portal for accounts, Edit the user make sure to save the changes on the check mark above

Restart/rebuild your docker container.

Doesn't seem to be a way to change the model, currently it's using gpt-4o-2024-08-06 based on my OpenAI API usage.

[richie80]

This worked like a charm! thanks! (also very easy)

[Heidieker]

Thank you, works great.
I dont know my "YOUR AFFINE SERVER NAME". Is it the URL?
But it is working just with "Affine"

[penleychan]

Thank you, works great. I dont know my "YOUR AFFINE SERVER NAME". Is it the URL? But it is working just with "Affine"

That can be anything, it's not necessary to include it by default I think it's set to "AFFiNE cloud" or something like that, which you can see in the admin page somewhere.

[PrestonCS]

Looks like this stopped working in v22 😞