From a96274c24b8d9fcc2639f0075e375ba536535111 Mon Sep 17 00:00:00 2001
From: Anton Dubovik <anton_dubovik@epam.com>
Date: Mon, 16 Sep 2024 10:33:23 +0000
Subject: [PATCH 01/16] feat: added docker-compose and test for llama3:8b via
 ollama

---
 .github/workflows/pr.yml                      |  5 ++
 .github/workflows/release.yml                 |  5 ++
 .gitignore                                    |  7 +-
 dial-docker-compose/addon/docker-compose.yml  |  2 +-
 dial-docker-compose/ci/ollama/.env            |  1 +
 .../ci/ollama/docker-compose.yml              | 17 +++++
 .../ci/ollama/test/.dockerignore              |  1 +
 dial-docker-compose/ci/ollama/test/Dockerfile |  7 ++
 dial-docker-compose/ci/ollama/test/app.py     | 66 +++++++++++++++++++
 .../ci/ollama/test/requirements.txt           |  2 +
 dial-docker-compose/common.yml                |  6 +-
 dial-docker-compose/model/docker-compose.yml  |  2 +-
 dial-docker-compose/ollama/.env               |  1 +
 dial-docker-compose/ollama/core/config.json   | 23 +++++++
 dial-docker-compose/ollama/docker-compose.yml | 30 +++++++++
 dial-docker-compose/ollama/ollama_setup.sh    | 26 ++++++++
 dial-docker-compose/settings/settings.json    |  4 +-
 17 files changed, 196 insertions(+), 9 deletions(-)
 create mode 100644 dial-docker-compose/ci/ollama/.env
 create mode 100644 dial-docker-compose/ci/ollama/docker-compose.yml
 create mode 100644 dial-docker-compose/ci/ollama/test/.dockerignore
 create mode 100644 dial-docker-compose/ci/ollama/test/Dockerfile
 create mode 100644 dial-docker-compose/ci/ollama/test/app.py
 create mode 100644 dial-docker-compose/ci/ollama/test/requirements.txt
 create mode 100644 dial-docker-compose/ollama/.env
 create mode 100644 dial-docker-compose/ollama/core/config.json
 create mode 100644 dial-docker-compose/ollama/docker-compose.yml
 create mode 100755 dial-docker-compose/ollama/ollama_setup.sh

diff --git a/.github/workflows/pr.yml b/.github/workflows/pr.yml
index 5a5fff39..06e5d3ce 100644
--- a/.github/workflows/pr.yml
+++ b/.github/workflows/pr.yml
@@ -38,6 +38,11 @@ jobs:
         with:
           cwd: "./dial-docker-compose/ci/addon"
           up-flags: "--abort-on-container-exit --exit-code-from test --timeout 300"
+      - uses: isbang/compose-action@802a148945af6399a338c7906c267331b39a71af # v2.0.0
+        name: Run quickstart ollama example
+        with:
+          cwd: "./dial-docker-compose/ci/ollama"
+          up-flags: "--abort-on-container-exit --exit-code-from test --timeout 300"
 
   build:
     needs: [run-notebooks, run-quickstart]
diff --git a/.github/workflows/release.yml b/.github/workflows/release.yml
index 38cdb0ba..9b7f475e 100644
--- a/.github/workflows/release.yml
+++ b/.github/workflows/release.yml
@@ -38,6 +38,11 @@ jobs:
         with:
           cwd: "./dial-docker-compose/ci/addon"
           up-flags: "--abort-on-container-exit --exit-code-from test --timeout 300"
+      - uses: isbang/compose-action@802a148945af6399a338c7906c267331b39a71af # v2.0.0
+        name: Run quickstart ollama example
+        with:
+          cwd: "./dial-docker-compose/ci/ollama"
+          up-flags: "--abort-on-container-exit --exit-code-from test --timeout 300"
 
   build-and-deploy:
     needs: [run-notebooks, run-quickstart]
diff --git a/.gitignore b/.gitignore
index 288b27ff..1c0d3cb8 100644
--- a/.gitignore
+++ b/.gitignore
@@ -23,8 +23,11 @@ yarn-error.log*
 .venv
 __pycache__
 
-# DIAL Core logs
-*.log
+# Docker container volumes
+core-data
+core-logs
+.ollama
+
 /.quarto/
 
 # Autogenerated files by Quarto
diff --git a/dial-docker-compose/addon/docker-compose.yml b/dial-docker-compose/addon/docker-compose.yml
index 9df5bd32..65c3715d 100644
--- a/dial-docker-compose/addon/docker-compose.yml
+++ b/dial-docker-compose/addon/docker-compose.yml
@@ -4,7 +4,7 @@ include:
 
 services:
   adapter-openai:
-    image: epam/ai-dial-adapter-openai:0.11.0
+    image: epam/ai-dial-adapter-openai:0.14.0
     environment:
       WEB_CONCURRENCY: "3"
 
diff --git a/dial-docker-compose/ci/ollama/.env b/dial-docker-compose/ci/ollama/.env
new file mode 100644
index 00000000..4daa1416
--- /dev/null
+++ b/dial-docker-compose/ci/ollama/.env
@@ -0,0 +1 @@
+DIAL_DIR="./ollama"
\ No newline at end of file
diff --git a/dial-docker-compose/ci/ollama/docker-compose.yml b/dial-docker-compose/ci/ollama/docker-compose.yml
new file mode 100644
index 00000000..b209b73f
--- /dev/null
+++ b/dial-docker-compose/ci/ollama/docker-compose.yml
@@ -0,0 +1,17 @@
+include:
+  - path: ../../ollama/docker-compose.yml
+    env_file: ./.env
+
+services:
+  test:
+    build: test
+    environment:
+      DIAL_URL: "http://core:8080"
+      DIAL_API_KEY: "dial_api_key"
+      DIAL_API_VERSION: "2024-02-01"
+      DIAL_DEPLOYMENT: "llama3:8b"
+    depends_on:
+      ollama-setup:
+        condition: service_healthy
+      core:
+        condition: service_healthy
\ No newline at end of file
diff --git a/dial-docker-compose/ci/ollama/test/.dockerignore b/dial-docker-compose/ci/ollama/test/.dockerignore
new file mode 100644
index 00000000..1d1fe94d
--- /dev/null
+++ b/dial-docker-compose/ci/ollama/test/.dockerignore
@@ -0,0 +1 @@
+Dockerfile
\ No newline at end of file
diff --git a/dial-docker-compose/ci/ollama/test/Dockerfile b/dial-docker-compose/ci/ollama/test/Dockerfile
new file mode 100644
index 00000000..c41fc702
--- /dev/null
+++ b/dial-docker-compose/ci/ollama/test/Dockerfile
@@ -0,0 +1,7 @@
+FROM python:3.11-alpine
+
+WORKDIR /app
+COPY * /app
+RUN pip install -r requirements.txt
+
+CMD ["python", "app.py"]
\ No newline at end of file
diff --git a/dial-docker-compose/ci/ollama/test/app.py b/dial-docker-compose/ci/ollama/test/app.py
new file mode 100644
index 00000000..95460af4
--- /dev/null
+++ b/dial-docker-compose/ci/ollama/test/app.py
@@ -0,0 +1,66 @@
+import os
+import aiohttp
+import asyncio
+import backoff
+
+import logging
+
+
+def get_env(name: str) -> str:
+    value = os.environ.get(name)
+    if value is None:
+        raise ValueError(f"'{name}' environment variable must be defined")
+    return value
+
+
+DIAL_URL = get_env("DIAL_URL")
+DIAL_API_KEY = get_env("DIAL_API_KEY")
+DIAL_API_VERSION = get_env("DIAL_API_VERSION")
+DIAL_DEPLOYMENT = get_env("DIAL_DEPLOYMENT")
+
+logging.basicConfig(level=logging.DEBUG)
+log = logging.getLogger(__name__)
+
+
+@backoff.on_exception(
+    backoff.expo,
+    (aiohttp.ClientError, aiohttp.ServerTimeoutError),
+    max_time=60,
+)
+async def post_with_retry(url: str, payload: dict, headers: dict, params: dict):
+    async with aiohttp.ClientSession() as session:
+        async with session.post(
+            url, json=payload, headers=headers, params=params
+        ) as response:
+            response.raise_for_status()
+            return await response.json()
+
+
+async def test_model(deployment_id: str):
+    api_url = f"{DIAL_URL}/openai/deployments/{deployment_id}/chat/completions"
+
+    message = "12 + 23 = ? Reply with a single number:"
+    payload = {
+        "model": deployment_id,
+        "messages": [{"role": "user", "content": message}],
+        "stream": False,
+    }
+    headers = {"api-key": DIAL_API_KEY}
+    params = {"api-version": DIAL_API_VERSION}
+
+    body = await post_with_retry(api_url, payload, headers, params)
+    log.debug(f"Response: {body}")
+
+    content = body.get("choices", [])[0].get("message", {}).get("content", "")
+
+    if "35" not in content:
+        raise ValueError(f"Test failed for {deployment_id!r}. ")
+
+
+async def tests():
+    await test_model(DIAL_DEPLOYMENT)
+
+
+if __name__ == "__main__":
+    loop = asyncio.get_event_loop()
+    loop.run_until_complete(tests())
diff --git a/dial-docker-compose/ci/ollama/test/requirements.txt b/dial-docker-compose/ci/ollama/test/requirements.txt
new file mode 100644
index 00000000..1c6d30a5
--- /dev/null
+++ b/dial-docker-compose/ci/ollama/test/requirements.txt
@@ -0,0 +1,2 @@
+aiohttp==3.9.4
+backoff==2.2.1
\ No newline at end of file
diff --git a/dial-docker-compose/common.yml b/dial-docker-compose/common.yml
index 6da6adbb..8e2ae74a 100644
--- a/dial-docker-compose/common.yml
+++ b/dial-docker-compose/common.yml
@@ -1,13 +1,13 @@
 services:
   themes:
-    image: epam/ai-dial-chat-themes:0.4.0
+    image: epam/ai-dial-chat-themes:0.6.0
     ports:
       - "3001:8080"
 
   chat:
     ports:
       - "3000:3000"
-    image: epam/ai-dial-chat:0.10.0
+    image: epam/ai-dial-chat:0.17.0
     depends_on:
       - themes
       - core
@@ -36,7 +36,7 @@ services:
     user: ${UID:-root}
     ports:
       - "8080:8080"
-    image: epam/ai-dial-core:0.9.0
+    image: epam/ai-dial-core:0.16.0
     environment:
       'AIDIAL_SETTINGS': '/opt/settings/settings.json'
       'JAVA_OPTS': '-Dgflog.config=/opt/settings/gflog.xml'
diff --git a/dial-docker-compose/model/docker-compose.yml b/dial-docker-compose/model/docker-compose.yml
index 1dd03ae0..a918cb85 100644
--- a/dial-docker-compose/model/docker-compose.yml
+++ b/dial-docker-compose/model/docker-compose.yml
@@ -4,6 +4,6 @@ include:
 
 services:
   adapter-openai:
-    image: epam/ai-dial-adapter-openai:0.11.0
+    image: epam/ai-dial-adapter-openai:0.14.0
     environment:
       WEB_CONCURRENCY: "3"
\ No newline at end of file
diff --git a/dial-docker-compose/ollama/.env b/dial-docker-compose/ollama/.env
new file mode 100644
index 00000000..4daa1416
--- /dev/null
+++ b/dial-docker-compose/ollama/.env
@@ -0,0 +1 @@
+DIAL_DIR="./ollama"
\ No newline at end of file
diff --git a/dial-docker-compose/ollama/core/config.json b/dial-docker-compose/ollama/core/config.json
new file mode 100644
index 00000000..10b0d035
--- /dev/null
+++ b/dial-docker-compose/ollama/core/config.json
@@ -0,0 +1,23 @@
+{
+  "routes": {},
+  "models": {
+    "llama3:8b": {
+      "type": "chat",
+      "displayName": "Llama3 8B (Locally hosted)",
+      "endpoint": "http://ollama:11434/v1/chat/completions"
+    }
+  },
+  "keys": {
+    "dial_api_key": {
+      "project": "TEST-PROJECT",
+      "role": "default"
+    }
+  },
+  "roles": {
+    "default": {
+      "limits": {
+        "llama3:8b": {}
+      }
+    }
+  }
+}
diff --git a/dial-docker-compose/ollama/docker-compose.yml b/dial-docker-compose/ollama/docker-compose.yml
new file mode 100644
index 00000000..418406d8
--- /dev/null
+++ b/dial-docker-compose/ollama/docker-compose.yml
@@ -0,0 +1,30 @@
+include:
+  - path: ../common.yml
+    env_file: ./.env
+
+services:
+  ollama:
+    image: ollama/ollama:0.3.10
+    volumes:
+      - ./.ollama:/root/.ollama
+    ports:
+      - "11434:11434"
+
+  ollama-setup:
+    depends_on:
+      ollama:
+        condition: service_started
+    image: alpine:3.20.3
+    environment:
+      - OLLAMA_URL=http://ollama:11434
+      - OLLAMA_ENABLED=${OLLAMA_ENABLED:-1}
+      - OLLAMA_MODEL=${OLLAMA_MODEL:-llama3:8b-instruct-q2_K}
+      - OLLAMA_MODEL_ALIAS=${OLLAMA_MODEL_ALIAS:-llama3:8b}
+    volumes:
+      - ./ollama_setup.sh:/setup.sh
+    command: sh /setup.sh
+    healthcheck:
+      test: ["CMD", "test", "-f", "/healthy"]
+      interval: 10s
+      start_period: 10s
+      retries: 10
diff --git a/dial-docker-compose/ollama/ollama_setup.sh b/dial-docker-compose/ollama/ollama_setup.sh
new file mode 100755
index 00000000..5bb8abb4
--- /dev/null
+++ b/dial-docker-compose/ollama/ollama_setup.sh
@@ -0,0 +1,26 @@
+#!/bin/sh
+set -e
+
+if [ "$OLLAMA_ENABLED" -eq 1 ]; then
+  apk add --no-cache curl
+
+  until curl -s ${OLLAMA_URL}; do
+    echo "Waiting for Ollama..."
+    sleep 5
+  done
+
+  echo "Pulling $OLLAMA_URL..."
+  curl -vL --fail-with-body "$OLLAMA_URL/api/pull" -d "{\"name\": \"$OLLAMA_MODEL\", \"stream\": false}"
+
+  echo "Making alias for $OLLAMA_URL: $OLLAMA_MODEL_ALIAS..."
+  curl -vL --fail-with-body "$OLLAMA_URL/api/copy" -d "{\"source\": \"$OLLAMA_MODEL\", \"destination\": \"$OLLAMA_MODEL_ALIAS\"}"
+
+  echo "Loading the model into memory..."
+  curl -vL --fail-with-body "$OLLAMA_URL/api/generate" -d "{\"model\": \"$OLLAMA_MODEL_ALIAS\"}"
+else
+  echo "Ollama is disabled"
+fi
+
+touch /healthy
+
+tail -f /dev/null
\ No newline at end of file
diff --git a/dial-docker-compose/settings/settings.json b/dial-docker-compose/settings/settings.json
index 446b1b6b..0471610d 100644
--- a/dial-docker-compose/settings/settings.json
+++ b/dial-docker-compose/settings/settings.json
@@ -9,7 +9,7 @@
     }
   },
   "encryption": {
-    "salt": "salt",
-    "password": "password"
+    "secret": "salt",
+    "key": "password"
   }
 }
\ No newline at end of file

From 79af610515ce1eaf19bc668024ed14d31859b3e2 Mon Sep 17 00:00:00 2001
From: Anton Dubovik <anton_dubovik@epam.com>
Date: Mon, 16 Sep 2024 12:16:02 +0000
Subject: [PATCH 02/16] feat: added quick-start tutorial for self-hosted models

---
 README.md                                     |  5 +--
 dial-docker-compose/ollama/core/config.json   |  6 ++--
 dial-docker-compose/ollama/docker-compose.yml |  2 --
 dial-docker-compose/ollama/ollama_setup.sh    | 28 ++++++++---------
 docs/README.md                                |  1 +
 .../quick-start-with-self-hosted-model.md     | 31 +++++++++++++++++++
 sidebars.js                                   |  9 ++++--
 7 files changed, 58 insertions(+), 24 deletions(-)
 create mode 100644 docs/tutorials/quick-start-with-self-hosted-model.md

diff --git a/README.md b/README.md
index c169e8f2..db8095ca 100644
--- a/README.md
+++ b/README.md
@@ -13,10 +13,11 @@
 ## Helm Deployment
 
 * [AI DIAL Generic Installation Simple Guide](https://github.com/epam/ai-dial-helm/tree/main/charts/dial/examples/generic/simple)
-  
+
 ## Tutorials
 
 * [Launch AI DIAL Chat with an Azure model](./docs/tutorials/quick-start-model.md)
+* [Launch AI DIAL Chat with a self-hosted model](./docs/tutorials/quick-start-with-self-hosted-model.md)
 * [Launch AI DIAL Chat with a Sample Application](./docs/tutorials/quick-start-with-application.md)
 * [Launch AI DIAL Chat with a Sample Addon](./docs/tutorials/quick-start-with-addon.md)
 
@@ -31,7 +32,7 @@
 ## Configuration
 
 * Refer to [Configuration](./docs/Deployment/configuration.md)
-  
+
 ## Other AI DIAL Project Open Source Repositories
 
 Here is the current list of repositories where you can find more details. You can also refer to [repository map](https://epam-rail.com/open-source).
diff --git a/dial-docker-compose/ollama/core/config.json b/dial-docker-compose/ollama/core/config.json
index 10b0d035..1a83c364 100644
--- a/dial-docker-compose/ollama/core/config.json
+++ b/dial-docker-compose/ollama/core/config.json
@@ -1,9 +1,9 @@
 {
   "routes": {},
   "models": {
-    "llama3:8b": {
+    "ollama-model": {
       "type": "chat",
-      "displayName": "Llama3 8B (Locally hosted)",
+      "displayName": "Self-hosted model",
       "endpoint": "http://ollama:11434/v1/chat/completions"
     }
   },
@@ -16,7 +16,7 @@
   "roles": {
     "default": {
       "limits": {
-        "llama3:8b": {}
+        "ollama-model": {}
       }
     }
   }
diff --git a/dial-docker-compose/ollama/docker-compose.yml b/dial-docker-compose/ollama/docker-compose.yml
index 418406d8..cbbc914e 100644
--- a/dial-docker-compose/ollama/docker-compose.yml
+++ b/dial-docker-compose/ollama/docker-compose.yml
@@ -17,9 +17,7 @@ services:
     image: alpine:3.20.3
     environment:
       - OLLAMA_URL=http://ollama:11434
-      - OLLAMA_ENABLED=${OLLAMA_ENABLED:-1}
       - OLLAMA_MODEL=${OLLAMA_MODEL:-llama3:8b-instruct-q2_K}
-      - OLLAMA_MODEL_ALIAS=${OLLAMA_MODEL_ALIAS:-llama3:8b}
     volumes:
       - ./ollama_setup.sh:/setup.sh
     command: sh /setup.sh
diff --git a/dial-docker-compose/ollama/ollama_setup.sh b/dial-docker-compose/ollama/ollama_setup.sh
index 5bb8abb4..d45c4c89 100755
--- a/dial-docker-compose/ollama/ollama_setup.sh
+++ b/dial-docker-compose/ollama/ollama_setup.sh
@@ -1,25 +1,23 @@
 #!/bin/sh
 set -e
 
-if [ "$OLLAMA_ENABLED" -eq 1 ]; then
-  apk add --no-cache curl
+apk add --no-cache curl
 
-  until curl -s ${OLLAMA_URL}; do
-    echo "Waiting for Ollama..."
-    sleep 5
-  done
+until curl -s ${OLLAMA_URL}; do
+  echo "Waiting for Ollama..."
+  sleep 5
+done
 
-  echo "Pulling $OLLAMA_URL..."
-  curl -vL --fail-with-body "$OLLAMA_URL/api/pull" -d "{\"name\": \"$OLLAMA_MODEL\", \"stream\": false}"
+echo "Pulling $OLLAMA_URL..."
+curl -vL --fail-with-body "$OLLAMA_URL/api/pull" -d "{\"name\": \"$OLLAMA_MODEL\", \"stream\": false}"
 
-  echo "Making alias for $OLLAMA_URL: $OLLAMA_MODEL_ALIAS..."
-  curl -vL --fail-with-body "$OLLAMA_URL/api/copy" -d "{\"source\": \"$OLLAMA_MODEL\", \"destination\": \"$OLLAMA_MODEL_ALIAS\"}"
+OLLAMA_MODEL_ALIAS=ollama-model
 
-  echo "Loading the model into memory..."
-  curl -vL --fail-with-body "$OLLAMA_URL/api/generate" -d "{\"model\": \"$OLLAMA_MODEL_ALIAS\"}"
-else
-  echo "Ollama is disabled"
-fi
+echo "Making alias for $OLLAMA_URL: $OLLAMA_MODEL_ALIAS..."
+curl -vL --fail-with-body "$OLLAMA_URL/api/copy" -d "{\"source\": \"$OLLAMA_MODEL\", \"destination\": \"$OLLAMA_MODEL_ALIAS\"}"
+
+echo "Loading the model into memory..."
+curl -vL --fail-with-body "$OLLAMA_URL/api/generate" -d "{\"model\": \"$OLLAMA_MODEL_ALIAS\"}"
 
 touch /healthy
 
diff --git a/docs/README.md b/docs/README.md
index 845f2859..de9d5580 100644
--- a/docs/README.md
+++ b/docs/README.md
@@ -21,6 +21,7 @@
 ## Tutorials
 
 * [Launch AI DIAL Chat with an Azure model](./tutorials/quick-start-model.md)
+* [Launch AI DIAL Chat with a self-hosted model](./tutorials/quick-start-with-self-hosted-model.md)
 * [Launch AI DIAL Chat with a Sample Application](./tutorials/quick-start-with-application.md)
 * [Launch AI DIAL Chat with a Sample Addon](./tutorials/quick-start-with-addon.md)
 
diff --git a/docs/tutorials/quick-start-with-self-hosted-model.md b/docs/tutorials/quick-start-with-self-hosted-model.md
new file mode 100644
index 00000000..098d2a4a
--- /dev/null
+++ b/docs/tutorials/quick-start-with-self-hosted-model.md
@@ -0,0 +1,31 @@
+# Launch AI DIAL Chat with a self-hosted model
+
+## Introduction
+
+In this tutorial, you will learn how to quickly launch AI DIAL Chat with a self-hosted model powered by [Ollama](https://ollama.com/).
+
+## Prerequisites
+
+Docker engine installed on your machine (Docker Compose Version 2.20.0 +).
+
+> Refer to [Docker](https://docs.docker.com/desktop/) documentation.
+
+## Step 1: Get AI DIAL
+
+[Download](https://github.com/epam/ai-dial/tree/main/dial-docker-compose/ollama/) AI DIAL.
+
+## Step 2: Launch AI DIAL Chat
+
+Run `docker compose up` from the folder with the [docker-compose file](https://github.com/epam/ai-dial/blob/main/dial-docker-compose/ollama/docker-compose.yml).
+
+By default a lightweight [llama3:8b-instruct-q2_K](https://ollama.com/library/llama3:8b-instruct-q2_K) model will be pulled and loaded into the memory of the Ollama server automatically.
+
+You could specify the model via the environment variable `OLLAMA_MODEL`:
+
+```sh
+OLLAMA_MODEL=model_of_your_choice docker compose up
+```
+
+Find the available models at the [Ollama model library](https://ollama.com/library).
+
+Finally, open http://localhost:3000/ in your browser to launch the AI DIAL Chat application and chat with the model.
\ No newline at end of file
diff --git a/sidebars.js b/sidebars.js
index 84d90389..9314a6c5 100644
--- a/sidebars.js
+++ b/sidebars.js
@@ -102,6 +102,11 @@ const sidebars = {
           id: 'tutorials/quick-start-model',
           label: 'Chat with OpenAI Model',
         },
+        {
+          type: 'doc',
+          id: 'tutorials/quick-start-with-self-hosted-model',
+          label: 'Chat with a self-hosted model',
+        },
         {
           type: 'doc',
           id: 'tutorials/quick-start-with-addon',
@@ -182,13 +187,13 @@ const sidebars = {
           Demos: [
             {
               type: 'autogenerated',
-              dirName: 'video demos/demos', 
+              dirName: 'video demos/demos',
             },
           ],
           'Demos For Developers': [
             {
               type: 'autogenerated',
-              dirName: 'video demos/demos-for-developers', 
+              dirName: 'video demos/demos-for-developers',
             },
           ],
         },

From cc717ff2595891aa1a5e24e95a537c052a992a6e Mon Sep 17 00:00:00 2001
From: Anton Dubovik <anton_dubovik@epam.com>
Date: Mon, 16 Sep 2024 12:19:40 +0000
Subject: [PATCH 03/16] chore: formatting fix

---
 .../quick-start-with-self-hosted-model.md      | 18 +++++++++---------
 1 file changed, 9 insertions(+), 9 deletions(-)

diff --git a/docs/tutorials/quick-start-with-self-hosted-model.md b/docs/tutorials/quick-start-with-self-hosted-model.md
index 098d2a4a..7c9bdce3 100644
--- a/docs/tutorials/quick-start-with-self-hosted-model.md
+++ b/docs/tutorials/quick-start-with-self-hosted-model.md
@@ -18,14 +18,14 @@ Docker engine installed on your machine (Docker Compose Version 2.20.0 +).
 
 Run `docker compose up` from the folder with the [docker-compose file](https://github.com/epam/ai-dial/blob/main/dial-docker-compose/ollama/docker-compose.yml).
 
-By default a lightweight [llama3:8b-instruct-q2_K](https://ollama.com/library/llama3:8b-instruct-q2_K) model will be pulled and loaded into the memory of the Ollama server automatically.
-
-You could specify the model via the environment variable `OLLAMA_MODEL`:
-
-```sh
-OLLAMA_MODEL=model_of_your_choice docker compose up
-```
-
-Find the available models at the [Ollama model library](https://ollama.com/library).
+> By default a lightweight [llama3:8b-instruct-q2_K](https://ollama.com/library/llama3:8b-instruct-q2_K) model will be pulled and loaded into the memory of the Ollama server automatically.
+>
+> You could specify the model via the environment variable `OLLAMA_MODEL`:
+>
+> ```sh
+> OLLAMA_MODEL=model_of_your_choice docker compose up
+> ```
+>
+> Find the available models at the [Ollama model library](https://ollama.com/library).
 
 Finally, open http://localhost:3000/ in your browser to launch the AI DIAL Chat application and chat with the model.
\ No newline at end of file

From a59961b9498e1da5a3fd959f6f81c624c8004e3a Mon Sep 17 00:00:00 2001
From: Anton Dubovik <anton_dubovik@epam.com>
Date: Mon, 16 Sep 2024 12:44:58 +0000
Subject: [PATCH 04/16] fix: fixed ci test for ollama model

---
 dial-docker-compose/ci/ollama/docker-compose.yml | 1 -
 dial-docker-compose/ci/ollama/test/app.py        | 3 +--
 2 files changed, 1 insertion(+), 3 deletions(-)

diff --git a/dial-docker-compose/ci/ollama/docker-compose.yml b/dial-docker-compose/ci/ollama/docker-compose.yml
index b209b73f..aff7e6cc 100644
--- a/dial-docker-compose/ci/ollama/docker-compose.yml
+++ b/dial-docker-compose/ci/ollama/docker-compose.yml
@@ -9,7 +9,6 @@ services:
       DIAL_URL: "http://core:8080"
       DIAL_API_KEY: "dial_api_key"
       DIAL_API_VERSION: "2024-02-01"
-      DIAL_DEPLOYMENT: "llama3:8b"
     depends_on:
       ollama-setup:
         condition: service_healthy
diff --git a/dial-docker-compose/ci/ollama/test/app.py b/dial-docker-compose/ci/ollama/test/app.py
index 95460af4..d3c133d1 100644
--- a/dial-docker-compose/ci/ollama/test/app.py
+++ b/dial-docker-compose/ci/ollama/test/app.py
@@ -16,7 +16,6 @@ def get_env(name: str) -> str:
 DIAL_URL = get_env("DIAL_URL")
 DIAL_API_KEY = get_env("DIAL_API_KEY")
 DIAL_API_VERSION = get_env("DIAL_API_VERSION")
-DIAL_DEPLOYMENT = get_env("DIAL_DEPLOYMENT")
 
 logging.basicConfig(level=logging.DEBUG)
 log = logging.getLogger(__name__)
@@ -58,7 +57,7 @@ async def test_model(deployment_id: str):
 
 
 async def tests():
-    await test_model(DIAL_DEPLOYMENT)
+    await test_model("ollama-model")
 
 
 if __name__ == "__main__":

From 20008a41a275395b303df60a7f5cc8af0f1646fc Mon Sep 17 00:00:00 2001
From: Anton Dubovik <anton_dubovik@epam.com>
Date: Mon, 16 Sep 2024 16:56:54 +0000
Subject: [PATCH 05/16] chore: bumped the default Ollama model to Llama 3.1

---
 dial-docker-compose/ollama/docker-compose.yml        | 2 +-
 docs/tutorials/quick-start-with-self-hosted-model.md | 2 +-
 2 files changed, 2 insertions(+), 2 deletions(-)

diff --git a/dial-docker-compose/ollama/docker-compose.yml b/dial-docker-compose/ollama/docker-compose.yml
index cbbc914e..7e1e8ab2 100644
--- a/dial-docker-compose/ollama/docker-compose.yml
+++ b/dial-docker-compose/ollama/docker-compose.yml
@@ -17,7 +17,7 @@ services:
     image: alpine:3.20.3
     environment:
       - OLLAMA_URL=http://ollama:11434
-      - OLLAMA_MODEL=${OLLAMA_MODEL:-llama3:8b-instruct-q2_K}
+      - OLLAMA_MODEL=${OLLAMA_MODEL:-llama3.1:8b-instruct-q4_K_M}
     volumes:
       - ./ollama_setup.sh:/setup.sh
     command: sh /setup.sh
diff --git a/docs/tutorials/quick-start-with-self-hosted-model.md b/docs/tutorials/quick-start-with-self-hosted-model.md
index 7c9bdce3..92282fb7 100644
--- a/docs/tutorials/quick-start-with-self-hosted-model.md
+++ b/docs/tutorials/quick-start-with-self-hosted-model.md
@@ -18,7 +18,7 @@ Docker engine installed on your machine (Docker Compose Version 2.20.0 +).
 
 Run `docker compose up` from the folder with the [docker-compose file](https://github.com/epam/ai-dial/blob/main/dial-docker-compose/ollama/docker-compose.yml).
 
-> By default a lightweight [llama3:8b-instruct-q2_K](https://ollama.com/library/llama3:8b-instruct-q2_K) model will be pulled and loaded into the memory of the Ollama server automatically.
+> By default [llama3.1:8b-instruct-q4_K_M](https://ollama.com/library/llama3.1:8b-instruct-q4_K_M) model will be pulled and loaded into the memory of the Ollama server automatically.
 >
 > You could specify the model via the environment variable `OLLAMA_MODEL`:
 >

From 626b1d634dcaffa821e0a28c7b068a08297b024d Mon Sep 17 00:00:00 2001
From: Anton Dubovik <anton_dubovik@epam.com>
Date: Tue, 17 Sep 2024 09:25:52 +0000
Subject: [PATCH 06/16] chore: fixed doc

---
 docs/tutorials/quick-start-with-self-hosted-model.md | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/docs/tutorials/quick-start-with-self-hosted-model.md b/docs/tutorials/quick-start-with-self-hosted-model.md
index 92282fb7..260b1dce 100644
--- a/docs/tutorials/quick-start-with-self-hosted-model.md
+++ b/docs/tutorials/quick-start-with-self-hosted-model.md
@@ -18,7 +18,7 @@ Docker engine installed on your machine (Docker Compose Version 2.20.0 +).
 
 Run `docker compose up` from the folder with the [docker-compose file](https://github.com/epam/ai-dial/blob/main/dial-docker-compose/ollama/docker-compose.yml).
 
-> By default [llama3.1:8b-instruct-q4_K_M](https://ollama.com/library/llama3.1:8b-instruct-q4_K_M) model will be pulled and loaded into the memory of the Ollama server automatically.
+> By default [llama3.1:8b-instruct-q4_K_M](https://ollama.com/library/llama3.1:8b-instruct-q4_K_M) model will be pulled and loaded into the memory of the Ollama server automatically. This is going to take around a minute on the first run.
 >
 > You could specify the model via the environment variable `OLLAMA_MODEL`:
 >

From c51418d48c94258cad30ecb0a723d604ce08cf1b Mon Sep 17 00:00:00 2001
From: Anton Dubovik <anton_dubovik@epam.com>
Date: Tue, 17 Sep 2024 17:25:00 +0000
Subject: [PATCH 07/16] feat: added instructions for self-hosted vision models

---
 dial-docker-compose/ci/ollama/test/app.py     |  59 ++++++++++--
 dial-docker-compose/ci/ollama/test/image.png  | Bin 0 -> 1976 bytes
 dial-docker-compose/ollama/core/config.json   |  24 ++++-
 dial-docker-compose/ollama/docker-compose.yml |  16 +++-
 dial-docker-compose/ollama/ollama_setup.sh    |  24 -----
 .../ollama/ollama_setup/Dockerfile            |   4 +
 .../ollama/ollama_setup/setup.py              |  87 ++++++++++++++++++
 .../quick-start-with-self-hosted-model.md     |  63 ++++++++++---
 8 files changed, 225 insertions(+), 52 deletions(-)
 create mode 100644 dial-docker-compose/ci/ollama/test/image.png
 delete mode 100755 dial-docker-compose/ollama/ollama_setup.sh
 create mode 100644 dial-docker-compose/ollama/ollama_setup/Dockerfile
 create mode 100755 dial-docker-compose/ollama/ollama_setup/setup.py

diff --git a/dial-docker-compose/ci/ollama/test/app.py b/dial-docker-compose/ci/ollama/test/app.py
index d3c133d1..6f805f43 100644
--- a/dial-docker-compose/ci/ollama/test/app.py
+++ b/dial-docker-compose/ci/ollama/test/app.py
@@ -1,9 +1,13 @@
+import base64
 import os
+from pathlib import Path
 import aiohttp
 import asyncio
 import backoff
 
 import logging
+import time
+from contextlib import asynccontextmanager
 
 
 def get_env(name: str) -> str:
@@ -21,6 +25,15 @@ def get_env(name: str) -> str:
 log = logging.getLogger(__name__)
 
 
+@asynccontextmanager
+async def timer(name: str):
+    log.debug(f"[{name}] Starting...")
+    start = time.perf_counter()
+    yield
+    elapsed = time.perf_counter() - start
+    log.debug(f"[{name}] Executed in {elapsed:.2f} seconds")
+
+
 @backoff.on_exception(
     backoff.expo,
     (aiohttp.ClientError, aiohttp.ServerTimeoutError),
@@ -35,13 +48,15 @@ async def post_with_retry(url: str, payload: dict, headers: dict, params: dict):
             return await response.json()
 
 
-async def test_model(deployment_id: str):
+def read_image_base64(png_file: Path) -> str:
+    return base64.b64encode(png_file.read_bytes()).decode("utf-8")
+
+async def dial_chat_completion(deployment_id: str, messages: list) -> str:
     api_url = f"{DIAL_URL}/openai/deployments/{deployment_id}/chat/completions"
 
-    message = "12 + 23 = ? Reply with a single number:"
     payload = {
         "model": deployment_id,
-        "messages": [{"role": "user", "content": message}],
+        "messages": messages,
         "stream": False,
     }
     headers = {"api-key": DIAL_API_KEY}
@@ -52,12 +67,44 @@ async def test_model(deployment_id: str):
 
     content = body.get("choices", [])[0].get("message", {}).get("content", "")
 
-    if "35" not in content:
-        raise ValueError(f"Test failed for {deployment_id!r}. ")
+    log.debug(f"Content: {content}")
+
+    return content
+
+async def test_chat_model(deployment_id: str):
+    message = "2 + 3 = ? Reply with a single number:"
+    messages = [{"role": "user", "content": message}]
+    content = await dial_chat_completion(deployment_id, messages)
+
+    if "5" not in content:
+        raise ValueError(f"Test failed for {deployment_id!r}")
 
 
+async def test_vision_model(deployment_id: str):
+    base64_data = read_image_base64(Path("./image.png"))
+    base64_image = f"data:image/png;base64,{base64_data}"
+
+    messages = [
+        {
+            "role": "user",
+            "content": [
+                {"type": "text", "text": "Describe the image"},
+                {"type": "image_url", "image_url": {"url": base64_image}},
+            ],
+        }
+    ]
+
+    content = await dial_chat_completion(deployment_id, messages)
+
+    if "vision" not in content.lower():
+        raise ValueError(f"Test failed for {deployment_id!r}")
+
 async def tests():
-    await test_model("ollama-model")
+    async with timer("Testing chat-model"):
+        await test_chat_model("chat-model")
+
+    async with timer("Testing vision-model"):
+        await test_vision_model("vision-model")
 
 
 if __name__ == "__main__":
diff --git a/dial-docker-compose/ci/ollama/test/image.png b/dial-docker-compose/ci/ollama/test/image.png
new file mode 100644
index 0000000000000000000000000000000000000000..bfbcfc7ed9307e4b17e48afc36567949da21c92b
GIT binary patch
literal 1976
zcmV;p2S@mcP)<h;3K|Lk000e1NJLTq0077U001Zm0ssI2G=3sW000MnNkl<Zc%1E9
zSxggM7(E3_6^ks2N>o5Zh(U2dqlUP~xS=r`)VNh}NsK(WMvcC>CTffi8Xt)90r!AL
zj2fe2lt>5|t3ecT!7WjuAVMn$sC@qe_wQ}T>7ACJ+M)lP%){6FoqOi?&gs1~cW!k=
z1PvNG8h+kxfCk@$W-JXJfo3cX9)V^o4IY7JEDaukW-JXJfo3cX9)V^o4IY7JEDat(
zfVUgq0kpK>^l1QCz8ryp>ecjZkndssd;$od1q)QI>C@1W@4YKD8qv_e0;^TMnm!F*
zyvLe0jRn)Et6I~i;fo_yb2Cnzf=-9!%V996SJS8Ai^kHR0{G%os6hqr#pMPU@%b}~
zi}C(FtX2dEBPt5x#-Vrb4p4sc24*v=t5H{n&`?A~AUPSavF(wnuSZc4UcZLLg5Y2b
z9Efq_5FXwEjr26U{8TDIN(y_b@7k3t7G?gPJ&do@(c8CD5oh17tc+4q$<L3|Y$CH~
z)3axyh6#GCsfi9Brnoq%Y}*kRM~4qnQ<Kx`HfBl+5z)twv|<Gr3{suIKr)%=^JnGG
z@hVhitdl2s4`yd8^Ai&pdCC-CC^kV(4(avs*S_uO(uFQxmez8`tOEv!WZRAb1L)~f
zc?El8efU5jAtIeVedxst<^J&=RAwxzm5fH#cQa-vke@u^9ew5uFBh1gyqvqrV4y8q
z=*A5yE~c|*Y04B{%V40#j~yzCjrH&$g@ri`PJ8##ty^R^)2&;ycdx@R78XViAIgZ>
z8!J7Xji8GcQ(hjK&2;xJ?b$=UdU0Apf@u8nHdJOTBFe~M9TyNlU%rT#J9aQ;NC-7H
z@^XO*OeUVn&6Q@iZ|B(w6C6?kO^bzMV|jMdCTeb$a+;fI(<aV|jU}s9%CTi5BgxN?
z3<kP>UCODdau^r0vIO1ZU8v7k_wVzLJAPcm95|5C)~<CZ7nq=5KYsE4#IIj8crdT~
z{JBe-8#eG%W~QjYnVFoqVS_X)ne0xSkQcdlk#nX`7oE<lP@l1gC_bJ~3eHec!u$8Z
z1BaBr1h#oFIaywI_ip*|z?o)kEp_Y0%41@vp+WS$8X71%nvuJ8quN@Bti7v}lBl&+
zUZkms0s|SXZ(q^ryb4vWV`XHppr{BRKFD(}TwuYFAs9W{RpqcS7Cd`~FJEK`J9olr
zWk-6tvS3aQdxr=hGZR5UB63g=)~#c4Lj!VhoH9)&=ydWag8|XeEdKUQxrJVWDvh;b
z1^oR1fC%T$%Mgf=o6F*rE8S3@JedVeO_(zWuU<(B{{Ao;+2P9en0xnlJb%72KP`>N
z1qDuN;y04ryR)FBMfr^0g13iECrC|Yqy4a9GTDLxmeuQJZtw+D;nSzor3=sc`_tUH
zbnYD0)F`yanPx%)E9}w3)vp#F&L8XSVbc!?pq3Vq!q~BlsnfZ7(!UN>j##i8he}IP
zUM|hr=D|605E0>qY+@p=UPaHI?Bnm>xN-$cmY{cUj31AqN8O%Ws;gNL8R=@3K7CkF
zQzK94-yeFtNTF>w2cY8jylOR;y<;S^;jdq~c#*|!&4X>2Iu+&R*tii!BQN~@8-;~f
zzaD-2V*7S{|L&@q#lix6H8FyNSx{fEK<La+tFiQYEM3ZCiLvh7VdF$(BxcNLk9b4`
zjvPT%6|%FDmd0;c0Q~rYg9nkEjCb!`lCsrx-HWz;f*~RD1o0C<CxKdx1z^=G7F1T^
z!2@1yn+HuMJ}I;VMk7on<mRHL1~+daEe(GE_QuPXm^BMOe=3XEX4OxhTs?19m9(w|
z|K&@m1}ZkzkRcc~ipA&7v7osbd3h{u$2=$ty&lu1AvYIAMSOR|^XE8oM!D{Y5iF>!
zMP;Qjzp|3w#@XunB2;WF*jXnh2d%9DaN`E*>Hr`m1@ZCjsXcy-!a@`lqNqroFlG$0
zviQgRd}Yl^lX!gTk}}_RcL-qOL}h;Gg|{C|g{Z5eu3g!qb$&h((V|7{<H?gwYdKpQ
zm^Y7?zj`HKx3Q6PT-h=4^(zGhvGV9><(2oSC`JwnqOV^avNnCe`v!4i)C6^pSRo;p
zH;={V&!e#s*RBD8(TIf$-Bz1AmB;oYlRau{dF;x5y6|v*FZljFwr&-Xw{GPx5>~J7
z9Q*a!4R1SJ3HkZ#rx_Yb+1V`4$PiU=_Jgjkr|#YP%Icqc2U=Tc%^F_!*fC|zHrbdM
zp53xVes{y>&72bxL-qAij&F@v>TMZAYb(EOw%thg?~AGk`rUpbp`RZuTSk{Ik=aZI
z1(cOV$;mt$73Fa4#n-Z=#9?3Fph2{MKi$4fW;5NsP5bsa?8gcXrIHdE5zSbC8@qNn
zy!nrhSKy2L{lEd4P05m@Uq31>bwiVgN=y091;G&=EmgL`w~VFkd#pC3r}Gsi*zJ92
zAGU2naWQ7jlwU6D)eGCUp{xu^N$$xeC844MhYmUY(?o164jn>8h1<%V6aVA?6;@ZH
zqy%r@!eT*SAR;0#bSQ=o?|}8m*RS#Dk;A`qMvg>WTt_sz)8HR%fN8`vOX}MI0000<
KMNUMnLSTX<hqe>|

literal 0
HcmV?d00001

diff --git a/dial-docker-compose/ollama/core/config.json b/dial-docker-compose/ollama/core/config.json
index 1a83c364..4a55bd1e 100644
--- a/dial-docker-compose/ollama/core/config.json
+++ b/dial-docker-compose/ollama/core/config.json
@@ -1,10 +1,25 @@
 {
   "routes": {},
   "models": {
-    "ollama-model": {
+    "chat-model": {
       "type": "chat",
-      "displayName": "Self-hosted model",
+      "displayName": "Self-hosted chat model",
       "endpoint": "http://ollama:11434/v1/chat/completions"
+    },
+    "vision-model": {
+      "type": "chat",
+      "displayName": "Self-hosted vision model",
+      "endpoint": "http://adapter-openai:5000/openai/deployments/vision-model/chat/completions",
+      "inputAttachmentTypes": [
+        "image/png",
+        "image/jpeg"
+      ],
+      "upstreams": [
+        {
+          "endpoint": "http://ollama:11434/v1/chat/completions",
+          "key": "dummy-key"
+        }
+      ]
     }
   },
   "keys": {
@@ -16,8 +31,9 @@
   "roles": {
     "default": {
       "limits": {
-        "ollama-model": {}
+        "chat-model": {},
+        "vision-model": {}
       }
     }
   }
-}
+}
\ No newline at end of file
diff --git a/dial-docker-compose/ollama/docker-compose.yml b/dial-docker-compose/ollama/docker-compose.yml
index 7e1e8ab2..9712aa6a 100644
--- a/dial-docker-compose/ollama/docker-compose.yml
+++ b/dial-docker-compose/ollama/docker-compose.yml
@@ -14,15 +14,21 @@ services:
     depends_on:
       ollama:
         condition: service_started
-    image: alpine:3.20.3
+    build: ./ollama_setup
     environment:
       - OLLAMA_URL=http://ollama:11434
-      - OLLAMA_MODEL=${OLLAMA_MODEL:-llama3.1:8b-instruct-q4_K_M}
-    volumes:
-      - ./ollama_setup.sh:/setup.sh
-    command: sh /setup.sh
+      - OLLAMA_CHAT_MODEL=${OLLAMA_CHAT_MODEL}
+      - OLLAMA_VISION_MODEL=${OLLAMA_VISION_MODEL}
     healthcheck:
       test: ["CMD", "test", "-f", "/healthy"]
       interval: 10s
       start_period: 10s
       retries: 10
+
+  adapter-openai:
+    image: epam/ai-dial-adapter-openai:0.14.0
+    environment:
+      WEB_CONCURRENCY: "3"
+      DIAL_URL: "http://core:8080"
+      DIAL_USE_FILE_STORAGE: "True"
+      GPT4_VISION_DEPLOYMENTS: "vision-model"
\ No newline at end of file
diff --git a/dial-docker-compose/ollama/ollama_setup.sh b/dial-docker-compose/ollama/ollama_setup.sh
deleted file mode 100755
index d45c4c89..00000000
--- a/dial-docker-compose/ollama/ollama_setup.sh
+++ /dev/null
@@ -1,24 +0,0 @@
-#!/bin/sh
-set -e
-
-apk add --no-cache curl
-
-until curl -s ${OLLAMA_URL}; do
-  echo "Waiting for Ollama..."
-  sleep 5
-done
-
-echo "Pulling $OLLAMA_URL..."
-curl -vL --fail-with-body "$OLLAMA_URL/api/pull" -d "{\"name\": \"$OLLAMA_MODEL\", \"stream\": false}"
-
-OLLAMA_MODEL_ALIAS=ollama-model
-
-echo "Making alias for $OLLAMA_URL: $OLLAMA_MODEL_ALIAS..."
-curl -vL --fail-with-body "$OLLAMA_URL/api/copy" -d "{\"source\": \"$OLLAMA_MODEL\", \"destination\": \"$OLLAMA_MODEL_ALIAS\"}"
-
-echo "Loading the model into memory..."
-curl -vL --fail-with-body "$OLLAMA_URL/api/generate" -d "{\"model\": \"$OLLAMA_MODEL_ALIAS\"}"
-
-touch /healthy
-
-tail -f /dev/null
\ No newline at end of file
diff --git a/dial-docker-compose/ollama/ollama_setup/Dockerfile b/dial-docker-compose/ollama/ollama_setup/Dockerfile
new file mode 100644
index 00000000..3afdb789
--- /dev/null
+++ b/dial-docker-compose/ollama/ollama_setup/Dockerfile
@@ -0,0 +1,4 @@
+FROM python:3.11-alpine
+RUN pip install requests
+COPY setup.py /setup.py
+CMD ["sh", "-c", "python /setup.py && tail -f /dev/null"]
diff --git a/dial-docker-compose/ollama/ollama_setup/setup.py b/dial-docker-compose/ollama/ollama_setup/setup.py
new file mode 100755
index 00000000..85e26858
--- /dev/null
+++ b/dial-docker-compose/ollama/ollama_setup/setup.py
@@ -0,0 +1,87 @@
+import asyncio
+from contextlib import asynccontextmanager
+import logging
+import os
+import time
+import requests
+
+OLLAMA_URL = os.getenv("OLLAMA_URL")
+if OLLAMA_URL is None:
+    raise RuntimeError("OLLAMA_URL env var isn't set")
+
+OLLAMA_CHAT_MODEL = os.getenv("OLLAMA_CHAT_MODEL")
+OLLAMA_VISION_MODEL = os.getenv("OLLAMA_VISION_MODEL")
+
+HEALTH_FILE = "/healthy"
+
+log = logging.getLogger(__name__)
+log.setLevel(logging.INFO)
+
+log.info(f"OLLAMA_URL = {OLLAMA_URL}")
+log.info(f"OLLAMA_CHAT_MODEL = {OLLAMA_CHAT_MODEL}")
+log.info(f"OLLAMA_VISION_MODEL = {OLLAMA_VISION_MODEL}")
+
+
+@asynccontextmanager
+async def timer(name: str):
+    log.info(f"[{name}] Starting...")
+    start = time.perf_counter()
+    yield
+    elapsed = time.perf_counter() - start
+    log.info(f"[{name}] Executed in {elapsed:.2f} seconds")
+
+
+async def wait_for_ollama():
+    while True:
+        try:
+            if requests.get(OLLAMA_URL).ok:
+                break
+        except requests.RequestException:
+            pass
+        await asyncio.sleep(1)
+
+
+async def pull_model(model):
+    data = {"name": model, "stream": False}
+    requests.post(f"{OLLAMA_URL}/api/pull", json=data).raise_for_status()
+
+
+async def create_alias(source, dest):
+    data = {"source": source, "destination": dest}
+    requests.post(f"{OLLAMA_URL}/api/copy", json=data).raise_for_status()
+
+
+async def load_model(model):
+    data = {"model": model}
+    requests.post(f"{OLLAMA_URL}/api/generate", json=data).raise_for_status()
+
+
+async def mark_as_healthy():
+    open(HEALTH_FILE, "w").close()
+
+
+async def main():
+    async with timer("Waiting for Ollama to start"):
+        await wait_for_ollama()
+
+    for model, alias in [
+        (OLLAMA_CHAT_MODEL, "chat-model"),
+        (OLLAMA_VISION_MODEL, "vision-model"),
+    ]:
+        if model:
+            async with timer(f"Pulling model {model}"):
+                await pull_model(model)
+
+            async with timer(f"Creating alias for {model}: {alias}"):
+                await create_alias(model, alias)
+
+    if model_to_load := OLLAMA_CHAT_MODEL or OLLAMA_VISION_MODEL:
+        async with timer(f"Loading model {model_to_load} into memory"):
+            await load_model(model_to_load)
+
+    await mark_as_healthy()
+
+
+if __name__ == "__main__":
+    loop = asyncio.get_event_loop()
+    loop.run_until_complete(main())
diff --git a/docs/tutorials/quick-start-with-self-hosted-model.md b/docs/tutorials/quick-start-with-self-hosted-model.md
index 260b1dce..c850eec3 100644
--- a/docs/tutorials/quick-start-with-self-hosted-model.md
+++ b/docs/tutorials/quick-start-with-self-hosted-model.md
@@ -12,20 +12,57 @@ Docker engine installed on your machine (Docker Compose Version 2.20.0 +).
 
 ## Step 1: Get AI DIAL
 
-[Download](https://github.com/epam/ai-dial/tree/main/dial-docker-compose/ollama/) AI DIAL.
+Clone [the repository](https://github.com/epam/ai-dial/) with the tutorials and change directory to the following folder:
 
-## Step 2: Launch AI DIAL Chat
+```sh
+cd dial-docker-compose/ollama
+```
 
-Run `docker compose up` from the folder with the [docker-compose file](https://github.com/epam/ai-dial/blob/main/dial-docker-compose/ollama/docker-compose.yml).
+## Step 2: Choose a model to run
 
-> By default [llama3.1:8b-instruct-q4_K_M](https://ollama.com/library/llama3.1:8b-instruct-q4_K_M) model will be pulled and loaded into the memory of the Ollama server automatically. This is going to take around a minute on the first run.
->
-> You could specify the model via the environment variable `OLLAMA_MODEL`:
->
-> ```sh
-> OLLAMA_MODEL=model_of_your_choice docker compose up
-> ```
->
-> Find the available models at the [Ollama model library](https://ollama.com/library).
+Ollama supports a wide range of popular open-source models.
 
-Finally, open http://localhost:3000/ in your browser to launch the AI DIAL Chat application and chat with the model.
\ No newline at end of file
+Consider first the modality your interested in - is a regular text-to-text chat model or a multi-modal vision model?
+
+Follow the feature tags at https://ollama.com/search to find an appropriate model, e.g. one working with `Code`, supporting `Tools` or a `Vision` feature.
+
+We recommend to choose one of the following models which have been tested:
+
+|Model|Vision|Tools|Streaming|
+|----|----|----|----|
+|[llama3.1:8b-instruct-q4_0](https://ollama.com/library/llama3.1:8b-instruct-q4_0)|❌|✅ *(only in non-streaming mode)*|✅|
+|[mistral:7b-instruct-q4_0](https://ollama.com/library/mistral:7b-instruct-q4_0)|❌|❌|✅|
+|[phi3.5:3.8b-mini-instruct-q4_0](https://ollama.com/library/phi3.5:3.8b-mini-instruct-q4_0)|❌|❌|✅|
+|[gemma2:2b-instruct-q4_0](https://ollama.com/library/gemma2:2b-instruct-q4_0)|❌|❌|✅|
+|[llava:7b-v1.6-mistral-q4_0](https://ollama.com/library/llava:7b-v1.6-mistral-q4_0)|✅|❌|❌|
+|[llava-phi3:3.8b-mini-q4_0](https://ollama.com/library/llava-phi3:3.8b-mini-q4_0)|✅|❌|❌|
+
+All the listed models support streaming.
+
+## Step 3: Launch AI DIAL Chat
+
+### Chat model
+
+If you have chosen a regular chat model _(e.g. llama3.1:8b-instruct-q4_0)_, then run the command:
+
+```sh
+OLLAMA_CHAT_MODEL=model_of_your_choice docker compose up --abort-on-container-exit
+```
+
+The model will be pulled and loaded into the memory of the Ollama server automatically. This may take a minute on the first run.
+
+Finally, open http://localhost:3000/ in your browser to launch the AI DIAL Chat application and select `Self-hosted chat model` deployment to converse with the model.
+
+### Vision model
+
+If you have chosen a vision model _(e.g. llava-phi3:3.8b-mini-q4_0)_, then run the command:
+
+```sh
+OLLAMA_VISION_MODEL=model_of_your_choice docker compose up --abort-on-container-exit
+```
+
+The model will be pulled and loaded into the memory of the Ollama server automatically. This may take a minute on the first run.
+
+Finally, open http://localhost:3000/ in your browser to launch the AI DIAL Chat application and select `Self-hosted vision model` deployment to converse with the model.
+
+> Note, that the vision models we tested, do not support streaming of response. Moreover, they are typically more computationally expensive than the chat models. So it may take minutes for a vision model to respond.
\ No newline at end of file

From 9003333e1cb224921c35ea715e3f8dfa5ee871de Mon Sep 17 00:00:00 2001
From: Anton Dubovik <anton_dubovik@epam.com>
Date: Tue, 17 Sep 2024 17:28:16 +0000
Subject: [PATCH 08/16] fix: updated env for ollama ci test

---
 dial-docker-compose/ci/ollama/.env | 4 +++-
 1 file changed, 3 insertions(+), 1 deletion(-)

diff --git a/dial-docker-compose/ci/ollama/.env b/dial-docker-compose/ci/ollama/.env
index 4daa1416..58a1c316 100644
--- a/dial-docker-compose/ci/ollama/.env
+++ b/dial-docker-compose/ci/ollama/.env
@@ -1 +1,3 @@
-DIAL_DIR="./ollama"
\ No newline at end of file
+DIAL_DIR="./ollama"
+OLLAMA_CHAT_MODEL=llama3.1:8b-instruct-q4_0
+OLLAMA_VISION_MODEL=llava-phi3:3.8b-mini-q4_0
\ No newline at end of file

From 4808b0958dc669ce7e18ba9c07d331445781780f Mon Sep 17 00:00:00 2001
From: Anton Dubovik <anton_dubovik@epam.com>
Date: Wed, 18 Sep 2024 08:33:01 +0000
Subject: [PATCH 09/16] chore: divided a single quickstart ci job into separate
 jobs

---
 .github/workflows/pr.yml      | 32 ++++++++++++++++++++++++++------
 .github/workflows/release.yml | 33 +++++++++++++++++++++++++++------
 2 files changed, 53 insertions(+), 12 deletions(-)

diff --git a/.github/workflows/pr.yml b/.github/workflows/pr.yml
index 06e5d3ce..0605935f 100644
--- a/.github/workflows/pr.yml
+++ b/.github/workflows/pr.yml
@@ -18,8 +18,8 @@ jobs:
           cwd: "./dial-cookbook/ci"
           up-flags: "--abort-on-container-exit --exit-code-from test --timeout 300"
 
-  run-quickstart:
-    name: Run quickstart examples
+  run-quickstart-model:
+    name: Run quickstart model example
     runs-on: ubuntu-latest
     steps:
       - uses: actions/checkout@b4ffde65f46336ab88eb53be808477a3936bae11 # v4.1.1
@@ -28,24 +28,44 @@ jobs:
         with:
           cwd: "./dial-docker-compose/ci/model"
           up-flags: "--abort-on-container-exit --exit-code-from test --timeout 300"
+
+  run-quickstart-application:
+    name: Run quickstart application example
+    runs-on: ubuntu-latest
+    steps:
+      - uses: actions/checkout@b4ffde65f46336ab88eb53be808477a3936bae11 # v4.1.1
       - uses: isbang/compose-action@802a148945af6399a338c7906c267331b39a71af # v2.0.0
-        name: Run quickstart application example
         with:
           cwd: "./dial-docker-compose/ci/application"
           up-flags: "--abort-on-container-exit --exit-code-from test --timeout 300"
+
+  run-quickstart-addon:
+    name: Run quickstart addon example
+    runs-on: ubuntu-latest
+    steps:
+      - uses: actions/checkout@b4ffde65f46336ab88eb53be808477a3936bae11 # v4.1.1
       - uses: isbang/compose-action@802a148945af6399a338c7906c267331b39a71af # v2.0.0
-        name: Run quickstart addon example
         with:
           cwd: "./dial-docker-compose/ci/addon"
           up-flags: "--abort-on-container-exit --exit-code-from test --timeout 300"
+
+  run-quickstart-self-hosted-model:
+    name: Run quickstart self-hosted model example
+    runs-on: ubuntu-latest
+    steps:
+      - uses: actions/checkout@b4ffde65f46336ab88eb53be808477a3936bae11 # v4.1.1
       - uses: isbang/compose-action@802a148945af6399a338c7906c267331b39a71af # v2.0.0
-        name: Run quickstart ollama example
         with:
           cwd: "./dial-docker-compose/ci/ollama"
           up-flags: "--abort-on-container-exit --exit-code-from test --timeout 300"
 
   build:
-    needs: [run-notebooks, run-quickstart]
+    needs:
+      - run-notebooks
+      - run-quickstart-model
+      - run-quickstart-application
+      - run-quickstart-addon
+      - run-quickstart-self-hosted-model
     runs-on: ubuntu-latest
     steps:
       - uses: actions/checkout@b4ffde65f46336ab88eb53be808477a3936bae11 # v4.1.1
diff --git a/.github/workflows/release.yml b/.github/workflows/release.yml
index 9b7f475e..a6677137 100644
--- a/.github/workflows/release.yml
+++ b/.github/workflows/release.yml
@@ -18,8 +18,9 @@ jobs:
           cwd: "./dial-cookbook/ci"
           up-flags: "--abort-on-container-exit --exit-code-from test --timeout 300"
 
-  run-quickstart:
-    name: Run quickstart examples
+
+  run-quickstart-model:
+    name: Run quickstart model example
     runs-on: ubuntu-latest
     steps:
       - uses: actions/checkout@b4ffde65f46336ab88eb53be808477a3936bae11 # v4.1.1
@@ -28,24 +29,44 @@ jobs:
         with:
           cwd: "./dial-docker-compose/ci/model"
           up-flags: "--abort-on-container-exit --exit-code-from test --timeout 300"
+
+  run-quickstart-application:
+    name: Run quickstart application example
+    runs-on: ubuntu-latest
+    steps:
+      - uses: actions/checkout@b4ffde65f46336ab88eb53be808477a3936bae11 # v4.1.1
       - uses: isbang/compose-action@802a148945af6399a338c7906c267331b39a71af # v2.0.0
-        name: Run quickstart application example
         with:
           cwd: "./dial-docker-compose/ci/application"
           up-flags: "--abort-on-container-exit --exit-code-from test --timeout 300"
+
+  run-quickstart-addon:
+    name: Run quickstart addon example
+    runs-on: ubuntu-latest
+    steps:
+      - uses: actions/checkout@b4ffde65f46336ab88eb53be808477a3936bae11 # v4.1.1
       - uses: isbang/compose-action@802a148945af6399a338c7906c267331b39a71af # v2.0.0
-        name: Run quickstart addon example
         with:
           cwd: "./dial-docker-compose/ci/addon"
           up-flags: "--abort-on-container-exit --exit-code-from test --timeout 300"
+
+  run-quickstart-self-hosted-model:
+    name: Run quickstart self-hosted model example
+    runs-on: ubuntu-latest
+    steps:
+      - uses: actions/checkout@b4ffde65f46336ab88eb53be808477a3936bae11 # v4.1.1
       - uses: isbang/compose-action@802a148945af6399a338c7906c267331b39a71af # v2.0.0
-        name: Run quickstart ollama example
         with:
           cwd: "./dial-docker-compose/ci/ollama"
           up-flags: "--abort-on-container-exit --exit-code-from test --timeout 300"
 
   build-and-deploy:
-    needs: [run-notebooks, run-quickstart]
+    needs:
+      - run-notebooks
+      - run-quickstart-model
+      - run-quickstart-application
+      - run-quickstart-addon
+      - run-quickstart-self-hosted-model
     runs-on: ubuntu-latest
     steps:
       - uses: actions/checkout@b4ffde65f46336ab88eb53be808477a3936bae11 # v4.1.1

From 64f4da8327d251d8930ae59a0e0c6a2c66e01607 Mon Sep 17 00:00:00 2001
From: Anton Dubovik <anton_dubovik@epam.com>
Date: Wed, 18 Sep 2024 08:39:38 +0000
Subject: [PATCH 10/16] chore: simplified names of the ci jobs

---
 .github/workflows/pr.yml      | 10 +++++-----
 .github/workflows/release.yml | 10 +++++-----
 2 files changed, 10 insertions(+), 10 deletions(-)

diff --git a/.github/workflows/pr.yml b/.github/workflows/pr.yml
index 0605935f..86f5ceb3 100644
--- a/.github/workflows/pr.yml
+++ b/.github/workflows/pr.yml
@@ -7,7 +7,7 @@ on:
 
 jobs:
   run-notebooks:
-    name: Run example Python notebooks
+    name: Cookbook notebooks
     runs-on: ubuntu-latest
     steps:
       - uses: actions/checkout@b4ffde65f46336ab88eb53be808477a3936bae11 # v4.1.1
@@ -19,7 +19,7 @@ jobs:
           up-flags: "--abort-on-container-exit --exit-code-from test --timeout 300"
 
   run-quickstart-model:
-    name: Run quickstart model example
+    name: Quickstart model
     runs-on: ubuntu-latest
     steps:
       - uses: actions/checkout@b4ffde65f46336ab88eb53be808477a3936bae11 # v4.1.1
@@ -30,7 +30,7 @@ jobs:
           up-flags: "--abort-on-container-exit --exit-code-from test --timeout 300"
 
   run-quickstart-application:
-    name: Run quickstart application example
+    name: Quickstart application
     runs-on: ubuntu-latest
     steps:
       - uses: actions/checkout@b4ffde65f46336ab88eb53be808477a3936bae11 # v4.1.1
@@ -40,7 +40,7 @@ jobs:
           up-flags: "--abort-on-container-exit --exit-code-from test --timeout 300"
 
   run-quickstart-addon:
-    name: Run quickstart addon example
+    name: Quickstart addon
     runs-on: ubuntu-latest
     steps:
       - uses: actions/checkout@b4ffde65f46336ab88eb53be808477a3936bae11 # v4.1.1
@@ -50,7 +50,7 @@ jobs:
           up-flags: "--abort-on-container-exit --exit-code-from test --timeout 300"
 
   run-quickstart-self-hosted-model:
-    name: Run quickstart self-hosted model example
+    name: Quickstart self-hosted model
     runs-on: ubuntu-latest
     steps:
       - uses: actions/checkout@b4ffde65f46336ab88eb53be808477a3936bae11 # v4.1.1
diff --git a/.github/workflows/release.yml b/.github/workflows/release.yml
index a6677137..e3828138 100644
--- a/.github/workflows/release.yml
+++ b/.github/workflows/release.yml
@@ -7,7 +7,7 @@ on:
 
 jobs:
   run-notebooks:
-    name: Run example Python notebooks
+    name: Cookbook notebooks
     runs-on: ubuntu-latest
     steps:
       - uses: actions/checkout@b4ffde65f46336ab88eb53be808477a3936bae11 # v4.1.1
@@ -20,7 +20,7 @@ jobs:
 
 
   run-quickstart-model:
-    name: Run quickstart model example
+    name: Quickstart model
     runs-on: ubuntu-latest
     steps:
       - uses: actions/checkout@b4ffde65f46336ab88eb53be808477a3936bae11 # v4.1.1
@@ -31,7 +31,7 @@ jobs:
           up-flags: "--abort-on-container-exit --exit-code-from test --timeout 300"
 
   run-quickstart-application:
-    name: Run quickstart application example
+    name: Quickstart application
     runs-on: ubuntu-latest
     steps:
       - uses: actions/checkout@b4ffde65f46336ab88eb53be808477a3936bae11 # v4.1.1
@@ -41,7 +41,7 @@ jobs:
           up-flags: "--abort-on-container-exit --exit-code-from test --timeout 300"
 
   run-quickstart-addon:
-    name: Run quickstart addon example
+    name: Quickstart addon
     runs-on: ubuntu-latest
     steps:
       - uses: actions/checkout@b4ffde65f46336ab88eb53be808477a3936bae11 # v4.1.1
@@ -51,7 +51,7 @@ jobs:
           up-flags: "--abort-on-container-exit --exit-code-from test --timeout 300"
 
   run-quickstart-self-hosted-model:
-    name: Run quickstart self-hosted model example
+    name: Quickstart self-hosted model
     runs-on: ubuntu-latest
     steps:
       - uses: actions/checkout@b4ffde65f46336ab88eb53be808477a3936bae11 # v4.1.1

From 11fa0ac4a7b51e6c299d1605bcd14dac59dce60b Mon Sep 17 00:00:00 2001
From: Anton Dubovik <anton_dubovik@epam.com>
Date: Wed, 18 Sep 2024 08:59:35 +0000
Subject: [PATCH 11/16] fix: migrated setup.py script for ollama to httpx

---
 .../ollama/ollama_setup/Dockerfile            |  9 +++--
 .../ollama/ollama_setup/requirements.txt      |  1 +
 .../ollama/ollama_setup/setup.py              | 36 ++++++++++---------
 3 files changed, 27 insertions(+), 19 deletions(-)
 create mode 100644 dial-docker-compose/ollama/ollama_setup/requirements.txt

diff --git a/dial-docker-compose/ollama/ollama_setup/Dockerfile b/dial-docker-compose/ollama/ollama_setup/Dockerfile
index 3afdb789..91b223bf 100644
--- a/dial-docker-compose/ollama/ollama_setup/Dockerfile
+++ b/dial-docker-compose/ollama/ollama_setup/Dockerfile
@@ -1,4 +1,7 @@
 FROM python:3.11-alpine
-RUN pip install requests
-COPY setup.py /setup.py
-CMD ["sh", "-c", "python /setup.py && tail -f /dev/null"]
+
+WORKDIR /app
+COPY * /app
+RUN pip install -r requirements.txt
+
+CMD ["sh", "-c", "python setup.py && tail -f /dev/null"]
diff --git a/dial-docker-compose/ollama/ollama_setup/requirements.txt b/dial-docker-compose/ollama/ollama_setup/requirements.txt
new file mode 100644
index 00000000..0329bccd
--- /dev/null
+++ b/dial-docker-compose/ollama/ollama_setup/requirements.txt
@@ -0,0 +1 @@
+httpx==0.27.2
\ No newline at end of file
diff --git a/dial-docker-compose/ollama/ollama_setup/setup.py b/dial-docker-compose/ollama/ollama_setup/setup.py
index 85e26858..cf5aa3b3 100755
--- a/dial-docker-compose/ollama/ollama_setup/setup.py
+++ b/dial-docker-compose/ollama/ollama_setup/setup.py
@@ -1,9 +1,8 @@
 import asyncio
 from contextlib import asynccontextmanager
-import logging
 import os
 import time
-import requests
+import httpx
 
 OLLAMA_URL = os.getenv("OLLAMA_URL")
 if OLLAMA_URL is None:
@@ -14,46 +13,51 @@
 
 HEALTH_FILE = "/healthy"
 
-log = logging.getLogger(__name__)
-log.setLevel(logging.INFO)
 
-log.info(f"OLLAMA_URL = {OLLAMA_URL}")
-log.info(f"OLLAMA_CHAT_MODEL = {OLLAMA_CHAT_MODEL}")
-log.info(f"OLLAMA_VISION_MODEL = {OLLAMA_VISION_MODEL}")
+def print_info(*args, **kwargs):
+    print(*args, **kwargs, flush=True)
+
+
+print_info(f"OLLAMA_URL = {OLLAMA_URL}")
+print_info(f"OLLAMA_CHAT_MODEL = {OLLAMA_CHAT_MODEL}")
+print_info(f"OLLAMA_VISION_MODEL = {OLLAMA_VISION_MODEL}")
 
 
 @asynccontextmanager
 async def timer(name: str):
-    log.info(f"[{name}] Starting...")
+    print_info(f"[{name}] Starting...")
     start = time.perf_counter()
     yield
     elapsed = time.perf_counter() - start
-    log.info(f"[{name}] Executed in {elapsed:.2f} seconds")
+    print_info(f"[{name}] Executed in {elapsed:.2f} seconds")
+
+
+ollama = httpx.AsyncClient(base_url=OLLAMA_URL, timeout=300)
 
 
-async def wait_for_ollama():
+async def wait_for_startup():
     while True:
         try:
-            if requests.get(OLLAMA_URL).ok:
+            if (await ollama.get("/")).is_success:
                 break
-        except requests.RequestException:
+        except Exception:
             pass
         await asyncio.sleep(1)
 
 
 async def pull_model(model):
     data = {"name": model, "stream": False}
-    requests.post(f"{OLLAMA_URL}/api/pull", json=data).raise_for_status()
+    (await ollama.post("/api/pull", json=data)).raise_for_status()
 
 
 async def create_alias(source, dest):
     data = {"source": source, "destination": dest}
-    requests.post(f"{OLLAMA_URL}/api/copy", json=data).raise_for_status()
+    (await ollama.post(f"/api/copy", json=data)).raise_for_status()
 
 
 async def load_model(model):
     data = {"model": model}
-    requests.post(f"{OLLAMA_URL}/api/generate", json=data).raise_for_status()
+    (await ollama.post(f"/api/generate", json=data)).raise_for_status()
 
 
 async def mark_as_healthy():
@@ -62,7 +66,7 @@ async def mark_as_healthy():
 
 async def main():
     async with timer("Waiting for Ollama to start"):
-        await wait_for_ollama()
+        await wait_for_startup()
 
     for model, alias in [
         (OLLAMA_CHAT_MODEL, "chat-model"),

From 517b1fae7d2d5617342e2279b5b7d3f6e06ca234 Mon Sep 17 00:00:00 2001
From: Anton Dubovik <anton_dubovik@epam.com>
Date: Wed, 18 Sep 2024 09:54:25 +0000
Subject: [PATCH 12/16] feat: added ci test for self-hosted embedding model

---
 dial-docker-compose/ci/ollama/.env            |  3 +-
 dial-docker-compose/ci/ollama/test/app.py     | 29 +++++++++++
 dial-docker-compose/ollama/core/config.json   |  7 ++-
 dial-docker-compose/ollama/docker-compose.yml |  1 +
 .../ollama/ollama_setup/setup.py              |  7 ++-
 .../quick-start-with-self-hosted-model.md     | 48 ++++++++++++++-----
 6 files changed, 78 insertions(+), 17 deletions(-)

diff --git a/dial-docker-compose/ci/ollama/.env b/dial-docker-compose/ci/ollama/.env
index 58a1c316..15ace1c5 100644
--- a/dial-docker-compose/ci/ollama/.env
+++ b/dial-docker-compose/ci/ollama/.env
@@ -1,3 +1,4 @@
 DIAL_DIR="./ollama"
 OLLAMA_CHAT_MODEL=llama3.1:8b-instruct-q4_0
-OLLAMA_VISION_MODEL=llava-phi3:3.8b-mini-q4_0
\ No newline at end of file
+OLLAMA_VISION_MODEL=llava-phi3:3.8b-mini-q4_0
+OLLAMA_EMBEDDING_MODEL=nomic-embed-text:137m-v1.5-fp16
\ No newline at end of file
diff --git a/dial-docker-compose/ci/ollama/test/app.py b/dial-docker-compose/ci/ollama/test/app.py
index 6f805f43..66e49085 100644
--- a/dial-docker-compose/ci/ollama/test/app.py
+++ b/dial-docker-compose/ci/ollama/test/app.py
@@ -1,6 +1,7 @@
 import base64
 import os
 from pathlib import Path
+from typing import Any
 import aiohttp
 import asyncio
 import backoff
@@ -71,6 +72,25 @@ async def dial_chat_completion(deployment_id: str, messages: list) -> str:
 
     return content
 
+async def dial_embeddings(deployment_id: str, input: Any) -> str:
+    api_url = f"{DIAL_URL}/openai/deployments/{deployment_id}/embeddings"
+
+    payload = {
+        "model": deployment_id,
+        "input": input,
+    }
+    headers = {"api-key": DIAL_API_KEY}
+    params = {"api-version": DIAL_API_VERSION}
+
+    body = await post_with_retry(api_url, payload, headers, params)
+    log.debug(f"Response: {body}")
+
+    embedding = body.get("data", [])[0].get("embedding", [])
+
+    log.debug(f"Len embedding vector: {len(embedding)}")
+
+    return embedding
+
 async def test_chat_model(deployment_id: str):
     message = "2 + 3 = ? Reply with a single number:"
     messages = [{"role": "user", "content": message}]
@@ -99,6 +119,13 @@ async def test_vision_model(deployment_id: str):
     if "vision" not in content.lower():
         raise ValueError(f"Test failed for {deployment_id!r}")
 
+async def test_embedding_model(deployment_id: str):
+    embeddings = await dial_embeddings(deployment_id, "cat")
+
+    if len(embeddings) == 0 or not isinstance(embeddings[0], float):
+        raise ValueError(f"Test failed for {deployment_id!r}")
+
+
 async def tests():
     async with timer("Testing chat-model"):
         await test_chat_model("chat-model")
@@ -106,6 +133,8 @@ async def tests():
     async with timer("Testing vision-model"):
         await test_vision_model("vision-model")
 
+    async with timer("Testing embedding-model"):
+        await test_embedding_model("embedding-model")
 
 if __name__ == "__main__":
     loop = asyncio.get_event_loop()
diff --git a/dial-docker-compose/ollama/core/config.json b/dial-docker-compose/ollama/core/config.json
index 4a55bd1e..e4b370cc 100644
--- a/dial-docker-compose/ollama/core/config.json
+++ b/dial-docker-compose/ollama/core/config.json
@@ -20,6 +20,10 @@
           "key": "dummy-key"
         }
       ]
+    },
+    "embedding-model": {
+      "type": "embedding",
+      "endpoint": "http://ollama:11434/v1/embeddings"
     }
   },
   "keys": {
@@ -32,7 +36,8 @@
     "default": {
       "limits": {
         "chat-model": {},
-        "vision-model": {}
+        "vision-model": {},
+        "embedding-model": {}
       }
     }
   }
diff --git a/dial-docker-compose/ollama/docker-compose.yml b/dial-docker-compose/ollama/docker-compose.yml
index 9712aa6a..68e316f0 100644
--- a/dial-docker-compose/ollama/docker-compose.yml
+++ b/dial-docker-compose/ollama/docker-compose.yml
@@ -19,6 +19,7 @@ services:
       - OLLAMA_URL=http://ollama:11434
       - OLLAMA_CHAT_MODEL=${OLLAMA_CHAT_MODEL}
       - OLLAMA_VISION_MODEL=${OLLAMA_VISION_MODEL}
+      - OLLAMA_EMBEDDING_MODEL=${OLLAMA_EMBEDDING_MODEL}
     healthcheck:
       test: ["CMD", "test", "-f", "/healthy"]
       interval: 10s
diff --git a/dial-docker-compose/ollama/ollama_setup/setup.py b/dial-docker-compose/ollama/ollama_setup/setup.py
index cf5aa3b3..ccc4d358 100755
--- a/dial-docker-compose/ollama/ollama_setup/setup.py
+++ b/dial-docker-compose/ollama/ollama_setup/setup.py
@@ -10,6 +10,7 @@
 
 OLLAMA_CHAT_MODEL = os.getenv("OLLAMA_CHAT_MODEL")
 OLLAMA_VISION_MODEL = os.getenv("OLLAMA_VISION_MODEL")
+OLLAMA_EMBEDDING_MODEL = os.getenv("OLLAMA_EMBEDDING_MODEL")
 
 HEALTH_FILE = "/healthy"
 
@@ -21,6 +22,7 @@ def print_info(*args, **kwargs):
 print_info(f"OLLAMA_URL = {OLLAMA_URL}")
 print_info(f"OLLAMA_CHAT_MODEL = {OLLAMA_CHAT_MODEL}")
 print_info(f"OLLAMA_VISION_MODEL = {OLLAMA_VISION_MODEL}")
+print_info(f"OLLAMA_EMBEDDING_MODEL = {OLLAMA_EMBEDDING_MODEL}")
 
 
 @asynccontextmanager
@@ -29,7 +31,7 @@ async def timer(name: str):
     start = time.perf_counter()
     yield
     elapsed = time.perf_counter() - start
-    print_info(f"[{name}] Executed in {elapsed:.2f} seconds")
+    print_info(f"[{name}] Finished in {elapsed:.2f} seconds")
 
 
 ollama = httpx.AsyncClient(base_url=OLLAMA_URL, timeout=300)
@@ -71,6 +73,7 @@ async def main():
     for model, alias in [
         (OLLAMA_CHAT_MODEL, "chat-model"),
         (OLLAMA_VISION_MODEL, "vision-model"),
+        (OLLAMA_EMBEDDING_MODEL, "embedding-model"),
     ]:
         if model:
             async with timer(f"Pulling model {model}"):
@@ -79,7 +82,7 @@ async def main():
             async with timer(f"Creating alias for {model}: {alias}"):
                 await create_alias(model, alias)
 
-    if model_to_load := OLLAMA_CHAT_MODEL or OLLAMA_VISION_MODEL:
+    if model_to_load := (OLLAMA_CHAT_MODEL or OLLAMA_VISION_MODEL):
         async with timer(f"Loading model {model_to_load} into memory"):
             await load_model(model_to_load)
 
diff --git a/docs/tutorials/quick-start-with-self-hosted-model.md b/docs/tutorials/quick-start-with-self-hosted-model.md
index c850eec3..54c96785 100644
--- a/docs/tutorials/quick-start-with-self-hosted-model.md
+++ b/docs/tutorials/quick-start-with-self-hosted-model.md
@@ -22,22 +22,32 @@ cd dial-docker-compose/ollama
 
 Ollama supports a wide range of popular open-source models.
 
-Consider first the modality your interested in - is a regular text-to-text chat model or a multi-modal vision model?
+Consider first the modality your are interested in - is it a regular text-to-text chat model, a multi-modal vision model or an embedding model?
 
-Follow the feature tags at https://ollama.com/search to find an appropriate model, e.g. one working with `Code`, supporting `Tools` or a `Vision` feature.
+Follow the feature tags _(`Embeddings`, `Code`, `Tools`, `Vision`)_ at https://ollama.com/search to find an appropriate model.
 
-We recommend to choose one of the following models which have been tested:
+We recommend to choose one of the following models which have been tested.
 
-|Model|Vision|Tools|Streaming|
-|----|----|----|----|
-|[llama3.1:8b-instruct-q4_0](https://ollama.com/library/llama3.1:8b-instruct-q4_0)|❌|✅ *(only in non-streaming mode)*|✅|
-|[mistral:7b-instruct-q4_0](https://ollama.com/library/mistral:7b-instruct-q4_0)|❌|❌|✅|
-|[phi3.5:3.8b-mini-instruct-q4_0](https://ollama.com/library/phi3.5:3.8b-mini-instruct-q4_0)|❌|❌|✅|
-|[gemma2:2b-instruct-q4_0](https://ollama.com/library/gemma2:2b-instruct-q4_0)|❌|❌|✅|
-|[llava:7b-v1.6-mistral-q4_0](https://ollama.com/library/llava:7b-v1.6-mistral-q4_0)|✅|❌|❌|
-|[llava-phi3:3.8b-mini-q4_0](https://ollama.com/library/llava-phi3:3.8b-mini-q4_0)|✅|❌|❌|
+### Chat models
 
-All the listed models support streaming.
+|Model|Tools|
+|----|----|
+|[llama3.1:8b-instruct-q4_0](https://ollama.com/library/llama3.1:8b-instruct-q4_0)|✅ *(only in non-streaming mode)*|
+|[mistral:7b-instruct-q4_0](https://ollama.com/library/mistral:7b-instruct-q4_0)|❌|
+|[phi3.5:3.8b-mini-instruct-q4_0](https://ollama.com/library/phi3.5:3.8b-mini-instruct-q4_0)|❌|
+|[gemma2:2b-instruct-q4_0](https://ollama.com/library/gemma2:2b-instruct-q4_0)|❌|
+
+All the models support streaming.
+
+### Vision models
+
+* [llava:7b-v1.6-mistral-q4_0](https://ollama.com/library/llava:7b-v1.6-mistral-q4_0)
+* [llava-phi3:3.8b-mini-q4_0](https://ollama.com/library/llava-phi3:3.8b-mini-q4_0)
+
+### Embedding models
+
+* [nomic-embed-text:137m-v1.5-fp16](https://ollama.com/library/nomic-embed-text:137m-v1.5-fp16)
+* [bge-m3:567m-fp16](https://ollama.com/library/bge-m3:567m-fp16)
 
 ## Step 3: Launch AI DIAL Chat
 
@@ -65,4 +75,16 @@ The model will be pulled and loaded into the memory of the Ollama server automat
 
 Finally, open http://localhost:3000/ in your browser to launch the AI DIAL Chat application and select `Self-hosted vision model` deployment to converse with the model.
 
-> Note, that the vision models we tested, do not support streaming of response. Moreover, they are typically more computationally expensive than the chat models. So it may take minutes for a vision model to respond.
\ No newline at end of file
+> Note, that the vision models we tested, do not support streaming of response. Moreover, they are typically more computationally expensive than the chat models. So it may take minutes for a vision model to respond.
+
+### Embedding model
+
+If you have chosen an embedding model _(e.g. nomic-embed-text:137m-v1.5-fp16)_, then run the command:
+
+```sh
+OLLAMA_EMBEDDING_MODEL=model_of_your_choice docker compose up --abort-on-container-exit
+```
+
+The model will be pulled and loaded into the memory of the Ollama server automatically. This may take a minute on the first run.
+
+The embedding model will become available in DIAL under the deployment name `embedding-model` and could be called via the endpoint: `localhost:8080/openai/deployments/embedding-model/embeddings`.
\ No newline at end of file

From bdd3536e4911f39a1eed9a67241825f77d4c5724 Mon Sep 17 00:00:00 2001
From: Anton Dubovik <anton_dubovik@epam.com>
Date: Thu, 19 Sep 2024 14:27:20 +0000
Subject: [PATCH 13/16] feat: used .env file instead of env vars in self-hosted
 model tutorial

---
 dial-docker-compose/ollama/.env               |  5 ++-
 .../quick-start-with-self-hosted-model.md     | 38 +++++++------------
 2 files changed, 17 insertions(+), 26 deletions(-)

diff --git a/dial-docker-compose/ollama/.env b/dial-docker-compose/ollama/.env
index 4daa1416..cdabc6b5 100644
--- a/dial-docker-compose/ollama/.env
+++ b/dial-docker-compose/ollama/.env
@@ -1 +1,4 @@
-DIAL_DIR="./ollama"
\ No newline at end of file
+DIAL_DIR="./ollama"
+OLLAMA_CHAT_MODEL=
+OLLAMA_VISION_MODEL=
+OLLAMA_EMBEDDING_MODEL=
\ No newline at end of file
diff --git a/docs/tutorials/quick-start-with-self-hosted-model.md b/docs/tutorials/quick-start-with-self-hosted-model.md
index 54c96785..337e6297 100644
--- a/docs/tutorials/quick-start-with-self-hosted-model.md
+++ b/docs/tutorials/quick-start-with-self-hosted-model.md
@@ -51,40 +51,28 @@ All the models support streaming.
 
 ## Step 3: Launch AI DIAL Chat
 
-### Chat model
+Configure `.env` file in the current directory according to the type of model you've chosen:
 
-If you have chosen a regular chat model _(e.g. llama3.1:8b-instruct-q4_0)_, then run the command:
+* Set `OLLAMA_CHAT_MODEL` for the name of a text model.
+* Set `OLLAMA_VISION_MODEL` for the name of a vision model.
+* Set `OLLAMA_EMBEDDING_MODEL` for the name of an embedding model.
 
-```sh
-OLLAMA_CHAT_MODEL=model_of_your_choice docker compose up --abort-on-container-exit
-```
-
-The model will be pulled and loaded into the memory of the Ollama server automatically. This may take a minute on the first run.
-
-Finally, open http://localhost:3000/ in your browser to launch the AI DIAL Chat application and select `Self-hosted chat model` deployment to converse with the model.
-
-### Vision model
+It's not necessary to configure all the models.
+If a model isn't set, then it won't be downloaded.
 
-If you have chosen a vision model _(e.g. llava-phi3:3.8b-mini-q4_0)_, then run the command:
+Then run the command:
 
 ```sh
-OLLAMA_VISION_MODEL=model_of_your_choice docker compose up --abort-on-container-exit
+docker compose up --abort-on-container-exit
 ```
 
-The model will be pulled and loaded into the memory of the Ollama server automatically. This may take a minute on the first run.
+It will pull and load into the memory of the Ollama server the specified models.
 
-Finally, open http://localhost:3000/ in your browser to launch the AI DIAL Chat application and select `Self-hosted vision model` deployment to converse with the model.
+> Keep in mind that a typical size of a lightweight Ollama model is around a few gigabytes. So it may take a few minutes _(or dozens of minutes)_ to download them on the first run depending on your Internet bandwidth.
 
-> Note, that the vision models we tested, do not support streaming of response. Moreover, they are typically more computationally expensive than the chat models. So it may take minutes for a vision model to respond.
-
-### Embedding model
-
-If you have chosen an embedding model _(e.g. nomic-embed-text:137m-v1.5-fp16)_, then run the command:
-
-```sh
-OLLAMA_EMBEDDING_MODEL=model_of_your_choice docker compose up --abort-on-container-exit
-```
+Finally, open http://localhost:3000/ in your browser to launch the AI DIAL Chat application and select an appropriate DIAL deployments to converse with:
 
-The model will be pulled and loaded into the memory of the Ollama server automatically. This may take a minute on the first run.
+* `Self-hosted chat model` deployment for the `OLLAMA_CHAT_MODEL`,
+* `Self-hosted vision model` deployment for the `OLLAMA_VISION_MODEL`,
 
 The embedding model will become available in DIAL under the deployment name `embedding-model` and could be called via the endpoint: `localhost:8080/openai/deployments/embedding-model/embeddings`.
\ No newline at end of file

From 1125e215e8d17b5ecb2c7fa18d9ccb31f51b9ba2 Mon Sep 17 00:00:00 2001
From: Anton Dubovik <anton_dubovik@epam.com>
Date: Thu, 19 Sep 2024 14:47:34 +0000
Subject: [PATCH 14/16] fix: increased timeout in the ollama setup script

---
 dial-docker-compose/ollama/ollama_setup/setup.py | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/dial-docker-compose/ollama/ollama_setup/setup.py b/dial-docker-compose/ollama/ollama_setup/setup.py
index ccc4d358..fdaad8b5 100755
--- a/dial-docker-compose/ollama/ollama_setup/setup.py
+++ b/dial-docker-compose/ollama/ollama_setup/setup.py
@@ -34,7 +34,7 @@ async def timer(name: str):
     print_info(f"[{name}] Finished in {elapsed:.2f} seconds")
 
 
-ollama = httpx.AsyncClient(base_url=OLLAMA_URL, timeout=300)
+ollama = httpx.AsyncClient(base_url=OLLAMA_URL, timeout=300000)
 
 
 async def wait_for_startup():

From 036a8409b09c38db5196d510d052785b71be2c75 Mon Sep 17 00:00:00 2001
From: sr-remsha <sr.remsha@gmail.com>
Date: Thu, 19 Sep 2024 18:49:56 +0200
Subject: [PATCH 15/16] review

---
 docs/README.md                                |  4 +-
 .../quick-start-with-self-hosted-model.md     | 39 +++++++++----------
 sidebars.js                                   |  2 +-
 3 files changed, 21 insertions(+), 24 deletions(-)

diff --git a/docs/README.md b/docs/README.md
index edde7112..93b9a2de 100644
--- a/docs/README.md
+++ b/docs/README.md
@@ -34,8 +34,8 @@
 
 * [Launch AI DIAL Chat with an Azure model](./tutorials/quick-start-model.md)
 * [Launch AI DIAL Chat with a self-hosted model](./tutorials/quick-start-with-self-hosted-model.md)
-* [Launch AI DIAL Chat with a Sample Application](./tutorials/quick-start-with-application.md)
-* [Launch AI DIAL Chat with a Sample Addon](./tutorials/quick-start-with-addon.md)
+* [Launch AI DIAL Chat with a sample application](./tutorials/quick-start-with-application.md)
+* [Launch AI DIAL Chat with a sample addon](./tutorials/quick-start-with-addon.md)
 
 ## AI DIAL Chat Application User Manual
 
diff --git a/docs/tutorials/quick-start-with-self-hosted-model.md b/docs/tutorials/quick-start-with-self-hosted-model.md
index 337e6297..bebdfd37 100644
--- a/docs/tutorials/quick-start-with-self-hosted-model.md
+++ b/docs/tutorials/quick-start-with-self-hosted-model.md
@@ -1,4 +1,4 @@
-# Launch AI DIAL Chat with a self-hosted model
+# Launch AI DIAL Chat with a Self-Hosted Model
 
 ## Introduction
 
@@ -24,9 +24,9 @@ Ollama supports a wide range of popular open-source models.
 
 Consider first the modality your are interested in - is it a regular text-to-text chat model, a multi-modal vision model or an embedding model?
 
-Follow the feature tags _(`Embeddings`, `Code`, `Tools`, `Vision`)_ at https://ollama.com/search to find an appropriate model.
+Follow the feature tags _(`Embeddings`, `Code`, `Tools`, `Vision`)_ at [Ollama Search](https://ollama.com/search) to find the appropriate model.
 
-We recommend to choose one of the following models which have been tested.
+We recommend choosing one of the following models which have been tested.
 
 ### Chat models
 
@@ -51,28 +51,25 @@ All the models support streaming.
 
 ## Step 3: Launch AI DIAL Chat
 
-Configure `.env` file in the current directory according to the type of model you've chosen:
+1. Configure `.env` file in the current directory according to the type of model you've chosen:
 
-* Set `OLLAMA_CHAT_MODEL` for the name of a text model.
-* Set `OLLAMA_VISION_MODEL` for the name of a vision model.
-* Set `OLLAMA_EMBEDDING_MODEL` for the name of an embedding model.
+    * Set `OLLAMA_CHAT_MODEL` for the name of a text model.
+    * Set `OLLAMA_VISION_MODEL` for the name of a vision model.
+    * Set `OLLAMA_EMBEDDING_MODEL` for the name of an embedding model.
+    
+    **Note**: It's not necessary to configure all the models. If a model isn't set, then it won't be downloaded.
 
-It's not necessary to configure all the models.
-If a model isn't set, then it won't be downloaded.
+2. Then run the following command to pull and load into the memory of the Ollama server the specified models:
 
-Then run the command:
+    ```sh
+    docker compose up --abort-on-container-exit
+    ```
 
-```sh
-docker compose up --abort-on-container-exit
-```
-
-It will pull and load into the memory of the Ollama server the specified models.
-
-> Keep in mind that a typical size of a lightweight Ollama model is around a few gigabytes. So it may take a few minutes _(or dozens of minutes)_ to download them on the first run depending on your Internet bandwidth.
+    > Keep in mind that a typical size of a lightweight Ollama model is around a few gigabytes. So it may take a few minutes _(or dozens of minutes)_ to download them on the first run depending on your Internet bandwidth.
 
-Finally, open http://localhost:3000/ in your browser to launch the AI DIAL Chat application and select an appropriate DIAL deployments to converse with:
+3. Finally, open http://localhost:3000/ in your browser to launch the AI DIAL Chat application and select an appropriate AI DIAL deployments to converse with:
 
-* `Self-hosted chat model` deployment for the `OLLAMA_CHAT_MODEL`,
-* `Self-hosted vision model` deployment for the `OLLAMA_VISION_MODEL`,
+    * `Self-hosted chat model` deployment for the `OLLAMA_CHAT_MODEL`
+    * `Self-hosted vision model` deployment for the `OLLAMA_VISION_MODEL`
 
-The embedding model will become available in DIAL under the deployment name `embedding-model` and could be called via the endpoint: `localhost:8080/openai/deployments/embedding-model/embeddings`.
\ No newline at end of file
+The embedding model will become available in AI DIAL under the deployment name `embedding-model` and could be called via the endpoint: `localhost:8080/openai/deployments/embedding-model/embeddings`.
\ No newline at end of file
diff --git a/sidebars.js b/sidebars.js
index eba6d6dd..dc81800b 100644
--- a/sidebars.js
+++ b/sidebars.js
@@ -105,7 +105,7 @@ const sidebars = {
         {
           type: 'doc',
           id: 'tutorials/quick-start-with-self-hosted-model',
-          label: 'Chat with a self-hosted model',
+          label: 'Chat with a Self-Hosted Model',
         },
         {
           type: 'doc',

From 3c3e4c81bfc3d0ab9c2577a5287d84642970d9cf Mon Sep 17 00:00:00 2001
From: Anton Dubovik <anton_dubovik@epam.com>
Date: Thu, 19 Sep 2024 16:59:46 +0000
Subject: [PATCH 16/16] feat: added progress bar for model downloading

---
 dial-docker-compose/ollama/.env.example       |  3 +
 .../ollama/ollama_setup/requirements.txt      |  4 +-
 .../ollama/ollama_setup/setup.py              | 81 +++++++++++++------
 .../quick-start-with-self-hosted-model.md     |  8 +-
 4 files changed, 69 insertions(+), 27 deletions(-)
 create mode 100644 dial-docker-compose/ollama/.env.example

diff --git a/dial-docker-compose/ollama/.env.example b/dial-docker-compose/ollama/.env.example
new file mode 100644
index 00000000..06216195
--- /dev/null
+++ b/dial-docker-compose/ollama/.env.example
@@ -0,0 +1,3 @@
+OLLAMA_CHAT_MODEL=llama3.1:8b-instruct-q4_0
+OLLAMA_VISION_MODEL=llava-phi3:3.8b-mini-q4_0
+OLLAMA_EMBEDDING_MODEL=bge-m3:567m-fp16
\ No newline at end of file
diff --git a/dial-docker-compose/ollama/ollama_setup/requirements.txt b/dial-docker-compose/ollama/ollama_setup/requirements.txt
index 0329bccd..ac6a93f1 100644
--- a/dial-docker-compose/ollama/ollama_setup/requirements.txt
+++ b/dial-docker-compose/ollama/ollama_setup/requirements.txt
@@ -1 +1,3 @@
-httpx==0.27.2
\ No newline at end of file
+httpx==0.27.2
+tqdm==4.66.5
+ollama==0.3.3
\ No newline at end of file
diff --git a/dial-docker-compose/ollama/ollama_setup/setup.py b/dial-docker-compose/ollama/ollama_setup/setup.py
index fdaad8b5..3e27663e 100755
--- a/dial-docker-compose/ollama/ollama_setup/setup.py
+++ b/dial-docker-compose/ollama/ollama_setup/setup.py
@@ -1,8 +1,11 @@
 import asyncio
 from contextlib import asynccontextmanager
 import os
+import sys
 import time
-import httpx
+import asyncio
+from ollama import AsyncClient
+from tqdm import tqdm
 
 OLLAMA_URL = os.getenv("OLLAMA_URL")
 if OLLAMA_URL is None:
@@ -15,9 +18,19 @@
 HEALTH_FILE = "/healthy"
 
 
-def print_info(*args, **kwargs):
-    print(*args, **kwargs, flush=True)
+class Writer:
+    @classmethod
+    def write(cls, s: str):
+        # NOTE: every tqdm progress bar update is deliberately ended with "\n",
+        # otherwise one wouldn't see the bar running in console upon running `docker compose up`.
+        print(s, file=sys.stdout, flush=True, end="\n")
+
+    @classmethod
+    def flush(cls):
+        sys.stdout.flush()
+
 
+print_info = Writer.write
 
 print_info(f"OLLAMA_URL = {OLLAMA_URL}")
 print_info(f"OLLAMA_CHAT_MODEL = {OLLAMA_CHAT_MODEL}")
@@ -34,39 +47,57 @@ async def timer(name: str):
     print_info(f"[{name}] Finished in {elapsed:.2f} seconds")
 
 
-ollama = httpx.AsyncClient(base_url=OLLAMA_URL, timeout=300000)
-
-
 async def wait_for_startup():
+    attempt = 0
     while True:
+        attempt += 1
         try:
-            if (await ollama.get("/")).is_success:
-                break
+            await AsyncClient(host=OLLAMA_URL, timeout=5).ps()
         except Exception:
-            pass
-        await asyncio.sleep(1)
+            print_info(f"[{attempt:>3}] Waiting for Ollama to start...")
+            await asyncio.sleep(5)
+        else:
+            break
 
 
-async def pull_model(model):
-    data = {"name": model, "stream": False}
-    (await ollama.post("/api/pull", json=data)).raise_for_status()
+async def pull_model(client: AsyncClient, model: str):
+    response = await client.pull(model, stream=True)
 
+    progress_bar = None
+    prev_status = None
 
-async def create_alias(source, dest):
-    data = {"source": source, "destination": dest}
-    (await ollama.post(f"/api/copy", json=data)).raise_for_status()
+    async for chunk in response:
+        status = chunk["status"]
+        total = chunk.get("total")
+        completed = chunk.get("completed")
 
+        if status != prev_status and total:
+            prev_status = status
+            if progress_bar:
+                progress_bar.close()
+            progress_bar = tqdm(
+                total=total, unit="B", unit_scale=True, desc=f"[{status}]", file=Writer
+            )
 
-async def load_model(model):
-    data = {"model": model}
-    (await ollama.post(f"/api/generate", json=data)).raise_for_status()
+        if completed and progress_bar and total:
+            progress_bar.n = completed
+            progress_bar.set_description(f"[{status}]")
+            progress_bar.refresh()
 
+        if total and total == completed and progress_bar:
+            progress_bar.close()
 
-async def mark_as_healthy():
+        if not completed and not total:
+            print_info(f"[{status}]")
+
+
+async def create_health_mark():
     open(HEALTH_FILE, "w").close()
 
 
 async def main():
+    client = AsyncClient(host=OLLAMA_URL, timeout=300000)
+
     async with timer("Waiting for Ollama to start"):
         await wait_for_startup()
 
@@ -77,16 +108,18 @@ async def main():
     ]:
         if model:
             async with timer(f"Pulling model {model}"):
-                await pull_model(model)
+                await pull_model(client, model)
 
             async with timer(f"Creating alias for {model}: {alias}"):
-                await create_alias(model, alias)
+                await client.copy(model, alias)
 
     if model_to_load := (OLLAMA_CHAT_MODEL or OLLAMA_VISION_MODEL):
         async with timer(f"Loading model {model_to_load} into memory"):
-            await load_model(model_to_load)
+            await client.generate(model_to_load)
+
+    await create_health_mark()
 
-    await mark_as_healthy()
+    print_info("The Ollama server is up and running.")
 
 
 if __name__ == "__main__":
diff --git a/docs/tutorials/quick-start-with-self-hosted-model.md b/docs/tutorials/quick-start-with-self-hosted-model.md
index bebdfd37..77b18828 100644
--- a/docs/tutorials/quick-start-with-self-hosted-model.md
+++ b/docs/tutorials/quick-start-with-self-hosted-model.md
@@ -56,7 +56,7 @@ All the models support streaming.
     * Set `OLLAMA_CHAT_MODEL` for the name of a text model.
     * Set `OLLAMA_VISION_MODEL` for the name of a vision model.
     * Set `OLLAMA_EMBEDDING_MODEL` for the name of an embedding model.
-    
+
     **Note**: It's not necessary to configure all the models. If a model isn't set, then it won't be downloaded.
 
 2. Then run the following command to pull and load into the memory of the Ollama server the specified models:
@@ -66,10 +66,14 @@ All the models support streaming.
     ```
 
     > Keep in mind that a typical size of a lightweight Ollama model is around a few gigabytes. So it may take a few minutes _(or dozens of minutes)_ to download them on the first run depending on your Internet bandwidth.
+    >
+    > The model is fully loaded once `ollama-setup` service prints `The Ollama server is up and running.`
 
 3. Finally, open http://localhost:3000/ in your browser to launch the AI DIAL Chat application and select an appropriate AI DIAL deployments to converse with:
 
     * `Self-hosted chat model` deployment for the `OLLAMA_CHAT_MODEL`
     * `Self-hosted vision model` deployment for the `OLLAMA_VISION_MODEL`
 
-The embedding model will become available in AI DIAL under the deployment name `embedding-model` and could be called via the endpoint: `localhost:8080/openai/deployments/embedding-model/embeddings`.
\ No newline at end of file
+> Note, that the vision models we tested, do not support streaming of response. Moreover, they are typically more computationally expensive than the chat models. So it may take minutes for a vision model to respond.
+
+The embedding model will become available in AI DIAL under the deployment name `embedding-model` and could be called via the endpoint: `localhost:8080/openai/deployments/embedding-model/embeddings`.