Kevin Burnett 5df1262dca
Proofing updates (#1838)
* update in place with python

* split files into chunks

* working chunking and updated quick start

* edits

* sanity check

* give up on faq page, long docs work

* debug

* system prompt updates, etc

* use temp file for output

* refactor

* remove dup import

* generate diff file

* check diff output to make sure it looks reasonable

* add overall results

* update script

* update script

* update script

* edits

* fix function

---------

Co-authored-by: burnettk <burnettk@users.noreply.github.com>
2024-06-28 08:03:25 -07:00

29 lines
900 B
Markdown

# Deployment
The minimal deployment is to mimic the docker-compose.yml file at the root of spiff-arena.
Steps for a more hardened production setup after that baseline include:
1. Setting up a MySQL or PostgreSQL database for Backend persistence (instead of SQLite)
2. Setting up a Redis/Valkey or RabbitMQ server for a Celery broker.
3. Separating out the Backend deployment into three deployments: 1) API, 2) Background, and 3) Celery worker.
```mermaid
graph TD;
subgraph Backend
A[API]
B[Background]
Ce[Celery Worker]
end
F[Frontend]
Co[Connector Proxy]
D[Database]
A --> D
B --> D
Ce --> D
F -- Communicates with --> Backend
Backend -- delegates to --> Co
```
API, Celery Worker, Connector Proxy, and Frontend can run any number of replicas.
The Background container is like a cron container, so it should run only one replica.