Update chat docs & resource limits

docs/install_setup/chat.md

@@ -52,6 +52,9 @@ If the model is not present in the local cache, it will be downloaded when Gramp
 
 Please share learnings about different models with the community!
 
+!!! info
+    The sentence transformers library consumes a significant amount of memory, which might cause worker processes being killed. As a rule of thumb, with semantic search enabled, each Gunicorn worker consumes around 200 MB of memory and each celery worker around 500 MB of memory even when idle, and up to 1 GB when computing embeddings. See [Limit CPU and memory usage](cpu-limited.md) for settings that limit memory usage. In addition, it is advisable to provision a sufficiently large swap partition to prevent OOM errors due to transient memory usage spikes.
+
 ## Setting up an LLM provider
 
 Communication with the LLM uses an OpenAI compatible API using the `openai-python` library. This allows using a locally deployed LLM via Ollama (see [Ollama OpenAI compatibility](https://ollama.com/blog/openai-compatibility)) or an API like OpenAI or Huggingface TGI. The LLM is configured via the configuration parameters `LLM_MODEL` and `LLM_BASE_URL`.

docs/install_setup/cpu-limited.md

@@ -1,27 +1,58 @@
-# Limit CPU usage
+# Limit CPU and memory usage
 
-In order to avoid high CPU/RAM usage, it is possible to set the number of workers
-using the environment variable `GUNICORN_NUM_WORKERS`.
+In the recommended docker-based setup, Gramps Web uses [Gunicorn](https://gunicorn.org/) to serve the
+backend and [Celery](https://docs.celeryq.dev) for background tasks. In both cases, several worker
+processes can be run in parallel, which makes the application more responsive from a user perspective.
+However, increasing the number of workers also increase the amount of RAM used (even when the application is idle)
+and allowing requests to be processed in parallel can lead to high CPU usage (in particular when many users
+are using the application simultaneously). Both Gunicorn and Celery allow to limit the number of parallel workers.
 
-Here, we will take a number of workers = 2. Adjust it according to your needs.
-It may be a good idea to check the CPU/Threads available before choosing the value:
+## Get information about your system
 
-> lscpu | grep CPU
+On Linux, you can check the number of cores available on your system with the following command:
 
-The easiest way is to declare the variable in the `docker-compose.yml` file,
-under the "environment".
+```bash
+lscpu | grep CPU
+```
 
+To see how much memory and swap space you have available, use
+
+```bash
+free -h
 ```
-version: "3.7"
+
+
+## Limiting the number of Gunicorn workers
+
+The easiest way to set the number of Gunicorn workers when using the default Gramps Web
+docker image is to set the environment variable `GUNICORN_NUM_WORKERS`, e.g. by declaring it
+in the `docker-compose.yml` file,
+under the "environment".
+
+```yaml
 services:
   grampsweb:
     environment:
       GUNICORN_NUM_WORKERS: 2
 ```
 
-Other ways are possible, for example by storing the variable in a file, 
-and calling it in the startup command:
+See [the Gunicorn documentation](https://docs.gunicorn.org/en/stable/design.html#how-many-workers) to decide
+about the ideal number of workers.
+
+
+
+## Limiting the number of Celery workers
+
+To set the number of Celery workers, adapt the `concurrency` setting in the Docker compose file:
+
+```yaml
+  grampsweb_celery:
+    command: celery -A gramps_webapi.celery worker --loglevel=INFO --concurrency=2
+```
 
-> docker compose --env-file ./env up
+See [the Celery documentation](https://docs.celeryq.dev/en/stable/userguide/workers.html#concurrency) to decide
+about the ideal number of workers.
 
-In this case, the `env` file would contain a single line: GUNICORN_NUM_WORKERS=2
+!!! info
+    If the `concurrency` flag is omitted (which was the case in the Gramps Web documentation until v2.5.0), it
+    defaults to the number of CPU cores available on the system, which might consume a substantial amount of memory.

docs/user-guide/chat.md

@@ -1,5 +1,9 @@
 # Using AI chat
 
+!!! info
+    AI chat requires Gramps Web API version 2.5.0 or higher and Gramps Web version 24.10.0 or higher.
+
+
 The chat view in Gramps Web (if available in your installation) gives access to an AI assistant that can answer questions about your family tree.
 
 !!! warning

examples/caprover-one-click-app.yml

@@ -49,7 +49,7 @@ services:
             notExposeAsWebApp: 'true'
             dockerfileLines:
                 - FROM ghcr.io/gramps-project/grampsweb:$$cap_version
-                - CMD exec celery -A gramps_webapi.celery worker --loglevel=INFO
+                - CMD exec celery -A gramps_webapi.celery worker --loglevel=INFO --concurrency=2
 volumes:
     $$cap_appname-users:
     $$cap_appname-index:

examples/digitalocean-1click/docker-compose.yml

@@ -36,7 +36,7 @@ services:
       VIRTUAL_HOST: ""
       LETSENCRYPT_HOST: ""
       LETSENCRYPT_EMAIL: ""
-    command: celery -A gramps_webapi.celery worker --loglevel=INFO
+    command: celery -A gramps_webapi.celery worker --loglevel=INFO --concurrency=2
 
   grampsweb_redis:
     image: docker.io/library/redis:7.2.4-alpine

examples/docker-compose-base/docker-compose.yml

@@ -27,7 +27,7 @@ services:
     container_name: grampsweb_celery
     depends_on:
       - grampsweb_redis
-    command: celery -A gramps_webapi.celery worker --loglevel=INFO
+    command: celery -A gramps_webapi.celery worker --loglevel=INFO --concurrency=2
 
   grampsweb_redis:
     image: docker.io/library/redis:7.2.4-alpine

examples/docker-compose-letsencrypt/docker-compose.yml

@@ -37,7 +37,7 @@ services:
       VIRTUAL_HOST: ""
       LETSENCRYPT_HOST: ""
       LETSENCRYPT_EMAIL: ""
-    command: celery -A gramps_webapi.celery worker --loglevel=INFO
+    command: celery -A gramps_webapi.celery worker --loglevel=INFO --concurrency=2
 
   grampsweb_redis:
     image: docker.io/library/redis:7.2.4-alpine

mkdocs.yml

@@ -18,7 +18,7 @@ nav:
           - Update:  install_setup/update.md
           - Using PostgreSQL:  install_setup/postgres.md
           - Hosting media on S3:  install_setup/s3.md
-          - Limit CPU usage:  install_setup/cpu-limited.md
+          - Limit CPU & memory usage:  install_setup/cpu-limited.md
           - 2.0 upgrade guide:  install_setup/v2.md
   - Administration:
       - Introduction:  administration/admin.md