OCI GenAI - Auto Model Loading (#238)

* Bump  Python Modules
* Dynamic OCI GenAI Model
* Boostrap GenAI Models
* OCI_GENAI_SERVICE_ENDPOINT replaced by OCI_GENAI_REGION

Author: gotsysdba

.github/workflows/pytest.yml

@@ -50,7 +50,7 @@ jobs:
         run: |
           cd src/
           python -m pip install --upgrade pip wheel setuptools
-          pip install torch==2.7.1+cpu -f https://download.pytorch.org/whl/cpu/torch
+          pip install torch==2.8.0+cpu -f https://download.pytorch.org/whl/cpu/torch
           pip install -e ".[all-test]"
 
       - name: Run All Tests

docs/content/client/configuration/model_config.md

@@ -89,14 +89,14 @@ To use OCI GenAI, the {{< short_app_ref >}} must be configured for [OCI access](
 >[!code]Skip the GUI!
 >You can set the following environment variables to automatically enable OCI GenAI models:
 >```shell
->export OCI_GENAI_SERVICE_ENDPOINT=<OCI GenAI Service Endpoint>
+>export OCI_GENAI_REGION=<OCI GenAI Service Region>
 >export OCI_GENAI_COMPARTMENT_ID=<OCI Compartment OCID of the OCI GenAI Service>
 >```
 >
 >Alternatively, you can specify the following in the `~/.oci/config` configfile under the appropriate OCI profile:
 >```shell
->service_endpoint=<OCI GenAI Service Endpoint>
->compartment_id=<OCI Compartment OCID of the OCI GenAI Service>
+>genai_compartment_id=<OCI Compartment OCID of the OCI GenAI Service>
+>genai_region=<OCI GenAI Region>
 >```
 
 {{% /tab %}}

docs/content/client/configuration/oci_config.md

@@ -42,8 +42,8 @@ Depending on the runtime environment, either [Bare Metal](#bare-metal) or [Conta
 
 In addition to the standard configuration file, two additional entries are required to enable OCI GenAI Services:
 
-- **service_endpoint**: the URL endpoint for the OCI GenAI Service
-- **compartment_id**: the compartment OCID of the OCI GenAI Service
+- **genai_region**: the Region for the OCI GenAI Service
+- **genai_compartment_id**: the Compartment OCID of the OCI GenAI Service
 
 #### Bare Metal
 

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

@@ -65,7 +65,7 @@ write_files:
       python3.11 -m venv .venv
       source .venv/bin/activate
       pip3.11 install --upgrade pip wheel setuptools
-      pip3.11 install torch==2.7.1+cpu -f https://download.pytorch.org/whl/cpu/torch
+      pip3.11 install torch==2.8.0+cpu -f https://download.pytorch.org/whl/cpu/torch
       pip3.11 install -e ".[all]" --quiet --no-input &
       INSTALL_PID=$!
 

src/Dockerfile

@@ -24,7 +24,7 @@ RUN for dir in server client common; do \
 COPY pyproject.toml /opt/package/pyproject.toml
 # Use the virtual environment for pip installations
 RUN /opt/.venv/bin/pip install --upgrade pip wheel setuptools && \
-    /opt/.venv/bin/pip install torch==2.7.1+cpu -f https://download.pytorch.org/whl/cpu/torch && \
+    /opt/.venv/bin/pip install torch==2.8.0+cpu -f https://download.pytorch.org/whl/cpu/torch && \
     /opt/.venv/bin/pip install --no-cache-dir "/opt/package[all]"
 
 ##################################################

src/client/content/config/models.py

@@ -108,6 +108,7 @@ def edit_model(model_type: str, action: Literal["add", "edit"], model_id: str =
         )
         api_values = get_model_apis(model_type)
         api_index = next((i for i, item in enumerate(api_values) if item == model["api"]), None)
+        disable_for_oci = model["api"] in ["ChatOCIGenAI", "OCIGenAIEmbeddings"]
         model["api"] = st.selectbox(
             "API (Required):",
             help=help_text.help_dict["model_api"],
@@ -122,13 +123,15 @@ def edit_model(model_type: str, action: Literal["add", "edit"], model_id: str =
             help=help_text.help_dict["model_api_url"],
             key="add_model_api_url",
             value=model.get("url", ""),
+            disabled=disable_for_oci
         )
         model["api_key"] = st.text_input(
             "API Key:",
             help=help_text.help_dict["model_api_key"],
             key="add_model_api_key",
             type="password",
             value=model.get("api_key", ""),
+            disabled=disable_for_oci
         )
         if model_type == "ll":
             model["context_length"] = st.number_input(
@@ -201,7 +204,7 @@ def edit_model(model_type: str, action: Literal["add", "edit"], model_id: str =
 
 def render_model_rows(model_type):
     """Render rows of the models"""
-    data_col_widths = [0.05, 0.25, 0.2, 0.28, 0.12]
+    data_col_widths = [0.07, 0.23, 0.2, 0.28, 0.12]
     table_col_format = st.columns(data_col_widths, vertical_alignment="center")
     col1, col2, col3, col4, col5 = table_col_format
     col1.markdown("​", help="Active", unsafe_allow_html=True)
@@ -213,7 +216,7 @@ def render_model_rows(model_type):
         model_id = model["id"]
         col1.text_input(
             "Enabled",
-            value="✅" if model["enabled"] else "⚪",
+            value=st_common.bool_to_emoji(model["enabled"]),
             key=f"{model_type}_{model_id}_enabled",
             label_visibility="collapsed",
             disabled=True,

src/client/content/config/oci.py

@@ -8,7 +8,7 @@
 # spell-checker:ignore streamlit, ocid, selectbox, genai, oraclecloud
 
 import inspect
-import re
+import pandas as pd
 
 import streamlit as st
 from streamlit import session_state as state
@@ -36,7 +36,21 @@ def get_oci(force: bool = False) -> None:
             state.oci_configs = {}
 
 
-def patch_oci(auth_profile: str, supplied: dict, namespace: str) -> bool:
+def get_genai_models() -> list[dict]:
+    """Get Subscribed OCI Regions"""
+    endpoint = f"v1/oci/genai/{state.client_settings['oci']['auth_profile']}"
+    genai_models = api_call.get(endpoint=endpoint)
+    return genai_models
+
+
+def create_genai_models() -> list[dict]:
+    """Create OCI GenAI Models"""
+    endpoint = f"v1/oci/genai/{state.client_settings['oci']['auth_profile']}"
+    genai_models = api_call.post(endpoint=endpoint)
+    return genai_models
+
+
+def patch_oci(auth_profile: str, supplied: dict, namespace: str, toast: bool = True) -> bool:
     """Update OCI"""
     rerun = False
     # Check if the OIC configuration is changed, or no namespace
@@ -49,19 +63,15 @@ def patch_oci(auth_profile: str, supplied: dict, namespace: str) -> bool:
                 supplied["authentication"] = "security_token"
 
             with st.spinner(text="Updating OCI Profile...", show_time=True):
-                _ = api_call.patch(
-                    endpoint=f"v1/oci/{auth_profile}",
-                    payload={"json": supplied},
-                )
+                _ = api_call.patch(endpoint=f"v1/oci/{auth_profile}", payload={"json": supplied}, toast=toast)
             logger.info("OCI Profile updated: %s", auth_profile)
         except api_call.ApiError as ex:
             logger.error("OCI Update failed: %s", ex)
             state.oci_error = ex
         st_common.clear_state_key("oci_configs")
-        if supplied.get("service_endpoint"):
-            st_common.clear_state_key("model_configs")
     else:
-        st.toast("No Changes Detected.", icon="ℹ️")
+        if toast:
+            st.toast("No Changes Detected.", icon="ℹ️")
 
     return rerun
 
@@ -155,31 +165,54 @@ def main() -> None:
         OCI Authentication must be configured above.
         """)
     with st.container(border=True):
-        supplied["compartment_id"] = st.text_input(
+        if "genai_models" not in state:
+            state.genai_models = []
+        supplied["genai_compartment_id"] = st.text_input(
             "OCI GenAI Compartment OCID:",
-            value=oci_lookup[selected_oci_auth_profile]["compartment_id"],
+            value=oci_lookup[selected_oci_auth_profile]["genai_compartment_id"],
             placeholder="Compartment OCID for GenAI Services",
             key="oci_genai_compartment_id",
             disabled=not namespace,
         )
-        match = re.search(
-            r"\.([a-zA-Z\-0-9]+)\.oci\.oraclecloud\.com", oci_lookup[selected_oci_auth_profile]["service_endpoint"]
-        )
-        supplied["service_endpoint"] = match.group(1) if match else None
-        supplied["service_endpoint"] = st.text_input(
-            "OCI GenAI Region:",
-            value=oci_lookup[selected_oci_auth_profile]["service_endpoint"],
-            help="Region of GenAI Service",
-            key="oci_genai_region",
-            disabled=not namespace,
-        )
-
-        if st.button("Save OCI GenAI", key="save_oci_genai", disabled=not namespace):
-            if not (supplied["compartment_id"] and supplied["service_endpoint"]):
-                st.error("All fields are required.", icon="🛑")
+        if st.button("Check for OCI GenAI Models", key="check_oci_genai", disabled=not namespace):
+            if not supplied["genai_compartment_id"]:
+                st.error("OCI GenAI Compartment OCID is required.", icon="🛑")
                 st.stop()
-            if patch_oci(selected_oci_auth_profile, supplied, namespace):
-                st.rerun()
+            with st.spinner("Looking for OCI GenAI Models... please be patient.", show_time=True):
+                patch_oci(selected_oci_auth_profile, supplied, namespace, toast=False)
+                state.genai_models = get_genai_models()
+        if state.genai_models:
+            regions = list({item["region"] for item in state.genai_models if "region" in item})
+            supplied["genai_region"] = st.selectbox(
+                "Region:",
+                regions,
+                key="selected_genai_region",
+            )
+            filtered_models = [
+                m
+                for m in state.genai_models
+                if m["region"] == supplied["genai_region"]
+                and ("CHAT" in m["capabilities"] or "TEXT_EMBEDDINGS" in m["capabilities"])
+            ]
+            table_data = []
+            for m in filtered_models:
+                table_data.append(
+                    {
+                        "Model Name": m["model_name"],
+                        "Large Language": st_common.bool_to_emoji("CHAT" in m["capabilities"]),
+                        "Embedding": st_common.bool_to_emoji("TEXT_EMBEDDINGS" in m["capabilities"]),
+                    }
+                )
+
+            # Convert to DataFrame and display
+            df = pd.DataFrame(table_data)
+            st.dataframe(df, hide_index=True)
+            if st.button("Enable Region Models", key="enable_oci_region_models", type="primary"):
+                with st.spinner("Enabling OCI GenAI Models... please be patient.", show_time=True):
+                    patch_oci(selected_oci_auth_profile, supplied, namespace, toast=False)
+                    _ = create_genai_models()
+                    st_common.clear_state_key("model_configs")
+                st.success("Oracle GenAI models - Enabled.", icon="✅")
 
 
 if __name__ == "__main__" or "page.py" in inspect.stack()[1].filename:

src/client/utils/st_common.py

@@ -52,6 +52,9 @@ def enabled_models_lookup(model_type: str) -> dict[str, dict[str, Any]]:
 #############################################################################
 # Common Helpers
 #############################################################################
+def bool_to_emoji(value):
+    "Return an Emoji for Bools"
+    return "✅" if value else "⚪"
 def local_file_payload(uploaded_files: Union[BytesIO, list[BytesIO]]) -> list:
     """Upload Single file from Streamlit to the Server"""
     # If it's a single file, convert it to a list for consistent processing

src/common/schema.py

@@ -2,7 +2,7 @@
 Copyright (c) 2024, 2025, Oracle and/or its affiliates.
 Licensed under the Universal Permissive License v1.0 as shown at http://oss.oracle.com/licenses/upl.
 """
-# spell-checker:ignore ollama, hnsw, mult, ocid, testset, selectai, explainsql, showsql, vector_search, aioptimizer
+# spell-checker:ignore ollama hnsw mult ocid testset selectai explainsql showsql vector_search aioptimizer genai
 
 import time
 from typing import Optional, Literal, Union, get_args, Any
@@ -180,11 +180,21 @@ class OracleCloudSettings(BaseModel):
 
     auth_profile: str = Field(default="DEFAULT", description="Config File Profile")
     namespace: Optional[str] = Field(default=None, description="Object Store Namespace", readOnly=True)
-    user: Optional[str] = Field(default=None, description="Optional if using Auth Token")
+    user: Optional[str] = Field(
+        default=None,
+        description="Optional if using Auth Token",
+        pattern=r"^([0-9a-zA-Z-_]+[.:])([0-9a-zA-Z-_]*[.:]){3,}([0-9a-zA-Z-_]+)$",
+    )
     security_token_file: Optional[str] = Field(default=None, description="Security Key File for Auth Token")
     authentication: Literal["api_key", "instance_principal", "oke_workload_identity", "security_token"] = Field(
         default="api_key", description="Authentication Method."
     )
+    genai_compartment_id: Optional[str] = Field(
+        default=None,
+        description="Optional Compartment OCID of OCI GenAI services",
+        pattern=r"^([0-9a-zA-Z-_]+[.:])([0-9a-zA-Z-_]*[.:]){3,}([0-9a-zA-Z-_]+)$",
+    )
+    genai_region: Optional[str] = Field(default=None, description="Optional Region OCID of OCI GenAI services")
 
     model_config = {
         "extra": "allow"  # enable extra fields
@@ -471,6 +481,7 @@ class EvaluationReport(Evaluation):
 ModelTypeType = Model.__annotations__["type"]
 ModelEnabledType = ModelAccess.__annotations__["enabled"]
 OCIProfileType = OracleCloudSettings.__annotations__["auth_profile"]
+OCIResourceOCID = OracleResource.__annotations__["ocid"]
 PromptNameType = Prompt.__annotations__["name"]
 PromptCategoryType = Prompt.__annotations__["category"]
 PromptPromptType = PromptText.__annotations__["prompt"]

src/pyproject.toml

@@ -14,7 +14,7 @@ authors = [
 
 # Common dependencies that are always needed
 dependencies = [
-    "langchain-core==0.3.72",
+    "langchain-core==0.3.74",
     "httpx==0.28.1",
     "oracledb~=3.1",
     "plotly==6.2.0",
@@ -32,16 +32,16 @@ server = [
     "langchain-community==0.3.27",
     "langchain-huggingface==0.3.1",
     "langchain-ollama==0.3.6",
-    "langchain-openai==0.3.28",
-    "langgraph==0.6.3",
-    "litellm==1.75.0",
-    "llama-index==0.13.0",
+    "langchain-openai==0.3.29",
+    "langgraph==0.6.4",
+    "litellm==1.75.3",
+    "llama-index==0.13.1",
     "lxml==6.0.0",
     "matplotlib==3.10.5",
     "oci~=2.0",
     "psutil==7.0.0",
     "python-multipart==0.0.20",
-    "torch==2.7.1",
+    "torch==2.8.0",
     "umap-learn==0.5.9.post2",
     "uvicorn==0.35.0",
 ]