@ -1,31 +1,26 @@
< p > < / p >
< figure class = "image image_resized" style = "width:63.68%;" > < figure class = "image image_resized" style = "width:63.68%;" >
< img style = "aspect-ratio:1363/1364;" src = "Introduction_image.png" width = "1363"
< img style = "aspect-ratio:1363/1364;" src = "Introduction_image.png" width = "1363"
height="1364">
height="1364">
< figcaption > An example chat with an LLM< / figcaption >
< figcaption > An example chat with an LLM< / figcaption >
< / figure > < / figure >
< p > < / p >
< p > The AI / LLM features within Trilium Notes are designed to allow you to< p > The AI / LLM features within Trilium Notes are designed to allow you to
interact with your Notes in a variety of ways, using as many of the major
interact with your Notes in a variety of ways, using as many of the major
providers as we can support. < / p >
providers as we can support. < / p >
< p > < / p >
< p > In addition to being able to send chats to LLM providers such as OpenAI,< p > In addition to being able to send chats to LLM providers such as OpenAI,
Anthropic, and Ollama - we also support agentic tool calling, and embeddings.< / p >
Anthropic, and Ollama - we also support agentic tool calling, and embeddings.< / p >
< p > < / p >
< p > The quickest way to get started is to navigate to the “AI/LLM” settings:< / p > < p > The quickest way to get started is to navigate to the “AI/LLM” settings:< / p >
< figure < figure
class="image image_resized" style="width:74.04%;">
class="image image_resized" style="width:74.04%;">
< img style = "aspect-ratio:1916/1906;" src = "5_Introduction_image.png" width = "1916"
< img style = "aspect-ratio:1916/1906;" src = "5_Introduction_image.png" width = "1916"
height="1906">
height="1906">
< / figure >
< / figure >
< p > < / p >
< p > Enable the feature:< / p >
< p > Enable the feature:< / p >
< figure class = "image image_resized" style = "width:82.82%;" >
< figure class = "image image_resized" style = "width:82.82%;" >
< img style = "aspect-ratio:1911/997;" src = "1_Introduction_image.png" width = "1911"
< img style = "aspect-ratio:1911/997;" src = "1_Introduction_image.png" width = "1911"
height="997">
height="997">
< / figure >
< / figure >
< p > < / p >
< h2 > Embeddings< / h2 > < h2 > Embeddings< / h2 >
< p > < strong > Embeddings< / strong > are important as it allows us to have an compact
< p > < strong > Embeddings< / strong > are important as it allows us to have an compact
AI “summary” (it's not human readable text) of each of your Notes, that
AI “summary” (it's not human readable text) of each of your Notes, that
we can then perform mathematical functions on (such as cosine similarity)
we can then perform mathematical functions on (such as cosine similarity)
@ -37,7 +32,7 @@ class="image image_resized" style="width:74.04%;">
< p > In the following example, we're going to use our self-hosted Ollama instance
< p > In the following example, we're going to use our self-hosted Ollama instance
to create the embeddings for our Notes. You can see additional documentation
to create the embeddings for our Notes. You can see additional documentation
about installing your own Ollama locally in < a class = "reference-link"
about installing your own Ollama locally in < a class = "reference-link"
href="#root/jdjRLhLV3TtI/LMAv4Uy3Wk6J/7EdTxPADv95W/ _help_vvUCN7FDkq7G">Installing Ollama< / a > .< / p >
href="#root/_help_vvUCN7FDkq7G">Installing Ollama< / a > .< / p >
< p > To see what embedding models Ollama has available, you can check out
< p > To see what embedding models Ollama has available, you can check out
< a
< a
href="https://ollama.com/search?c=embedding">this search< / a > on their website, and then < code > pull< / code > whichever one
href="https://ollama.com/search?c=embedding">this search< / a > on their website, and then < code > pull< / code > whichever one
@ -51,7 +46,6 @@ class="image image_resized" style="width:74.04%;">
< img style = "aspect-ratio:1912/1075;" src = "4_Introduction_image.png" width = "1912"
< img style = "aspect-ratio:1912/1075;" src = "4_Introduction_image.png" width = "1912"
height="1075">
height="1075">
< / figure >
< / figure >
< p > < / p >
< p > When selecting the dropdown for the “Embedding Model”, embedding models
< p > When selecting the dropdown for the “Embedding Model”, embedding models
should be at the top of the list, separated by regular chat models with
should be at the top of the list, separated by regular chat models with
a horizontal line, as seen below:< / p >
a horizontal line, as seen below:< / p >
@ -60,7 +54,6 @@ class="image image_resized" style="width:74.04%;">
< img style = "aspect-ratio:1232/959;" src = "8_Introduction_image.png" width = "1232"
< img style = "aspect-ratio:1232/959;" src = "8_Introduction_image.png" width = "1232"
height="959">
height="959">
< / figure >
< / figure >
< p > < / p >
< p > After selecting an embedding model, embeddings should automatically begin
< p > After selecting an embedding model, embeddings should automatically begin
to be generated by checking the embedding statistics at the top of the
to be generated by checking the embedding statistics at the top of the
“AI/LLM” settings panel:< / p >
“AI/LLM” settings panel:< / p >
@ -68,7 +61,6 @@ class="image image_resized" style="width:74.04%;">
< img style = "aspect-ratio:1333/499;" src = "7_Introduction_image.png" width = "1333"
< img style = "aspect-ratio:1333/499;" src = "7_Introduction_image.png" width = "1333"
height="499">
height="499">
< / figure >
< / figure >
< p > < / p >
< p > If you don't see any embeddings being created, you will want to scroll
< p > If you don't see any embeddings being created, you will want to scroll
to the bottom of the settings, and hit “Recreate All Embeddings”:< / p >
to the bottom of the settings, and hit “Recreate All Embeddings”:< / p >
< figure
< figure
@ -76,19 +68,15 @@ class="image image_resized" style="width:74.04%;">
< img style = "aspect-ratio:1337/1490;" src = "3_Introduction_image.png" width = "1337"
< img style = "aspect-ratio:1337/1490;" src = "3_Introduction_image.png" width = "1337"
height="1490">
height="1490">
< / figure >
< / figure >
< p > < / p >
< p > Creating the embeddings will take some time, and will be regenerated when
< p > Creating the embeddings will take some time, and will be regenerated when
a Note is created, updated, or deleted (removed).< / p >
a Note is created, updated, or deleted (removed).< / p >
< p > If for some reason you choose to change your embedding provider, or the
< p > If for some reason you choose to change your embedding provider, or the
model used, you'll need to recreate all embeddings.< / p >
model used, you'll need to recreate all embeddings.< / p >
< p > < / p >
< p > < / p >
< h2 > Tools< / h2 >
< h2 > Tools< / h2 >
< p > Tools are essentially functions that we provide to the various LLM providers,
< p > Tools are essentially functions that we provide to the various LLM providers,
and then LLMs can respond in a specific format that tells us what tool
and then LLMs can respond in a specific format that tells us what tool
function and parameters they would like to invoke. We then execute these
function and parameters they would like to invoke. We then execute these
tools, and provide it as additional context in the Chat conversation. < / p >
tools, and provide it as additional context in the Chat conversation. < / p >
< p > < / p >
< p > These are the tools that currently exist, and will certainly be updated
< p > These are the tools that currently exist, and will certainly be updated
to be more effectively (and even more to be added!):< / p >
to be more effectively (and even more to be added!):< / p >
< ul >
< ul >
@ -148,7 +136,6 @@ class="image image_resized" style="width:74.04%;">
< / ul >
< / ul >
< / li >
< / li >
< / ul >
< / ul >
< p > < / p >
< p > When Tools are executed within your Chat, you'll see output like the following:< / p >
< p > When Tools are executed within your Chat, you'll see output like the following:< / p >
< figure
< figure
class="image image_resized" style="width:66.88%;">
class="image image_resized" style="width:66.88%;">
@ -157,9 +144,7 @@ class="image image_resized" style="width:74.04%;">
< / figure >
< / figure >
< p > You don't need to tell the LLM to execute a certain tool, it should “smartly”
< p > You don't need to tell the LLM to execute a certain tool, it should “smartly”
call tools and automatically execute them as needed.< / p >
call tools and automatically execute them as needed.< / p >
< p > < / p >
< h2 > Overview< / h2 >
< h2 > Overview< / h2 >
< p > < / p >
< p > Now that you know about embeddings and tools, you can just go ahead and
< p > Now that you know about embeddings and tools, you can just go ahead and
use the “Chat with Notes” button, where you can go ahead and start chatting!:< / p >
use the “Chat with Notes” button, where you can go ahead and start chatting!:< / p >
< figure
< figure
@ -167,7 +152,6 @@ class="image image_resized" style="width:74.04%;">
< img style = "aspect-ratio:1378/539;" src = "2_Introduction_image.png" width = "1378"
< img style = "aspect-ratio:1378/539;" src = "2_Introduction_image.png" width = "1378"
height="539">
height="539">
< / figure >
< / figure >
< p > < / p >
< p > If you don't see the “Chat with Notes” button on your side launchbar,
< p > If you don't see the “Chat with Notes” button on your side launchbar,
you might need to move it from the “Available Launchers” section to the
you might need to move it from the “Available Launchers” section to the
“Visible Launchers” section:< / p >
“Visible Launchers” section:< / p >
Elian Doran