.github/workflows/ci-playground.yaml

@@ -60,7 +60,7 @@ jobs:
           push: ${{ github.event_name != 'pull_request' }}
           tags: ${{ steps.meta.outputs.tags }}
           labels: ${{ steps.meta.outputs.labels }}
-          context: .
+          context: llama_stack/distribution/ui
           file: llama_stack/distribution/ui/Containerfile
           provenance: mode=max
           sbom: true

kvant_start_local.sh

@@ -1,17 +1,25 @@
 #!/usr/bin/env bash
 
+export INFERENCE_MODEL="inference-llama4-maverick"
+export EMBEDDING_MODEL="inference-bge-m3"
+export EMBEDDING_DIMENSION="1024"
 export LLAMA_STACK_PORT=8321
-# VLLM_API_TOKEN= env file
+export OPENAI_BASE_URL=https://maas.ai-2.kvant.cloud/v1
-# KEYCLOAK_CLIENT_SECRET= env file
+export OPENAI_API_KEY=sk-ZqAWqBKFXjb6y3tVej2AaA
-
+export VLLM_MAX_TOKENS=125000
 
 docker run -it \
   -p $LLAMA_STACK_PORT:$LLAMA_STACK_PORT \
   -v $(pwd)/data:/root/.llama \
   --mount type=bind,source="$(pwd)"/llama_stack/templates/kvant/run.yaml,target=/root/.llama/config.yaml,readonly \
   --entrypoint python \
-  --env-file ./.env \
   distribution-kvant:dev \
   -m llama_stack.distribution.server.server  --config /root/.llama/config.yaml \
   --port $LLAMA_STACK_PORT \
-
+  --env VLLM_URL=$OPENAI_BASE_URL \
+  --env VLLM_API_TOKEN=$OPENAI_API_KEY \
+  --env PASSTHROUGH_URL=$OPENAI_BASE_URL \
+  --env PASSTHROUGH_API_KEY=$OPENAI_API_KEY \
+  --env INFERENCE_MODEL=$INFERENCE_MODEL \
+  --env EMBEDDING_MODEL=$EMBEDDING_MODEL \
+  --env EMBEDDING_DIMENSION=$EMBEDDING_DIMENSION \

llama_stack/distribution/ui/Containerfile

@@ -5,8 +5,7 @@ FROM python:3.12-slim
 WORKDIR /app
 COPY . /app/
 RUN /usr/local/bin/python -m pip install --upgrade pip && \
-    /usr/local/bin/pip3 install -r requirements.txt && \
+    /usr/local/bin/pip3 install -r requirements.txt
-    /usr/local/bin/pip3 install -r llama_stack/distribution/ui/requirements.txt
 EXPOSE 8501
 
-ENTRYPOINT ["streamlit", "run", "llama_stack/distribution/ui/app.py", "--server.port=8501", "--server.address=0.0.0.0"]
+ENTRYPOINT ["streamlit", "run", "app.py", "--server.port=8501", "--server.address=0.0.0.0"]

llama_stack/distribution/ui/README.md

@@ -48,6 +48,3 @@ uv run --with ".[ui]" streamlit run llama_stack/distribution/ui/app.py
 | TOGETHER_API_KEY           | API key for Together provider      | (empty string)            |
 | SAMBANOVA_API_KEY          | API key for SambaNova provider     | (empty string)            |
 | OPENAI_API_KEY             | API key for OpenAI provider        | (empty string)            |
-| KEYCLOAK_URL               | URL for keycloak authentication    | (empty string)            |
-| KEYCLOAK_REALM             | Keycloak realm                     | default                   |
-| KEYCLOAK_CLIENT_ID         | Client ID for keycloak auth        | (empty string)            |

llama_stack/distribution/ui/app.py

@@ -50,42 +50,6 @@ def main():
     )
     pg.run()
 
-def main2():
-    from dataclasses import asdict
-    st.subheader(f"Welcome {keycloak.user_info['preferred_username']}!")
-    st.write(f"Here is your user information:")
-    st.write(asdict(keycloak))
-
-def get_access_token() -> str|None:
-    return st.session_state.get('access_token')
 
 if __name__ == "__main__":
-    
+    main()
-    from streamlit_keycloak import login
-    import os
-    
-    keycloak_url = os.environ.get("KEYCLOAK_URL")
-    keycloak_realm = os.environ.get("KEYCLOAK_REALM", "default")
-    keycloak_client_id = os.environ.get("KEYCLOAK_CLIENT_ID")
-    
-    if keycloak_url and keycloak_client_id:
-        keycloak = login(
-            url=keycloak_url,
-            realm=keycloak_realm,
-            client_id=keycloak_client_id,
-            custom_labels={
-                "labelButton": "Sign in to kvant",
-                "labelLogin": "Please sign in to your kvant account.",
-                "errorNoPopup": "Unable to open the authentication popup. Allow popups and refresh the page to proceed.",
-                "errorPopupClosed": "Authentication popup was closed manually.",
-                "errorFatal": "Unable to connect to Keycloak using the current configuration."   
-            },
-            auto_refresh=True,
-        )
-
-        if keycloak.authenticated:
-            st.session_state['access_token'] = keycloak.access_token
-            main()
-    # TBD - add other authentications
-    else:
-        main()

llama_stack/distribution/ui/modules/api.py

@@ -7,13 +7,11 @@
 import os
 
 from llama_stack_client import LlamaStackClient
-from llama_stack.distribution.ui.app import get_access_token
 
 
 class LlamaStackApi:
     def __init__(self):
         self.client = LlamaStackClient(
-            api_key=get_access_token(),
             base_url=os.environ.get("LLAMA_STACK_ENDPOINT", "http://localhost:8321"),
             provider_data={
                 "fireworks_api_key": os.environ.get("FIREWORKS_API_KEY", ""),
@@ -30,3 +28,5 @@ class LlamaStackApi:
             scoring_params = {fn_id: None for fn_id in scoring_function_ids}
         return self.client.scoring.score(input_rows=[row], scoring_functions=scoring_params)
 
+
+llama_stack_api = LlamaStackApi()

llama_stack/distribution/ui/page/distribution/datasets.py

@@ -6,13 +6,13 @@
 
 import streamlit as st
 
-from llama_stack.distribution.ui.modules.api import LlamaStackApi
+from llama_stack.distribution.ui.modules.api import llama_stack_api
 
 
 def datasets():
     st.header("Datasets")
 
-    datasets_info = {d.identifier: d.to_dict() for d in LlamaStackApi().client.datasets.list()}
+    datasets_info = {d.identifier: d.to_dict() for d in llama_stack_api.client.datasets.list()}
     if len(datasets_info) > 0:
         selected_dataset = st.selectbox("Select a dataset", list(datasets_info.keys()))
         st.json(datasets_info[selected_dataset], expanded=True)

llama_stack/distribution/ui/page/distribution/eval_tasks.py

@@ -6,14 +6,14 @@
 
 import streamlit as st
 
-from llama_stack.distribution.ui.modules.api import LlamaStackApi
+from llama_stack.distribution.ui.modules.api import llama_stack_api
 
 
 def benchmarks():
     # Benchmarks Section
     st.header("Benchmarks")
 
-    benchmarks_info = {d.identifier: d.to_dict() for d in LlamaStackApi().client.benchmarks.list()}
+    benchmarks_info = {d.identifier: d.to_dict() for d in llama_stack_api.client.benchmarks.list()}
 
     if len(benchmarks_info) > 0:
         selected_benchmark = st.selectbox("Select an eval task", list(benchmarks_info.keys()), key="benchmark_inspect")

llama_stack/distribution/ui/page/distribution/models.py

@@ -6,13 +6,13 @@
 
 import streamlit as st
 
-from llama_stack.distribution.ui.modules.api import LlamaStackApi
+from llama_stack.distribution.ui.modules.api import llama_stack_api
 
 
 def models():
     # Models Section
     st.header("Models")
-    models_info = {m.identifier: m.to_dict() for m in LlamaStackApi().client.models.list()}
+    models_info = {m.identifier: m.to_dict() for m in llama_stack_api.client.models.list()}
 
     selected_model = st.selectbox("Select a model", list(models_info.keys()))
     st.json(models_info[selected_model])

llama_stack/distribution/ui/page/distribution/providers.py

@@ -6,12 +6,12 @@
 
 import streamlit as st
 
-from llama_stack.distribution.ui.modules.api import LlamaStackApi
+from llama_stack.distribution.ui.modules.api import llama_stack_api
 
 
 def providers():
     st.header("🔍 API Providers")
-    apis_providers_lst = LlamaStackApi().client.providers.list()
+    apis_providers_lst = llama_stack_api.client.providers.list()
     api_to_providers = {}
     for api_provider in apis_providers_lst:
         if api_provider.api in api_to_providers:

llama_stack/distribution/ui/page/distribution/scoring_functions.py

@@ -6,13 +6,13 @@
 
 import streamlit as st
 
-from llama_stack.distribution.ui.modules.api import LlamaStackApi
+from llama_stack.distribution.ui.modules.api import llama_stack_api
 
 
 def scoring_functions():
     st.header("Scoring Functions")
 
-    scoring_functions_info = {s.identifier: s.to_dict() for s in LlamaStackApi().client.scoring_functions.list()}
+    scoring_functions_info = {s.identifier: s.to_dict() for s in llama_stack_api.client.scoring_functions.list()}
 
     selected_scoring_function = st.selectbox("Select a scoring function", list(scoring_functions_info.keys()))
     st.json(scoring_functions_info[selected_scoring_function], expanded=True)

llama_stack/distribution/ui/page/distribution/shields.py

@@ -6,14 +6,14 @@
 
 import streamlit as st
 
-from llama_stack.distribution.ui.modules.api import LlamaStackApi
+from llama_stack.distribution.ui.modules.api import llama_stack_api
 
 
 def shields():
     # Shields Section
     st.header("Shields")
 
-    shields_info = {s.identifier: s.to_dict() for s in LlamaStackApi().client.shields.list()}
+    shields_info = {s.identifier: s.to_dict() for s in llama_stack_api.client.shields.list()}
 
     selected_shield = st.selectbox("Select a shield", list(shields_info.keys()))
     st.json(shields_info[selected_shield])

llama_stack/distribution/ui/page/distribution/vector_dbs.py

@@ -6,12 +6,12 @@
 
 import streamlit as st
 
-from llama_stack.distribution.ui.modules.api import LlamaStackApi
+from llama_stack.distribution.ui.modules.api import llama_stack_api
 
 
 def vector_dbs():
     st.header("Vector Databases")
-    vector_dbs_info = {v.identifier: v.to_dict() for v in LlamaStackApi().client.vector_dbs.list()}
+    vector_dbs_info = {v.identifier: v.to_dict() for v in llama_stack_api.client.vector_dbs.list()}
 
     if len(vector_dbs_info) > 0:
         selected_vector_db = st.selectbox("Select a vector database", list(vector_dbs_info.keys()))

llama_stack/distribution/ui/page/evaluations/app_eval.py

@@ -9,7 +9,7 @@ import json
 import pandas as pd
 import streamlit as st
 
-from llama_stack.distribution.ui.modules.api import LlamaStackApi
+from llama_stack.distribution.ui.modules.api import llama_stack_api
 from llama_stack.distribution.ui.modules.utils import process_dataset
 
 
@@ -39,7 +39,7 @@ def application_evaluation_page():
 
     # Select Scoring Functions to Run Evaluation On
     st.subheader("Select Scoring Functions")
-    scoring_functions = LlamaStackApi().client.scoring_functions.list()
+    scoring_functions = llama_stack_api.client.scoring_functions.list()
     scoring_functions = {sf.identifier: sf for sf in scoring_functions}
     scoring_functions_names = list(scoring_functions.keys())
     selected_scoring_functions = st.multiselect(
@@ -48,7 +48,7 @@ def application_evaluation_page():
         help="Choose one or more scoring functions.",
     )
 
-    available_models = LlamaStackApi().client.models.list()
+    available_models = llama_stack_api.client.models.list()
     available_models = [m.identifier for m in available_models]
 
     scoring_params = {}
@@ -108,7 +108,7 @@ def application_evaluation_page():
                 progress_bar.progress(progress, text=progress_text)
 
                 # Run evaluation for current row
-                score_res = LlamaStackApi().run_scoring(
+                score_res = llama_stack_api.run_scoring(
                     r,
                     scoring_function_ids=selected_scoring_functions,
                     scoring_params=scoring_params,

llama_stack/distribution/ui/page/evaluations/native_eval.py

@@ -9,13 +9,13 @@ import json
 import pandas as pd
 import streamlit as st
 
-from llama_stack.distribution.ui.modules.api import LlamaStackApi
+from llama_stack.distribution.ui.modules.api import llama_stack_api
 
 
 def select_benchmark_1():
     # Select Benchmarks
     st.subheader("1. Choose An Eval Task")
-    benchmarks = LlamaStackApi().client.benchmarks.list()
+    benchmarks = llama_stack_api.client.benchmarks.list()
     benchmarks = {et.identifier: et for et in benchmarks}
     benchmarks_names = list(benchmarks.keys())
     selected_benchmark = st.selectbox(
@@ -47,7 +47,7 @@ def define_eval_candidate_2():
         # Define Eval Candidate
         candidate_type = st.radio("Candidate Type", ["model", "agent"])
 
-        available_models = LlamaStackApi().client.models.list()
+        available_models = llama_stack_api.client.models.list()
         available_models = [model.identifier for model in available_models]
         selected_model = st.selectbox(
             "Choose a model",
@@ -167,7 +167,7 @@ def run_evaluation_3():
     eval_candidate = st.session_state["eval_candidate"]
 
     dataset_id = benchmarks[selected_benchmark].dataset_id
-    rows = LlamaStackApi().client.datasets.iterrows(
+    rows = llama_stack_api.client.datasets.iterrows(
         dataset_id=dataset_id,
     )
     total_rows = len(rows.data)
@@ -208,7 +208,7 @@ def run_evaluation_3():
             progress = i / len(rows)
             progress_bar.progress(progress, text=progress_text)
             # Run evaluation for current row
-            eval_res = LlamaStackApi().client.eval.evaluate_rows(
+            eval_res = llama_stack_api.client.eval.evaluate_rows(
                 benchmark_id=selected_benchmark,
                 input_rows=[r],
                 scoring_functions=benchmarks[selected_benchmark].scoring_functions,

llama_stack/distribution/ui/page/playground/chat.py

@@ -6,12 +6,12 @@
 
 import streamlit as st
 
-from llama_stack.distribution.ui.modules.api import LlamaStackApi
+from llama_stack.distribution.ui.modules.api import llama_stack_api
 
 # Sidebar configurations
 with st.sidebar:
     st.header("Configuration")
-    available_models = LlamaStackApi().client.models.list()
+    available_models = llama_stack_api.client.models.list()
     available_models = [model.identifier for model in available_models if model.model_type == "llm"]
     selected_model = st.selectbox(
         "Choose a model",
@@ -103,7 +103,7 @@ if prompt := st.chat_input("Example: What is Llama Stack?"):
         else:
             strategy = {"type": "greedy"}
 
-        response = LlamaStackApi().client.inference.chat_completion(
+        response = llama_stack_api.client.inference.chat_completion(
             messages=[
                 {"role": "system", "content": system_prompt},
                 {"role": "user", "content": prompt},

llama_stack/distribution/ui/page/playground/rag.py

@@ -10,7 +10,7 @@ import streamlit as st
 from llama_stack_client import Agent, AgentEventLogger, RAGDocument
 
 from llama_stack.apis.common.content_types import ToolCallDelta
-from llama_stack.distribution.ui.modules.api import LlamaStackApi
+from llama_stack.distribution.ui.modules.api import llama_stack_api
 from llama_stack.distribution.ui.modules.utils import data_url_from_file
 
 
@@ -57,14 +57,14 @@ def rag_chat_page():
                     for i, uploaded_file in enumerate(uploaded_files)
                 ]
 
-                providers = LlamaStackApi().client.providers.list()
+                providers = llama_stack_api.client.providers.list()
                 vector_io_provider = None
 
                 for x in providers:
                     if x.api == "vector_io":
                         vector_io_provider = x.provider_id
 
-                LlamaStackApi().client.vector_dbs.register(
+                llama_stack_api.client.vector_dbs.register(
                     vector_db_id=vector_db_name,  # Use the user-provided name
                     embedding_dimension=384,
                     embedding_model="all-MiniLM-L6-v2",
@@ -72,7 +72,7 @@ def rag_chat_page():
                 )
 
                 # insert documents using the custom vector db name
-                LlamaStackApi().client.tool_runtime.rag_tool.insert(
+                llama_stack_api.client.tool_runtime.rag_tool.insert(
                     vector_db_id=vector_db_name,  # Use the user-provided name
                     documents=documents,
                     chunk_size_in_tokens=512,
@@ -93,7 +93,7 @@ def rag_chat_page():
         )
 
         # select memory banks
-        vector_dbs = LlamaStackApi().client.vector_dbs.list()
+        vector_dbs = llama_stack_api.client.vector_dbs.list()
         vector_dbs = [vector_db.identifier for vector_db in vector_dbs]
         selected_vector_dbs = st.multiselect(
             label="Select Document Collections to use in RAG queries",
@@ -103,7 +103,7 @@ def rag_chat_page():
         )
 
         st.subheader("Inference Parameters", divider=True)
-        available_models = LlamaStackApi().client.models.list()
+        available_models = llama_stack_api.client.models.list()
         available_models = [model.identifier for model in available_models if model.model_type == "llm"]
         selected_model = st.selectbox(
             label="Choose a model",
@@ -167,7 +167,7 @@ def rag_chat_page():
     @st.cache_resource
     def create_agent():
         return Agent(
-            LlamaStackApi().client,
+            llama_stack_api.client,
             model=selected_model,
             instructions=system_prompt,
             sampling_params={
@@ -232,7 +232,7 @@ def rag_chat_page():
             st.session_state.messages.append({"role": "system", "content": system_prompt})
 
         # Query the vector DB
-        rag_response = LlamaStackApi().client.tool_runtime.rag_tool.query(
+        rag_response = llama_stack_api.client.tool_runtime.rag_tool.query(
             content=prompt, vector_db_ids=list(selected_vector_dbs)
         )
         prompt_context = rag_response.content
@@ -251,7 +251,7 @@ def rag_chat_page():
 
             # Run inference directly
             st.session_state.messages.append({"role": "user", "content": extended_prompt})
-            response = LlamaStackApi().client.inference.chat_completion(
+            response = llama_stack_api.client.inference.chat_completion(
                 messages=st.session_state.messages,
                 model_id=selected_model,
                 sampling_params={

llama_stack/distribution/ui/page/playground/tools.py

@@ -13,7 +13,7 @@ from llama_stack_client import Agent
 from llama_stack_client.lib.agents.react.agent import ReActAgent
 from llama_stack_client.lib.agents.react.tool_parser import ReActOutput
 
-from llama_stack.distribution.ui.modules.api import LlamaStackApi
+from llama_stack.distribution.ui.modules.api import llama_stack_api
 
 
 class AgentType(enum.Enum):
@@ -24,7 +24,7 @@ class AgentType(enum.Enum):
 def tool_chat_page():
     st.title("🛠 Tools")
 
-    client = LlamaStackApi().client
+    client = llama_stack_api.client
     models = client.models.list()
     model_list = [model.identifier for model in models if model.api_model_type == "llm"]
 
@@ -55,7 +55,7 @@ def tool_chat_page():
         )
 
         if "builtin::rag" in toolgroup_selection:
-            vector_dbs = LlamaStackApi().client.vector_dbs.list() or []
+            vector_dbs = llama_stack_api.client.vector_dbs.list() or []
             if not vector_dbs:
                 st.info("No vector databases available for selection.")
             vector_dbs = [vector_db.identifier for vector_db in vector_dbs]

llama_stack/distribution/ui/requirements.txt

@@ -1,5 +1,5 @@
-llama-stack-client>=0.2.9
+llama-stack>=0.2.1
+llama-stack-client>=0.2.1
 pandas
 streamlit
 streamlit-option-menu
-streamlit-keycloak

llama_stack/templates/kvant/run.yaml

@@ -154,17 +154,3 @@ tool_groups:
   provider_id: rag-runtime
 server:
   port: 8321
-  auth:
-    provider_type: "oauth2_token"
-    config:
-      jwks:
-      introspection:
-        url: ${env.KEYCLOAK_INSTROSPECT:https://iam.phoenix-systems.ch/realms/kvant/protocol/openid-connect/token/introspect} 
-        client_id: ${env.KEYCLOAK_CLIENT_ID:llama-stack}
-        client_secret: ${env.KEYCLOAK_CLIENT_SECRET}
-      claims_mapping:
-        sub: projects
-        scope: roles
-        #groups: teams
-        customer/id: teams
-        aud: namespaces

playground_start_local.sh

@@ -1,7 +1,3 @@
 #!/usr/bin/env bash
 
-export KEYCLOAK_URL="https://iam.phoenix-systems.ch"
-export KEYCLOAK_REALM="kvant"
-export KEYCLOAK_CLIENT_ID="llama-stack-playground"
-
 uv run --with ".[ui]" streamlit run llama_stack/distribution/ui/app.py

pyproject.toml

@@ -49,7 +49,6 @@ ui = [
     "pandas",
     "llama-stack-client>=0.2.9",
     "streamlit-option-menu",
-    "streamlit-keycloak",
 ]
 
 [dependency-groups]