Partial implementation of dual-mode templates

templates/components.jsonnet

@@ -1,47 +1,70 @@
 {
+
+   // Essentials
+   "trustgraph-base": import "components/trustgraph.jsonnet",
+   "pulsar": import "components/pulsar.jsonnet",
+
+   // LLMs
    "azure": import "components/azure.jsonnet",
    "azure-openai": import "components/azure-openai.jsonnet",
    "bedrock": import "components/bedrock.jsonnet",
    "claude": import "components/claude.jsonnet",
    "cohere": import "components/cohere.jsonnet",
-   "document-rag": import "components/document-rag.jsonnet",
-   "embeddings-hf": import "components/embeddings-hf.jsonnet",
-   "embeddings-ollama": import "components/embeddings-ollama.jsonnet",
    "googleaistudio": import "components/googleaistudio.jsonnet",
-   "grafana": import "components/grafana.jsonnet",
+   "ollama": import "components/ollama.jsonnet",
+   "openai": import "components/openai.jsonnet",
+   "vertexai": import "components/vertexai.jsonnet",
+
+   // LLMs for RAG
+   "azure-rag": import "components/azure-rag.jsonnet",
+   "azure-openai-rag": import "components/azure-openai-rag.jsonnet",
+   "bedrock-rag": import "components/bedrock-rag.jsonnet",
+   "claude-rag": import "components/claude-rag.jsonnet",
+   "cohere-rag": import "components/cohere-rag.jsonnet",
+   "googleaistudio-rag": import "components/googleaistudio-rag.jsonnet",
+   "ollama-rag": import "components/ollama-rag.jsonnet",
+   "openai-rag": import "components/openai-rag.jsonnet",
+   "vertexai-rag": import "components/vertexai-rag.jsonnet",
+
+   // Embeddings
+   "embeddings-ollama": import "components/embeddings-ollama.jsonnet",
+   "embeddings-hf": import "components/embeddings-hf.jsonnet",
+
+   // Processing pipelines
    "graph-rag": import "components/graph-rag.jsonnet",
+   "document-rag": import "components/document-rag.jsonnet",
+
+   // Vector stores
+   "vector-store-milvus": import "components/milvus.jsonnet",
+   "vector-store-qdrant": import "components/qdrant.jsonnet",
+   "vector-store-pinecone": import "components/pinecone.jsonnet",
+
+   // Triples stores
    "triple-store-cassandra": import "components/cassandra.jsonnet",
    "triple-store-neo4j": import "components/neo4j.jsonnet",
    "triple-store-falkordb": import "components/falkordb.jsonnet",
    "triple-store-memgraph": import "components/memgraph.jsonnet",
+
+   // Observability support
+   "grafana": import "components/grafana.jsonnet",
+
+   // Pulsar manager is a UI for Pulsar.  Uses a LOT of memory
+   "pulsar-manager": import "components/pulsar-manager.jsonnet",
+
    "llamafile": import "components/llamafile.jsonnet",
-   "ollama": import "components/ollama.jsonnet",
-   "openai": import "components/openai.jsonnet",
    "override-recursive-chunker": import "components/chunker-recursive.jsonnet",
 
+   // The prompt manager
    "prompt-template": import "components/prompt-template.jsonnet",
    "prompt-overrides": import "components/prompt-overrides.jsonnet",
 
-   "pulsar": import "components/pulsar.jsonnet",
+   // ReAct agent
-   "pulsar-manager": import "components/pulsar-manager.jsonnet",
-   "trustgraph-base": import "components/trustgraph.jsonnet",
-   "vector-store-milvus": import "components/milvus.jsonnet",
-   "vector-store-qdrant": import "components/qdrant.jsonnet",
-   "vector-store-pinecone": import "components/pinecone.jsonnet",
-   "vertexai": import "components/vertexai.jsonnet",
-   "workbench-ui": import "components/workbench-ui.jsonnet",
-   "null": {},
-
    "agent-manager-react": import "components/agent-manager-react.jsonnet",
 
-   // FIXME: Dupes
+   // Optional UI
-   "cassandra": import "components/cassandra.jsonnet",
+   "workbench-ui": import "components/workbench-ui.jsonnet",
-   "neo4j": import "components/neo4j.jsonnet",
+
-   "memgraph": import "components/memgraph.jsonnet",
+   // Does nothing.  But, can be a hack to overwrite parameters
-   "qdrant": import "components/qdrant.jsonnet",
+   "null": {},
-   "pinecone": import "components/pinecone.jsonnet",
-   "milvus": import "components/milvus.jsonnet",
-   "falkordb": import "components/falkordb.jsonnet",
-   "trustgraph": import "components/trustgraph.jsonnet",
 
 }

templates/components/azure-openai-rag.jsonnet

@@ -0,0 +1,61 @@
+local base = import "base/base.jsonnet";
+local images = import "values/images.jsonnet";
+local url = import "values/url.jsonnet";
+local prompts = import "prompts/mixtral.jsonnet";
+
+{
+
+    with:: function(key, value)
+        self + {
+            ["ollama-rag-" + key]:: value,
+        },
+
+    "azure-openai-rag-model":: "GPT-3.5-Turbo",
+    "azure-openai-rag-max-output-tokens":: 4192,
+    "azure-openai-rag-temperature":: 0.0,
+
+    "text-completion" +: {
+    
+        create:: function(engine)
+
+            local envSecrets = engine.envSecrets("azure-openai-credentials")
+                .with_env_var("AZURE_TOKEN", "azure-token");
+
+            local containerRag =
+                engine.container("text-completion-rag")
+                    .with_image(images.trustgraph)
+                    .with_command([
+                        "text-completion-azure",
+                        "-p",
+                        url.pulsar,
+                        "-x",
+                        std.toString($["azure-openai-rag-max-output-tokens"]),
+                        "-t",
+                        "%0.3f" % $["azure-openai-rag-temperature"],
+                        "-i",
+                        "non-persistent://tg/request/text-completion-rag",
+                        "-o",
+                        "non-persistent://tg/response/text-completion-rag",
+                    ])
+                    .with_env_var_secrets(envSecrets)
+                    .with_limits("0.5", "128M")
+                    .with_reservations("0.1", "128M");
+
+            local containerSetRag = engine.containers(
+                "text-completion-rag", [ containerRag ]
+            );
+
+            local serviceRag =
+                engine.internalService(containerSetRag)
+                .with_port(8000, 8000, "metrics");
+
+            engine.resources([
+                envSecrets,
+                containerSetRag,
+                serviceRag,
+            ])
+
+    },
+
+} + prompts
+

templates/components/azure-openai.jsonnet

@@ -5,6 +5,11 @@ local prompts = import "prompts/mixtral.jsonnet";
 
 {
 
+    with:: function(key, value)
+        self + {
+            ["azure-openai-" + key]:: value,
+        },
+
     "azure-openai-model":: "GPT-3.5-Turbo",
     "azure-openai-max-output-tokens":: 4192,
     "azure-openai-temperature":: 0.0,
@@ -34,48 +39,18 @@ local prompts = import "prompts/mixtral.jsonnet";
                     .with_limits("0.5", "128M")
                     .with_reservations("0.1", "128M");
 
-            local containerRag =
-                engine.container("text-completion-rag")
-                    .with_image(images.trustgraph)
-                    .with_command([
-                        "text-completion-azure",
-                        "-p",
-                        url.pulsar,
-                        "-x",
-                        std.toString($["azure-openai-max-output-tokens"]),
-                        "-t",
-                        "%0.3f" % $["azure-openai-temperature"],
-                        "-i",
-                        "non-persistent://tg/request/text-completion-rag",
-                        "-o",
-                        "non-persistent://tg/response/text-completion-rag",
-                    ])
-                    .with_env_var_secrets(envSecrets)
-                    .with_limits("0.5", "128M")
-                    .with_reservations("0.1", "128M");
-
             local containerSet = engine.containers(
                 "text-completion", [ container ]
             );
 
-            local containerSetRag = engine.containers(
-                "text-completion-rag", [ containerRag ]
-            );
-
             local service =
                 engine.internalService(containerSet)
                 .with_port(8000, 8000, "metrics");
 
-            local serviceRag =
-                engine.internalService(containerSetRag)
-                .with_port(8000, 8000, "metrics");
-
             engine.resources([
                 envSecrets,
                 containerSet,
-                containerSetRag,
                 service,
-                serviceRag,
             ])
 
     },

templates/components/azure-rag.jsonnet

@@ -0,0 +1,60 @@
+local images = import "values/images.jsonnet";
+local url = import "values/url.jsonnet";
+local prompts = import "prompts/mixtral.jsonnet";
+
+{
+
+    with:: function(key, value)
+        self + {
+            ["azure-rag-" + key]:: value,
+        },
+
+    "azure-rag-max-output-tokens":: 4096,
+    "azure-rag-temperature":: 0.0,
+
+    "text-completion-rag" +: {
+    
+        create:: function(engine)
+
+            local envSecrets = engine.envSecrets("azure-credentials")
+                .with_env_var("AZURE_TOKEN", "azure-token")
+                .with_env_var("AZURE_ENDPOINT", "azure-endpoint");
+
+            local containerRag =
+                engine.container("text-completion-rag")
+                    .with_image(images.trustgraph)
+                    .with_command([
+                        "text-completion-azure",
+                        "-p",
+                        url.pulsar,
+                        "-x",
+                        std.toString($["azure-rag-max-output-tokens"]),
+                        "-t",
+                        "%0.3f" % $["azure-rag-temperature"],
+                        "-i",
+                        "non-persistent://tg/request/text-completion-rag",
+                        "-o",
+                        "non-persistent://tg/response/text-completion-rag",
+                    ])
+                    .with_env_var_secrets(envSecrets)
+                    .with_limits("0.5", "128M")
+                    .with_reservations("0.1", "128M");
+
+            local containerSetRag = engine.containers(
+                "text-completion-rag", [ containerRag ]
+            );
+
+            local serviceRag =
+                engine.internalService(containerSetRag)
+                .with_port(8000, 8000, "metrics");
+
+            engine.resources([
+                envSecrets,
+                containerSetRag,
+                serviceRag,
+            ])
+
+    }
+
+} + prompts
+

templates/components/azure.jsonnet

@@ -1,10 +1,14 @@
-local base = import "base/base.jsonnet";
 local images = import "values/images.jsonnet";
 local url = import "values/url.jsonnet";
 local prompts = import "prompts/mixtral.jsonnet";
 
 {
 
+    with:: function(key, value)
+        self + {
+            ["azure-" + key]:: value,
+        },
+
     "azure-max-output-tokens":: 4096,
     "azure-temperature":: 0.0,
 
@@ -32,48 +36,18 @@ local prompts = import "prompts/mixtral.jsonnet";
                     .with_limits("0.5", "128M")
                     .with_reservations("0.1", "128M");
 
-            local containerRag =
-                engine.container("text-completion-rag")
-                    .with_image(images.trustgraph)
-                    .with_command([
-                        "text-completion-azure",
-                        "-p",
-                        url.pulsar,
-                        "-x",
-                        std.toString($["azure-max-output-tokens"]),
-                        "-t",
-                        "%0.3f" % $["azure-temperature"],
-                        "-i",
-                        "non-persistent://tg/request/text-completion-rag",
-                        "-o",
-                        "non-persistent://tg/response/text-completion-rag",
-                    ])
-                    .with_env_var_secrets(envSecrets)
-                    .with_limits("0.5", "128M")
-                    .with_reservations("0.1", "128M");
-
             local containerSet = engine.containers(
                 "text-completion", [ container ]
             );
 
-            local containerSetRag = engine.containers(
-                "text-completion-rag", [ containerRag ]
-            );
-
             local service =
                 engine.internalService(containerSet)
                 .with_port(8000, 8000, "metrics");
 
-            local serviceRag =
-                engine.internalService(containerSetRag)
-                .with_port(8000, 8000, "metrics");
-
             engine.resources([
                 envSecrets,
                 containerSet,
-                containerSetRag,
                 service,
-                serviceRag,
             ])
 
     }

templates/components/bedrock-rag.jsonnet

@@ -0,0 +1,66 @@
+local base = import "base/base.jsonnet";
+local images = import "values/images.jsonnet";
+local url = import "values/url.jsonnet";
+local prompts = import "prompts/mixtral.jsonnet";
+local chunker = import "chunker-recursive.jsonnet";
+
+{
+
+    with:: function(key, value)
+        self + {
+            ["bedrock-rag-" + key]:: value,
+        },
+
+    "bedrock-rag-max-output-tokens":: 4096,
+    "bedrock-rag-temperature":: 0.0,
+    "bedrock-rag-model":: "mistral.mixtral-8x7b-instruct-v0:1",
+
+    "text-completion-rag" +: {
+    
+        create:: function(engine)
+
+            local envSecrets = engine.envSecrets("bedrock-credentials")
+                .with_env_var("AWS_ID_KEY", "aws-id-key")
+                .with_env_var("AWS_SECRET", "aws-secret")
+                .with_env_var("AWS_REGION", "aws-region");
+
+            local containerRag =
+                engine.container("text-completion-rag")
+                    .with_image(images.trustgraph)
+                    .with_command([
+                        "text-completion-bedrock",
+                        "-p",
+                        url.pulsar,
+                        "-x",
+                        std.toString($["bedrock-rag-max-output-tokens"]),
+                        "-t",
+                        "%0.3f" % $["bedrock-rag-temperature"],
+                        "-m",
+                        $["bedrock-rag-model"],
+                        "-i",
+                        "non-persistent://tg/request/text-completion-rag",
+                        "-o",
+                        "non-persistent://tg/response/text-completion-rag",
+              	    ])
+                    .with_env_var_secrets(envSecrets)
+                    .with_limits("0.5", "128M")
+                    .with_reservations("0.1", "128M");
+
+            local containerSetRag = engine.containers(
+                "text-completion-rag", [ containerRag ]
+            );
+
+            local serviceRag =
+                engine.internalService(containerSetRag)
+                .with_port(8000, 8000, "metrics");
+
+            engine.resources([
+                envSecrets,
+                containerSetRag,
+                serviceRag,
+            ])
+
+    },
+
+} + prompts + chunker
+

templates/components/bedrock.jsonnet

@@ -6,6 +6,11 @@ local chunker = import "chunker-recursive.jsonnet";
 
 {
 
+    with:: function(key, value)
+        self + {
+            ["bedrock-" + key]:: value,
+        },
+
     "bedrock-max-output-tokens":: 4096,
     "bedrock-temperature":: 0.0,
     "bedrock-model":: "mistral.mixtral-8x7b-instruct-v0:1",
@@ -37,50 +42,18 @@ local chunker = import "chunker-recursive.jsonnet";
                     .with_limits("0.5", "128M")
                     .with_reservations("0.1", "128M");
 
-            local containerRag =
-                engine.container("text-completion-rag")
-                    .with_image(images.trustgraph)
-                    .with_command([
-                        "text-completion-bedrock",
-                        "-p",
-                        url.pulsar,
-                        "-x",
-                        std.toString($["bedrock-max-output-tokens"]),
-                        "-t",
-                        "%0.3f" % $["bedrock-temperature"],
-                        "-m",
-                        $["bedrock-model"],
-                        "-i",
-                        "non-persistent://tg/request/text-completion-rag",
-                        "-o",
-                        "non-persistent://tg/response/text-completion-rag",
-              	    ])
-                    .with_env_var_secrets(envSecrets)
-                    .with_limits("0.5", "128M")
-                    .with_reservations("0.1", "128M");
-
             local containerSet = engine.containers(
                 "text-completion", [ container ]
             );
 
-            local containerSetRag = engine.containers(
-                "text-completion-rag", [ containerRag ]
-            );
-
             local service =
                 engine.internalService(containerSet)
                 .with_port(8000, 8000, "metrics");
 
-            local serviceRag =
-                engine.internalService(containerSetRag)
-                .with_port(8000, 8000, "metrics");
-
             engine.resources([
                 envSecrets,
                 containerSet,
-                containerSetRag,
                 service,
-                serviceRag,
             ])
 
     },

templates/components/claude-rag.jsonnet

@@ -0,0 +1,60 @@
+local base = import "base/base.jsonnet";
+local images = import "values/images.jsonnet";
+local url = import "values/url.jsonnet";
+local prompts = import "prompts/mixtral.jsonnet";
+
+{
+
+    with:: function(key, value)
+        self + {
+            ["claude-rag-" + key]:: value,
+        },
+
+    "claude-rag-max-output-tokens":: 4096,
+    "claude-rag-temperature":: 0.0,
+
+    "text-completion" +: {
+    
+        create:: function(engine)
+
+            local envSecrets = engine.envSecrets("claude-credentials")
+                .with_env_var("CLAUDE_KEY", "claude-key");
+
+            local containerRag =
+                engine.container("text-completion-rag")
+                    .with_image(images.trustgraph)
+                    .with_command([
+                        "text-completion-claude",
+                        "-p",
+                        url.pulsar,
+                        "-x",
+                        std.toString($["claude-rag-max-output-tokens"]),
+                        "-t",
+                        "%0.3f" % $["claude-rag-temperature"],
+                        "-i",
+                        "non-persistent://tg/request/text-completion-rag",
+                        "-o",
+                        "non-persistent://tg/response/text-completion-rag",
+                    ])
+                    .with_env_var_secrets(envSecrets)
+                    .with_limits("0.5", "128M")
+                    .with_reservations("0.1", "128M");
+
+            local containerSetRag = engine.containers(
+                "text-completion-rag", [ containerRag ]
+            );
+
+            local serviceRag =
+                engine.internalService(containerSetRag)
+                .with_port(8000, 8000, "metrics");
+
+            engine.resources([
+                envSecrets,
+                containerSetRag,
+                serviceRag,
+            ])
+
+    },
+
+} + prompts
+

templates/components/claude.jsonnet

@@ -5,6 +5,11 @@ local prompts = import "prompts/mixtral.jsonnet";
 
 {
 
+    with:: function(key, value)
+        self + {
+            ["claude-" + key]:: value,
+        },
+
     "claude-max-output-tokens":: 4096,
     "claude-temperature":: 0.0,
 
@@ -31,48 +36,18 @@ local prompts = import "prompts/mixtral.jsonnet";
                     .with_limits("0.5", "128M")
                     .with_reservations("0.1", "128M");
 
-            local containerRag =
-                engine.container("text-completion-rag")
-                    .with_image(images.trustgraph)
-                    .with_command([
-                        "text-completion-claude",
-                        "-p",
-                        url.pulsar,
-                        "-x",
-                        std.toString($["claude-max-output-tokens"]),
-                        "-t",
-                        "%0.3f" % $["claude-temperature"],
-                        "-i",
-                        "non-persistent://tg/request/text-completion-rag",
-                        "-o",
-                        "non-persistent://tg/response/text-completion-rag",
-                    ])
-                    .with_env_var_secrets(envSecrets)
-                    .with_limits("0.5", "128M")
-                    .with_reservations("0.1", "128M");
-
             local containerSet = engine.containers(
                 "text-completion", [ container ]
             );
 
-            local containerSetRag = engine.containers(
-                "text-completion-rag", [ containerRag ]
-            );
-
             local service =
                 engine.internalService(containerSet)
                 .with_port(8000, 8000, "metrics");
 
-            local serviceRag =
-                engine.internalService(containerSetRag)
-                .with_port(8000, 8000, "metrics");
-
             engine.resources([
                 envSecrets,
                 containerSet,
-                containerSetRag,
                 service,
-                serviceRag,
             ])
 
     },

templates/components/cohere-rag.jsonnet

@@ -0,0 +1,56 @@
+local base = import "base/base.jsonnet";
+local images = import "values/images.jsonnet";
+local url = import "values/url.jsonnet";
+local prompts = import "prompts/mixtral.jsonnet";
+
+{
+
+    with:: function(key, value)
+        self + {
+            ["cohere-rag-" + key]:: value,
+        },
+
+    "cohere-rag-temperature":: 0.0,
+
+    "text-completion" +: {
+    
+        create:: function(engine)
+
+            local envSecrets = engine.envSecrets("cohere-credentials")
+                .with_env_var("COHERE_KEY", "cohere-key");
+
+            local containerRag =
+                engine.container("text-completion-rag")
+                    .with_image(images.trustgraph)
+                    .with_command([
+                        "text-completion-cohere",
+                        "-p",
+                        url.pulsar,
+                        "-t",
+                        "%0.3f" % $["cohere-rag-temperature"],
+                        "-i",
+                        "non-persistent://tg/request/text-completion-rag",
+                        "-o",
+                        "non-persistent://tg/response/text-completion-rag",
+                    ])
+                    .with_limits("0.5", "128M")
+                    .with_reservations("0.1", "128M");
+
+            local containerSetRag = engine.containers(
+                "text-completion-rag", [ containerRag ]
+            );
+
+            local serviceRag =
+                engine.internalService(containerSetRag)
+                .with_port(8000, 8000, "metrics");
+
+            engine.resources([
+                envSecrets,
+                containerSetRag,
+                serviceRag,
+            ])
+
+    },
+
+} + prompts
+

templates/components/cohere.jsonnet

@@ -5,9 +5,10 @@ local prompts = import "prompts/mixtral.jsonnet";
 
 {
 
-    // Override chunking
+    with:: function(key, value)
-    "chunk-size":: 150,
+        self + {
-    "chunk-overlap":: 10,
+            ["cohere-" + key]:: value,
+        },
 
     "cohere-temperature":: 0.0,
 
@@ -31,45 +32,18 @@ local prompts = import "prompts/mixtral.jsonnet";
                     .with_limits("0.5", "128M")
                     .with_reservations("0.1", "128M");
 
-            local containerRag =
-                engine.container("text-completion-rag")
-                    .with_image(images.trustgraph)
-                    .with_command([
-                        "text-completion-cohere",
-                        "-p",
-                        url.pulsar,
-                        "-t",
-                        "%0.3f" % $["cohere-temperature"],
-                        "-i",
-                        "non-persistent://tg/request/text-completion-rag",
-                        "-o",
-                        "non-persistent://tg/response/text-completion-rag",
-                    ])
-                    .with_limits("0.5", "128M")
-                    .with_reservations("0.1", "128M");
-
             local containerSet = engine.containers(
                 "text-completion", [ container ]
             );
 
-            local containerSetRag = engine.containers(
-                "text-completion-rag", [ containerRag ]
-            );
-
             local service =
                 engine.internalService(containerSet)
                 .with_port(8000, 8000, "metrics");
 
-            local serviceRag =
-                engine.internalService(containerSetRag)
-                .with_port(8000, 8000, "metrics");
-
             engine.resources([
                 envSecrets,
                 containerSet,
-                containerSetRag,
                 service,
-                serviceRag,
             ])
 
     },

templates/components/googleaistudio-rag.jsonnet

@@ -0,0 +1,65 @@
+local base = import "base/base.jsonnet";
+local images = import "values/images.jsonnet";
+local url = import "values/url.jsonnet";
+local prompts = import "prompts/mixtral.jsonnet";
+
+{
+
+    with:: function(key, value)
+        self + {
+            ["googleaistudio-rag-" + key]:: value,
+        },
+
+    "googleaistudio-rag-max-output-tokens":: 4096,
+    "googleaistudio-rag-temperature":: 0.0,
+    "googleaistudio-rag-model":: "gemini-1.5-flash-002",
+
+    "text-completion" +: {
+    
+        create:: function(engine)
+
+            local envSecrets = engine.envSecrets("googleaistudio-key")
+                .with_env_var("GOOGLE_AI_STUDIO_KEY", "googleaistudio-key");
+
+            local containerRag =
+                engine.container("text-completion-rag")
+                    .with_image(images.trustgraph)
+                    .with_command([
+                        "text-completion-googleaistudio",
+                        "-p",
+                        url.pulsar,
+                        "-x",
+                        std.toString(
+                            $["googleaistudio-rag-max-output-tokens"]
+                        ),
+                        "-t",
+                        "%0.3f" % $["googleaistudio-rag-temperature"],
+                        "-m",
+                        $["googleaistudio-rag-model"],
+                        "-i",
+                        "non-persistent://tg/request/text-completion-rag",
+                        "-o",
+                        "non-persistent://tg/response/text-completion-rag",
+                    ])
+                    .with_env_var_secrets(envSecrets)
+                    .with_limits("0.5", "128M")
+                    .with_reservations("0.1", "128M");
+
+            local containerSetRag = engine.containers(
+                "text-completion-rag", [ containerRag ]
+            );
+
+            local serviceRag =
+                engine.internalService(containerSetRag)
+                .with_port(8000, 8000, "metrics");
+
+            engine.resources([
+                envSecrets,
+                containerSetRag,
+                serviceRag,
+            ])
+
+    },
+
+} + prompts
+

templates/components/googleaistudio.jsonnet

@@ -34,50 +34,18 @@ local prompts = import "prompts/mixtral.jsonnet";
                     .with_limits("0.5", "128M")
                     .with_reservations("0.1", "128M");
 
-            local containerRag =
-                engine.container("text-completion-rag")
-                    .with_image(images.trustgraph)
-                    .with_command([
-                        "text-completion-googleaistudio",
-                        "-p",
-                        url.pulsar,
-                        "-x",
-                        std.toString($["googleaistudio-max-output-tokens"]),
-                        "-t",
-                        "%0.3f" % $["googleaistudio-temperature"],
-                        "-m",
-                        $["googleaistudio-model"],
-                        "-i",
-                        "non-persistent://tg/request/text-completion-rag",
-                        "-o",
-                        "non-persistent://tg/response/text-completion-rag",
-                    ])
-                    .with_env_var_secrets(envSecrets)
-                    .with_limits("0.5", "128M")
-                    .with_reservations("0.1", "128M");
-
             local containerSet = engine.containers(
                 "text-completion", [ container ]
             );
 
-            local containerSetRag = engine.containers(
-                "text-completion-rag", [ containerRag ]
-            );
-
             local service =
                 engine.internalService(containerSet)
                 .with_port(8000, 8000, "metrics");
 
-            local serviceRag =
-                engine.internalService(containerSetRag)
-                .with_port(8000, 8000, "metrics");
-
             engine.resources([
                 envSecrets,
                 containerSet,
-                containerSetRag,
                 service,
-                serviceRag,
             ])
 
     },

templates/components/llamafile-rag.jsonnet

@@ -0,0 +1,57 @@
+local base = import "base/base.jsonnet";
+local images = import "values/images.jsonnet";
+local url = import "values/url.jsonnet";
+local prompts = import "prompts/slm.jsonnet";
+
+{
+
+    with:: function(key, value)
+        self + {
+            ["llamafile-rag-" + key]:: value,
+        },
+
+    "llamafile-rag-model":: "LLaMA_CPP",
+
+    "text-completion-rag" +: {
+    
+        create:: function(engine)
+
+            local envSecrets = engine.envSecrets("llamafile-credentials")
+                .with_env_var("LLAMAFILE_URL", "llamafile-url");
+
+            local containerRag =
+                engine.container("text-completion-rag")
+                    .with_image(images.trustgraph)
+                    .with_command([
+                        "text-completion-llamafile",
+                        "-p",
+                        url.pulsar,
+                        "-m",
+                        $["llamafile-rag-model"],
+                        "-i",
+                        "non-persistent://tg/request/text-completion-rag",
+                        "-o",
+                        "non-persistent://tg/response/text-completion-rag",
+                    ])
+                    .with_env_var_secrets(envSecrets)
+                    .with_limits("0.5", "128M")
+                    .with_reservations("0.1", "128M");
+
+            local containerSetRag = engine.containers(
+                "text-completion-rag", [ containerRag ]
+            );
+
+            local serviceRag =
+                engine.internalService(containerSetRag)
+                .with_port(8080, 8080, "metrics");
+
+            engine.resources([
+                envSecrets,
+                containerSetRag,
+                serviceRag,
+            ])
+
+    },
+
+} + prompts
+

templates/components/llamafile.jsonnet

@@ -5,6 +5,11 @@ local prompts = import "prompts/slm.jsonnet";
 
 {
 
+    with:: function(key, value)
+        self + {
+            ["llamafile-" + key]:: value,
+        },
+
     "llamafile-model":: "LLaMA_CPP",
 
     "text-completion" +: {
@@ -28,46 +33,18 @@ local prompts = import "prompts/slm.jsonnet";
                     .with_limits("0.5", "128M")
                     .with_reservations("0.1", "128M");
 
-            local containerRag =
-                engine.container("text-completion-rag")
-                    .with_image(images.trustgraph)
-                    .with_command([
-                        "text-completion-llamafile",
-                        "-p",
-                        url.pulsar,
-                        "-m",
-                        $["llamafile-model"],
-                        "-i",
-                        "non-persistent://tg/request/text-completion-rag",
-                        "-o",
-                        "non-persistent://tg/response/text-completion-rag",
-                    ])
-                    .with_env_var_secrets(envSecrets)
-                    .with_limits("0.5", "128M")
-                    .with_reservations("0.1", "128M");
-
             local containerSet = engine.containers(
                 "text-completion", [ container ]
             );
 
-            local containerSetRag = engine.containers(
-                "text-completion-rag", [ containerRag ]
-            );
-
             local service =
                 engine.internalService(containerSet)
                 .with_port(8080, 8080, "metrics");
 
-            local serviceRag =
-                engine.internalService(containerSetRag)
-                .with_port(8080, 8080, "metrics");
-
             engine.resources([
                 envSecrets,
                 containerSet,
-                containerSetRag,
                 service,
-                serviceRag,
             ])
 
     },

templates/components/ollama-rag.jsonnet

@@ -0,0 +1,57 @@
+local base = import "base/base.jsonnet";
+local images = import "values/images.jsonnet";
+local url = import "values/url.jsonnet";
+local prompts = import "prompts/mixtral.jsonnet";
+
+{
+
+    with:: function(key, value)
+        self + {
+            ["ollama-rag-" + key]:: value,
+        },
+
+    "ollama-rag-model":: "gemma2:9b",
+
+    "text-completion-rag" +: {
+    
+        create:: function(engine)
+
+            local envSecrets = engine.envSecrets("ollama-credentials")
+                .with_env_var("OLLAMA_HOST", "ollama-host");
+
+            local containerRag =
+                engine.container("text-completion-rag")
+                    .with_image(images.trustgraph)
+                    .with_command([
+                        "text-completion-ollama",
+                        "-p",
+                        url.pulsar,
+                        "-m",
+                        $["ollama-rag-model"],
+                        "-i",
+                        "non-persistent://tg/request/text-completion-rag",
+                        "-o",
+                        "non-persistent://tg/response/text-completion-rag",
+                    ])
+                    .with_env_var_secrets(envSecrets)
+                    .with_limits("0.5", "128M")
+                    .with_reservations("0.1", "128M");
+
+            local containerSetRag = engine.containers(
+                "text-completion-rag", [ containerRag ]
+            );
+
+            local serviceRag =
+                engine.internalService(containerSetRag)
+                .with_port(8080, 8080, "metrics");
+
+            engine.resources([
+                envSecrets,
+                containerSetRag,
+                serviceRag,
+            ])
+
+    },
+
+} + prompts
+

templates/components/ollama.jsonnet

@@ -5,6 +5,11 @@ local prompts = import "prompts/mixtral.jsonnet";
 
 {
 
+    with:: function(key, value)
+        self + {
+            ["ollama-" + key]:: value,
+        },
+
     "ollama-model":: "gemma2:9b",
 
     "text-completion" +: {
@@ -28,46 +33,18 @@ local prompts = import "prompts/mixtral.jsonnet";
                     .with_limits("0.5", "128M")
                     .with_reservations("0.1", "128M");
 
-            local containerRag =
-                engine.container("text-completion-rag")
-                    .with_image(images.trustgraph)
-                    .with_command([
-                        "text-completion-ollama",
-                        "-p",
-                        url.pulsar,
-                        "-m",
-                        $["ollama-model"],
-                        "-i",
-                        "non-persistent://tg/request/text-completion-rag",
-                        "-o",
-                        "non-persistent://tg/response/text-completion-rag",
-                    ])
-                    .with_env_var_secrets(envSecrets)
-                    .with_limits("0.5", "128M")
-                    .with_reservations("0.1", "128M");
-
             local containerSet = engine.containers(
                 "text-completion", [ container ]
             );
 
-            local containerSetRag = engine.containers(
-                "text-completion-rag", [ containerRag ]
-            );
-
             local service =
                 engine.internalService(containerSet)
                 .with_port(8080, 8080, "metrics");
 
-            local serviceRag =
-                engine.internalService(containerSetRag)
-                .with_port(8080, 8080, "metrics");
-
             engine.resources([
                 envSecrets,
                 containerSet,
-                containerSetRag,
                 service,
-                serviceRag,
             ])
 
     },

templates/components/openai-rag.jsonnet

@@ -0,0 +1,63 @@
+local base = import "base/base.jsonnet";
+local images = import "values/images.jsonnet";
+local url = import "values/url.jsonnet";
+local prompts = import "prompts/mixtral.jsonnet";
+
+{
+
+    with:: function(key, value)
+        self + {
+            ["openai-rag-" + key]:: value,
+        },
+
+    "openai-rag-max-output-tokens":: 4096,
+    "openai-rag-temperature":: 0.0,
+    "openai-rag-model":: "GPT-3.5-Turbo",
+
+    "text-completion-rag" +: {
+    
+        create:: function(engine)
+
+            local envSecrets = engine.envSecrets("openai-credentials")
+                .with_env_var("OPENAI_TOKEN", "openai-token");
+
+            local containerRag =
+                engine.container("text-completion-rag")
+                    .with_image(images.trustgraph)
+                    .with_command([
+                        "text-completion-openai",
+                        "-p",
+                        url.pulsar,
+                        "-x",
+                        std.toString($["openai-rag-max-output-tokens"]),
+                        "-t",
+                        "%0.3f" % $["openai-rag-temperature"],
+                        "-m",
+                        $["openai-rag-model"],
+                        "-i",
+                        "non-persistent://tg/request/text-completion-rag",
+                        "-o",
+                        "non-persistent://tg/response/text-completion-rag",
+                    ])
+                    .with_env_var_secrets(envSecrets)
+                    .with_limits("0.5", "128M")
+                    .with_reservations("0.1", "128M");
+
+            local containerSetRag = engine.containers(
+                "text-completion-rag", [ containerRag ]
+            );
+
+            local serviceRag =
+                engine.internalService(containerSetRag)
+                .with_port(8080, 8080, "metrics");
+
+            engine.resources([
+                envSecrets,
+                containerSetRag,
+                serviceRag,
+            ])
+
+    },
+
+} + prompts
+

templates/components/openai.jsonnet

@@ -5,6 +5,11 @@ local prompts = import "prompts/mixtral.jsonnet";
 
 {
 
+    with:: function(key, value)
+        self + {
+            ["openai-" + key]:: value,
+        },
+
     "openai-max-output-tokens":: 4096,
     "openai-temperature":: 0.0,
     "openai-model":: "GPT-3.5-Turbo",
@@ -34,50 +39,18 @@ local prompts = import "prompts/mixtral.jsonnet";
                     .with_limits("0.5", "128M")
                     .with_reservations("0.1", "128M");
 
-            local containerRag =
-                engine.container("text-completion-rag")
-                    .with_image(images.trustgraph)
-                    .with_command([
-                        "text-completion-openai",
-                        "-p",
-                        url.pulsar,
-                        "-x",
-                        std.toString($["openai-max-output-tokens"]),
-                        "-t",
-                        "%0.3f" % $["openai-temperature"],
-                        "-m",
-                        $["openai-model"],
-                        "-i",
-                        "non-persistent://tg/request/text-completion-rag",
-                        "-o",
-                        "non-persistent://tg/response/text-completion-rag",
-                    ])
-                    .with_env_var_secrets(envSecrets)
-                    .with_limits("0.5", "128M")
-                    .with_reservations("0.1", "128M");
-
             local containerSet = engine.containers(
                 "text-completion", [ container ]
             );
 
-            local containerSetRag = engine.containers(
-                "text-completion-rag", [ containerRag ]
-            );
-
             local service =
                 engine.internalService(containerSet)
                 .with_port(8080, 8080, "metrics");
 
-            local serviceRag =
-                engine.internalService(containerSetRag)
-                .with_port(8080, 8080, "metrics");
-
             engine.resources([
                 envSecrets,
                 containerSet,
-                containerSetRag,
                 service,
-                serviceRag,
             ])
 
     },

templates/components/prompt-overrides.jsonnet

@@ -1,7 +1,3 @@
-local base = import "base/base.jsonnet";
-local images = import "values/images.jsonnet";
-local url = import "values/url.jsonnet";
-local prompts = import "prompts/mixtral.jsonnet";
 local default_prompts = import "prompts/default-prompts.jsonnet";
 
 {

templates/components/trustgraph.jsonnet

@@ -1,7 +1,6 @@
 local base = import "base/base.jsonnet";
 local images = import "values/images.jsonnet";
 local url = import "values/url.jsonnet";
-local prompt = import "prompt-template.jsonnet";
 
 {
 
@@ -181,5 +180,5 @@ local prompt = import "prompt-template.jsonnet";
 
     },
 
-} + prompt
+}