Move LM clients to LiteLMM (#252)

* Use LiteLLM for all LMs 
* Use API_BASE for Model
* Fix Streaming Issue
* Removed cyclic import on databases
* Removed cyclic import on oci
* IaC Updates
* LangChain MCP template fix
* API Server Control

---------

Co-authored-by: corradodebari <corrado.de.bari@oracle.com>

Author: gotsysdba

.pylintrc

@@ -80,6 +80,7 @@ resource "oci_database_autonomous_database" "default_adb" {
 module "vm" {
   count                 = var.infrastructure == "VM" ? 1 : 0
   source                = "./modules/vm"
+  optimizer_version     = var.optimizer_version
   label_prefix          = local.label_prefix
   tenancy_id            = var.tenancy_ocid
   compartment_id        = local.compartment_ocid

opentofu/main.tf

@@ -42,8 +42,8 @@ resource "oci_identity_policy" "workers_policies" {
     format("allow any-user to read objects in compartment id %s where all {request.principal.type = 'workload', request.principal.namespace = '%s', request.principal.cluster_id = '%s'}", var.compartment_id, var.label_prefix, oci_containerengine_cluster.default_cluster.id),
     format("allow any-user to manage repos in compartment id %s where all {request.principal.type = 'workload', request.principal.namespace = '%s', request.principal.cluster_id = '%s'}", var.compartment_id, var.label_prefix, oci_containerengine_cluster.default_cluster.id),
     # Instance Principles
-    format("allow dynamic-group %s to use generative-ai-family in compartment id %s", oci_identity_dynamic_group.workers_dynamic_group.name, var.compartment_id),
     format("allow dynamic-group %s to manage repos in compartment id %s", oci_identity_dynamic_group.workers_dynamic_group.name, var.compartment_id),
+    format("allow dynamic-group %s to use generative-ai-family in tenancy", oci_identity_dynamic_group.workers_dynamic_group.name),
   ]
   provider = oci.home_region
 }

opentofu/modules/kubernetes/iam.tf

@@ -34,6 +34,10 @@ resource "oci_identity_policy" "identity_node_policies" {
       "allow dynamic-group %s to read objects in compartment id %s",
       oci_identity_dynamic_group.compute_dynamic_group.name, var.compartment_id
     ),
+    format(
+      "allow dynamic-group %s to use generative-ai-family in tenancy",
+      oci_identity_dynamic_group.compute_dynamic_group.name
+    ),
   ]
   provider = oci.home_region
 }

opentofu/modules/vm/iam.tf

@@ -4,12 +4,13 @@
 
 locals {
   cloud_init_compute = templatefile("${path.module}/templates/cloudinit-compute.tpl", {
-    tenancy_id     = var.tenancy_id
-    compartment_id = var.compartment_id
-    oci_region     = var.region
-    db_name        = var.adb_name
-    db_password    = var.adb_password
-    install_ollama = var.vm_is_gpu_shape ? true : false
+    tenancy_id        = var.tenancy_id
+    compartment_id    = var.compartment_id
+    oci_region        = var.region
+    db_name           = var.adb_name
+    db_password       = var.adb_password
+    optimizer_version = var.optimizer_version
+    install_ollama    = var.vm_is_gpu_shape ? true : false
   })
 
   cloud_init_database = templatefile("${path.module}/templates/cloudinit-database.tpl", {

opentofu/modules/vm/locals.tf

@@ -57,11 +57,14 @@ write_files:
     permissions: '0755'
     content: |
       #!/bin/bash
-      # Setup for Instance Principles
-
       # Download/Setup Source Code
-      curl -L https://github.com/oracle/ai-optimizer/releases/latest/download/ai-optimizer-src.tar.gz \
-      | tar -xz -C /app
+      if [ "${optimizer_version}" = "main" ]; then
+          URL="https://github.com/oracle/ai-optimizer/archive/refs/heads/main.tar.gz"
+          curl -L "$URL" | tar -xz -C /app --strip-components=2 ai-optimizer-main/src
+      else
+          URL="https://github.com/oracle/ai-optimizer/releases/latest/download/ai-optimizer-src.tar.gz"
+          curl -L "$URL" | tar -xz -C /app
+      fi
       cd /app
       python3.11 -m venv .venv
       source .venv/bin/activate
@@ -84,6 +87,7 @@ write_files:
     content: |
       #!/bin/bash
       export OCI_CLI_AUTH=instance_principal
+      export API_SERVER_CONTROL="True"
       export DB_USERNAME='AI_OPTIMIZER'
       export DB_PASSWORD='${db_password}'
       export DB_DSN='${db_name}_TP'

opentofu/modules/vm/templates/cloudinit-compute.tpl

@@ -2,6 +2,10 @@
 # All rights reserved. The Universal Permissive License (UPL), Version 1.0 as shown at http://oss.oracle.com/licenses/upl
 # spell-checker: disable
 
+variable "optimizer_version" {
+  type = string
+}
+
 variable "tenancy_id" {
   type = string
 }

opentofu/modules/vm/variables.tf

@@ -28,6 +28,7 @@ variableGroups:
 
   - title: "Hidden (Defaults)"
     variables:
+      - optimizer_version
       - adb_version
       - k8s_run_cfgmgt
     visible: false

opentofu/schema.yaml

@@ -3,6 +3,16 @@
 # spell-checker: disable
 
 // Standard Default Vars
+variable "optimizer_version" {
+  description = "Determines if latest release or main is used"
+  type        = string
+  default     = "latest"
+  validation {
+    condition     = var.optimizer_version == "latest" || var.optimizer_version == "main"
+    error_message = "optimizer_version must be either 'latest' or 'main'."
+  }
+}
+
 variable "tenancy_ocid" {
   description = "The Tenancy ID of the OCI Cloud Account in which to create the resources."
   type        = string

opentofu/variables.tf

@@ -14,18 +14,15 @@
 import streamlit as st
 from streamlit import session_state as state
 
-import client.utils.client as client
-import client.utils.api_call as api_call
-import common.logging_config as logging_config
+from client.utils import client, api_call
+from common import logging_config
 
 logger = logging_config.logging.getLogger("client.content.api_server")
 
 try:
     import launch_server
-
-    REMOTE_SERVER = False
 except ImportError:
-    REMOTE_SERVER = True
+    pass
 
 
 #####################################################
@@ -72,16 +69,16 @@ async def main() -> None:
         key="user_server_port",
         min_value=1,
         max_value=65535,
-        disabled=REMOTE_SERVER,
+        disabled=not state.server["control"],
     )
     right.text_input(
         "API Server Key:",
         value=state.server["key"],
         key="user_server_key",
         type="password",
-        disabled=REMOTE_SERVER,
+        disabled=not state.server["control"],
     )
-    if not REMOTE_SERVER:
+    if state.server["control"]:
         st.button("Restart Server", type="primary", on_click=server_restart)
 
     st.header("Server Settings", divider="red")