Added llamafile template support

Makefile

@@ -32,12 +32,12 @@ clean:
 set-version:
 	echo '"${VERSION}"' > templates/values/version.jsonnet
 
-TEMPLATES=azure bedrock claude cohere mix ollama openai vertexai \
+TEMPLATES=azure bedrock claude cohere mix llamafile ollama openai vertexai \
     openai-neo4j storage
 
 DCS=$(foreach template,${TEMPLATES},${template:%=tg-launch-%.yaml})
 
-MODELS=azure bedrock claude cohere ollama openai vertexai
+MODELS=azure bedrock claude cohere llamafile ollama openai vertexai
 GRAPHS=cassandra neo4j
 
 # tg-launch-%.yaml: templates/%.jsonnet templates/components/version.jsonnet

templates/all-patterns.jsonnet

@@ -10,6 +10,7 @@
     import "patterns/llm-bedrock.jsonnet",
     import "patterns/llm-claude.jsonnet",
     import "patterns/llm-cohere.jsonnet",
+    import "patterns/llm-llamafile.jsonnet",
     import "patterns/llm-ollama.jsonnet",
     import "patterns/llm-openai.jsonnet",
     import "patterns/llm-vertexai.jsonnet",

templates/components.jsonnet

@@ -10,6 +10,7 @@
    "graph-rag": import "components/graph-rag.jsonnet",
    "triple-store-cassandra": import "components/cassandra.jsonnet",
    "triple-store-neo4j": import "components/neo4j.jsonnet",
+   "llamafile": import "components/llamafile.jsonnet",
    "ollama": import "components/ollama.jsonnet",
    "openai": import "components/openai.jsonnet",
    "override-recursive-chunker": import "components/chunker-recursive.jsonnet",

templates/components/llamafile.jsonnet

@@ -0,0 +1,75 @@
+local base = import "base/base.jsonnet";
+local images = import "values/images.jsonnet";
+local url = import "values/url.jsonnet";
+local prompts = import "prompts/slm.jsonnet";
+
+{
+
+    "llamafile-model":: "LLaMA_CPP",
+    "llamafile-url":: "${LLAMAFILE_URL}",
+
+    "text-completion" +: {
+    
+        create:: function(engine)
+
+            local container =
+                engine.container("text-completion")
+                    .with_image(images.trustgraph)
+                    .with_command([
+                        "text-completion-llamafile",
+                        "-p",
+                        url.pulsar,
+                        "-m",
+                        $["llamafile-model"],
+                        "-r",
+                        $["llamafile-url"],
+                    ])
+                    .with_limits("0.5", "128M")
+                    .with_reservations("0.1", "128M");
+
+            local containerSet = engine.containers(
+                "text-completion", [ container ]
+            );
+
+            engine.resources([
+                containerSet,
+            ])
+
+    },
+
+    "text-completion-rag" +: {
+    
+        create:: function(engine)
+
+            local container =
+                engine.container("text-completion-rag")
+                    .with_image(images.trustgraph)
+                    .with_command([
+                        "text-completion-llamafile",
+                        "-p",
+                        url.pulsar,
+                        "-m",
+                        $["llamafile-model"],
+                        "-r",
+                        $["llamafile-url"],
+                        "-i",
+                        "non-persistent://tg/request/text-completion-rag",
+                        "-o",
+                        "non-persistent://tg/response/text-completion-rag-response",
+                    ])
+                    .with_limits("0.5", "128M")
+                    .with_reservations("0.1", "128M");
+
+            local containerSet = engine.containers(
+                "text-completion-rag", [ container ]
+            );
+
+            engine.resources([
+                containerSet,
+            ])
+
+
+    }
+
+} + prompts
+

templates/patterns/llm-llamafile.jsonnet

@@ -0,0 +1,41 @@
+{
+    pattern: {
+	name: "llamafile",
+        icon: "🤖💬",
+        title: "Add Llamafile-invoked LLMs for text completion",
+	description: "This pattern integrates a Llamafile service for text completion operations.  You need to have a running Llamafile implementation executing the necessary model in order to be able to use this service.",
+        requires: ["pulsar", "trustgraph"],
+        features: ["llm"],
+	args: [
+	    {
+		name: "llamafile-max-output-tokens",
+                label: "Maximum output tokens",
+		type: "integer",
+		description: "Limit on number tokens to generate",
+                default: 4096,
+		required: true,
+            },
+	    {
+		name: "llamafile-temperature",
+                label: "Temperature",
+		type: "slider",
+		description: "Controlling predictability / creativity balance",
+                min: 0,
+                max: 1,
+                step: 0.05,
+                default: 0.5,
+            },
+	    {
+		name: "llamafile-url",
+                label: "URL",
+		type: "text",
+		width: 120,
+		description: "URL of the Llamafile service",
+                default: "http://llamafile:8080",
+		required: true,
+	    },
+	],
+        category: [ "llm" ],
+    },
+    module: "components/llamafile.jsonnet",
+}