Add AI integration documentation and update release notes

Introduced initial Large Language Model (LLM) support with details on Ollama and OpenAI engines. Added related configuration options, installation instructions, and usage documentation. Published release notes for version 4.0.0 highlighting AI integration and other updates.

Author: gjong

_config.yml

@@ -1,8 +1,8 @@
 title: Pledger.io
 
 release:
-  version: 3.8.0
-  date: 2025-01-03
+  version: 4.0.0
+  date: 2025-02-20
 
 github:
   repository: pledger-io/user-documentation

_release-notes/index.adoc

@@ -1,4 +1,5 @@
 
+include::v4.x/index.adoc[]
 include::v3.x/index.adoc[]
 include::v2.x/index.adoc[]
 include::v1.x/index.adoc[]

_release-notes/v4.x/index.adoc

@@ -0,0 +1,2 @@
+
+include::v4.0.0.adoc[]

_release-notes/v4.x/v4.0.0.adoc

@@ -0,0 +1,14 @@
+
+== 4.0.0
+
+*Release date: 2025-02-20*
+
+=== Added
+
+- Feature/initial AI integration
+
+=== Changed
+
+- Feature/move to vite for the UX build engine
+- Refactor form input handling and enhance transaction suggestions
+- Move the eventbus to only the command system

docs/getting-started/advanced/index.adoc

@@ -5,7 +5,7 @@
 :page-section: getting-started
 :page-layout: section
 :page-order: 2
-:page-status: draft
+:page-status: published
 
 ifndef::document-root[:document-root: ../../]
 include::../../_document-setup.adoc[]

docs/getting-started/advanced/using-an-llm.adoc

@@ -0,0 +1,65 @@
+:page-nav-title: Using the AI features
+:page-parent: getting-started::advanced
+:page-title: Using the AI features in Pledger.io
+:page-section: getting-started
+:page-layout: section
+:page-order: 1
+:page-status: published
+
+ifndef::document-root[:document-root: ../../]
+include::../../_document-setup.adoc[]
+
+WARNING: This functionality is still under active development and should be considered a `preview` feature.
+Some of the configuration or usage may be a bit more difficult.
+
+Since version `4.0.0` of {application} support for Large Language Models (LLM) has been added.
+A Large Language Model (LLM) is an advanced machine learning model trained on vast amounts of text data to understand, generate, and process human-like natural language.
+
+{application} supports the following types of LLM engines:
+
+1. link:https://platform.openai.com[Open AI], a cloud based LLM
+2. link:https://ollama.com[Ollama], a local hosted LLM
+
+== How will the AI help you
+
+{application} currently only has very limited features build using the LLM.
+The following is added at this time:
+
+* Making suggestions for budget, category and tags based upon the description, bank accounts and amount in the transaction form.
+
+== Privacy focussed option
+
+For those users that want their financial data to remain private you can use the Ollama version.
+A special version of the Docker Image is released that contains the Ollama software alongside {application}.
+
+NOTE: Use this docker image `ghcr.io/pledger-io/amd64-embedded-llm`.
+
+When starting the docker image it will automatically download the model configured in the environment.
+Please see the page `xref:{document-root}/how-to/installation/configuration.adoc#configure_llm[Large Language Model options]` for how to configure {application} correctly.
+
+For most users it is advised to not change the model used by Ollama, the picked model has the best accuracy vs performance.
+If you have an Nvidia or AMD GPU you may be able to utilize that to power Ollama.
+This would allow you to pick a larger model like `llama3.3` or `mistral`.
+
+To utilize the GPU you would have to adopt the docker command to something like this:
+
+[source,shell,linenums]
+....
+ docker run -d \
+         -v pledger_logs:/opt/storage/logs \
+         -v pledger_uploads:/opt/storage/upload \
+         -p 8080:8080 \
+         --gpus=all \
+         --name pledger \
+         -e DB_TYPE=mysql
+         -e DATABASE_HOST=mariadb
+         -e DATABASE_SCHEMA=pledger
+         -e DATABASE_USER=pledger
+         -e DATABASE_PASSWORD=pledger
+         -e AI_MODEL=llama3.3
+         ghcr.io/pledger-io/amd64-embedded-llm:stable
+....
+
+== OpenAI based AI
+
+TIP: Still to be documented....

docs/how-to/installation/configuration.adoc

@@ -51,4 +51,24 @@ When using the mail feature of {application} the following environment variables
 | `SMTP_USER` | The SMTP user for authentication | -
 | `SMTP_PASSWORD` | The SMTP password for authentication | -
 
-|===
+|===
+
+[[configure_llm]]
+== Large Language Model options
+
+When either using the docker image `ghcr.io/pledger-io/amd64-embedded-llm` you can configure the following options.
+Alternatively you can use the normal image but use link:https://platform.openai.com[OpenAI], at the cost of some privacy.
+
+[%header,cols="1,2,1"]
+|===
+| Variable | Description | Default
+
+| `AI_ENGINE` | The engine to use. Supported are `ollama` and `open-ai` | ollama
+| `AI_MODEL` | The Ollama model to use, only used with AI_ENGINE 'ollama' | qwen2.5-coder:1.5b
+| `AI_AUGMENTER_ENABLED` | Should the context for the LLM be expanded with budget, category and tag information | true
+| `OPENAI_TOKEN` | The application token for OpenAI, must be set when AI_ENGINE is 'open-ai' | -
+
+|===
+
+NOTE: Please leave `AI_AUGMENTER_ENABLED` enabled with the default model.
+This can be disabled when using larger models as they have a better understanding of the transaction information and tools.

docs/how-to/installation/docker.adoc

@@ -25,6 +25,7 @@ There are various builds for various different target platforms.
 Personal Ledger is currently built for:
 
 - AMD64, under image `{docker-image}`
+- AMD64 with Ollama embedded, under image `ghcr.io/pledger-io/amd64-embedded-llm`
 - ARMv8, under image `ghcr.io/pledger-io/arm64-core`
 
 The instructions here will assume usage of the tag `stable`.