55 lines
No EOL
1.8 KiB
XML
55 lines
No EOL
1.8 KiB
XML
<?xml version="1.0" encoding="UTF-8"?>
|
||
<!--
|
||
imixs-llm.xml - LLM Model Registry
|
||
======================================
|
||
Place this file in your deployment so it is available on the classpath,
|
||
e.g. in src/main/resources/ for a Maven project.
|
||
|
||
Each <model> element defines one logical LLM service endpoint.
|
||
A model is either a completion model OR an embedding model - never both.
|
||
The BPMN configuration references them separately by their id:
|
||
|
||
<imixs-ai name="CONDITION">
|
||
<endpoint>my-llm</endpoint>
|
||
...
|
||
</imixs-ai>
|
||
|
||
<imixs-ai name="RAG_INDEX">
|
||
<endpoint-completion>my-llm</endpoint-completion>
|
||
<endpoint-embeddings>my-embeddings</endpoint-embeddings>
|
||
...
|
||
</imixs-ai>
|
||
|
||
Environment variable placeholders are supported via ${env.VAR_NAME}.
|
||
-->
|
||
<imixs-llm>
|
||
|
||
<!--
|
||
Completion model – used for chat completions, conditions, analysis etc.
|
||
This example connects to a local llama.cpp server.
|
||
-->
|
||
<endpoint id="api.llama.cpp.imixs.com">
|
||
<url>https://api.llama.cpp.imixs.com/</url>
|
||
<apikey>7fdb663fa87ea16731c1e0ece8ad3214e358423eabf2403188110c731ce07321</apikey>
|
||
<options>{
|
||
"temperature": 0.2,
|
||
"max_tokens": 1024
|
||
}
|
||
</options>
|
||
</endpoint>
|
||
|
||
<!--
|
||
Embedding model – used for RAG indexing and retrieval.
|
||
This example connects to a separate embedding server.
|
||
No apikey needed for a locally hosted instance.
|
||
-->
|
||
<endpoint id="embeddings.llama.cpp.imixs.com">
|
||
<url>https://embeddings.llama.cpp.imixs.com/</url>
|
||
<apikey>7fdb663fa87ea16731c1e0ece8ad3214e358423eabf2403188110c731ce07321</apikey>
|
||
<options>{
|
||
"max_tokens": 512
|
||
}
|
||
</options>
|
||
</endpoint>
|
||
|
||
</imixs-llm> |