### Complete Example Configuration File

Source: https://github.com/microsoft/call-center-ai/blob/main/_autodocs/03-configuration.md

A comprehensive example of a configuration file (config.yaml) showing all available settings for various services like AI Search, Translation, Cognitive Services, Communication Services, Database, LLM, Queue, SMS, Conversation, App Configuration, and Resources.

```yaml
# config.yaml

public_domain: "https://example.com"
version: "1.0.0"

ai_search:
  endpoint: "https://example.search.windows.net"
  api_key: "${AI_SEARCH_API_KEY}"
  index_name: "documents"

ai_translation:
  endpoint: "https://api.cognitive.microsofttranslator.com"
  api_key: "${AI_TRANSLATION_API_KEY}"
  region: "westeurope"

cognitive_service:
  endpoint: "https://westeurope.tts.speech.microsoft.com"
  api_key: "${COGNITIVE_SERVICE_API_KEY}"
  region: "westeurope"

communication_services:
  endpoint: "https://xxx.communication.azure.com"
  access_key: "${COMMUNICATION_SERVICES_ACCESS_KEY}"
  phone_number: "+33612345678"
  resource_id: "/subscriptions/xxx/resourceGroups/xxx/providers/Microsoft.Communication/communicationServices/xxx"

database:
  endpoint: "https://xxx.documents.azure.com:443/"
  api_key: "${COSMOS_DB_API_KEY}"
  database: "call-center"
  container: "calls"

llm:
  fast:
    endpoint: "https://xxx.openai.azure.com"
    api_key: "${LLM_FAST_API_KEY}"
    deployment: "gpt-4-1-mini"
    max_tokens: 2000
  slow:
    endpoint: "https://yyy.openai.azure.com"
    api_key: "${LLM_SLOW_API_KEY}"
    deployment: "gpt-4-1"
    max_tokens: 4000

queue:
  connection_string: "${AZURE_STORAGE_CONNECTION_STRING}"
  call_queue_name: "call-events"
  post_queue_name: "post-events"
  sms_queue_name: "sms-events"
  training_queue_name: "training-events"

sms:
  mode: "azure"

conversation:
  initiate:
    bot_name: "Amélie"
    bot_company: "Contoso"
    agent_phone_number: "+33698765432"
    task: "Help the customer with their insurance claim"
    lang:
      default_short_code: "fr-FR"

app_configuration:
  connection_string: "${APP_CONFIGURATION_CONNECTION_STRING}"
  ttl_sec: 60

resources:
  sounds_container_url: "https://xxx.blob.core.windows.net/sounds"
  recordings_container_url: "https://xxx.blob.core.windows.net/recordings"
```

--------------------------------

### Configuration Example

Source: https://github.com/microsoft/call-center-ai/blob/main/_autodocs/README.md

Example of AI Search configuration settings in YAML format. This includes the endpoint and API key required for connecting to the AI Search service.

```yaml
# Configuration example
ai_search:
  endpoint: "https://example.search.windows.net"
  api_key: "xxx"
```

--------------------------------

### REST API Example

Source: https://github.com/microsoft/call-center-ai/blob/main/_autodocs/README.md

Example of how to make a POST request to the REST API for call management. Ensure to set the Content-Type header and provide a JSON payload.

```bash
# REST API example
curl -X POST http://localhost:8080/call \
  -H "Content-Type: application/json" \
  -d '{...}'
```

--------------------------------

### Start the Call Center AI Server

Source: https://github.com/microsoft/call-center-ai/blob/main/_autodocs/00-index.md

Run the main application module using Python to start the server. The server will be accessible at `http://localhost:8080`.

```bash
python -m app.main
# Server runs on http://localhost:8080
```

--------------------------------

### Install Python Packages

Source: https://github.com/microsoft/call-center-ai/blob/main/examples/blocklist.ipynb

Install the required Azure SDKs and pandas library for the project. This command ensures specific versions are used for compatibility.

```python
%pip install azure-identity==1.15.0 azure-ai-contentsafety==1.0.0 pandas==2.2.0
```

--------------------------------

### Python: Handle Play Started Event

Source: https://github.com/microsoft/call-center-ai/blob/main/_autodocs/05-workflows-and-handlers.md

Records that audio playback has started and updates the call state. Triggered by the `Microsoft.Communication.PlayStarted` CloudEvent.

```python
async def on_play_started(
    call: CallStateModel,
    scheduler: Scheduler,
) -> None:
    # Records that audio playback started
    # Updates call state
    pass
```

--------------------------------

### Start Azure Dev Tunnel

Source: https://github.com/microsoft/call-center-ai/blob/main/README.md

Starts a tunnel for connecting to your local development environment. This must be run in a separate terminal and kept running.

```zsh
# Start the tunnel
make tunnel
```

--------------------------------

### AI Search Configuration Example

Source: https://github.com/microsoft/call-center-ai/blob/main/_autodocs/03-configuration.md

Example YAML configuration for Azure AI Search, specifying endpoint, API key, index name, and search parameters.

```yaml
ai_search:
  endpoint: https://example.search.windows.net
  api_key: "xxxx"
  index_name: documents
  search_top_k: 5
  strictness: 0.5
  expansion_n_messages: 3
```

--------------------------------

### System Prompt Template Example

Source: https://github.com/microsoft/call-center-ai/blob/main/_autodocs/07-advanced-topics.md

Defines a system prompt template using Jinja2 for placeholder substitution. Placeholders like bot name, company, date, and phone number can be dynamically inserted.

```yaml
prompts:
  llm:
    default_system_tpl: |
      Assistant is called {bot_name} and works for {bot_company}.
      Today is {date}.
      Customer is calling from {phone_number}.

```

--------------------------------

### Custom Prompt Templates Example

Source: https://github.com/microsoft/call-center-ai/blob/main/_autodocs/07-advanced-topics.md

Illustrates defining multiple custom prompt templates for text-to-speech. One template is randomly selected for each call, allowing for varied greetings.

```yaml
prompts:
  tts:
    hello_tpl:
      - "Hello, I'm {bot_name} from {bot_company}!"
      - "Hi, this is {bot_name} speaking."

```

--------------------------------

### Get Deployment Logs

Source: https://github.com/microsoft/call-center-ai/blob/main/README.md

Retrieve the deployment logs for the Call Center AI project using the Make command. Specify the resource group name.

```zsh
make logs name=my-rg-name
```

--------------------------------

### Python PhoneNumber Example

Source: https://github.com/microsoft/call-center-ai/blob/main/_autodocs/README.md

Example demonstrating the usage of the PhoneNumber class in Python to extract timezone information. This requires the PhoneNumber library to be installed.

```python
# Python example
phone = PhoneNumber("+33612345678")
tz = PhoneNumber.tz(phone)
```

--------------------------------

### Run Development Server with Hot-Reloading

Source: https://github.com/microsoft/call-center-ai/blob/main/README.md

Starts the development server with automatic code reloading on file changes. The API server will be accessible at http://localhost:8080.

```zsh
make dev
```

--------------------------------

### Example Error Response

Source: https://github.com/microsoft/call-center-ai/blob/main/_autodocs/02-types.md

Illustrates the JSON format for an error response, including a summary message and specific details about the error. This example is useful for understanding how to parse and handle API errors.

```json
{
  "error": {
    "message": "Validation error",
    "details": ["Invalid phone number format"]
  }
}
```

--------------------------------

### Search Training Documents

Source: https://github.com/microsoft/call-center-ai/blob/main/_autodocs/04-persistence-api.md

Example of searching for training documents using the ISearch interface. Specify language, query text, and optionally use cache_only.

```python
docs = await search.training_search_all(
    lang="fr-FR",
    text="Comment résilier mon contrat?",
)
if docs:
    for doc in docs:
        print(f"[{doc.score}] {doc.title}: {doc.content}")
```

--------------------------------

### Start Audio Streaming from Call

Source: https://github.com/microsoft/call-center-ai/blob/main/_autodocs/06-helpers-and-utilities.md

Initiates the capture of audio streams from an active call. Requires a CallAutomationClient and call connection ID.

```python
async def start_audio_streaming(
    client: CallAutomationClient,
    call_connection_id: str,
) -> None:
    ...
```

--------------------------------

### Message Merging Example

Source: https://github.com/microsoft/call-center-ai/blob/main/_autodocs/07-advanced-topics.md

Demonstrates how consecutive messages from the same persona with the same action are merged. This reduces LLM context size when streaming responses in multiple chunks.

```python
[
    MessageModel(persona=ASSISTANT, action=TALK, content="Hello"),
    MessageModel(persona=ASSISTANT, action=TALK, content="How are you?"),
    MessageModel(persona=HUMAN, action=TALK, content="I'm fine"),
]
```

```python
[
    MessageModel(persona=ASSISTANT, action=TALK, content="Hello How are you?"),
    MessageModel(persona=HUMAN, action=TALK, content="I'm fine"),
]
```

--------------------------------

### API Reference Documentation

Source: https://github.com/microsoft/call-center-ai/blob/main/_autodocs/MANIFEST.txt

This section outlines the structure and content of the generated API documentation, including details on type signatures, parameter descriptions, return types, usage examples, and error conditions.

```APIDOC
## API Documentation Structure

### Description
Provides detailed information for every exported symbol, ensuring comprehensive coverage of the API surface.

### Content:
- Full type signature
- Parameter descriptions with types
- Return type and description
- Usage examples (code samples)
- Source file location with line numbers
- Related types and cross-references
- Configuration examples (where applicable)
- Error conditions and exceptions
- Performance considerations

### Format:
- Markdown with clear hierarchy
- Code syntax highlighting
- Tables for quick reference
- Runnable examples
- YAML configuration examples
- Architecture diagrams
- State machine diagrams

### Completeness:
- COMPREHENSIVE: All public APIs documented.
- PRECISE: Signatures verified against source code.
- PRACTICAL: Includes runnable examples and patterns.
- ORGANIZED: Cross-referenced and logically structured.
- ACCURATE: Generated directly from source analysis.
- USEFUL: Standalone reference.

### Navigation:
- Table of contents with 9 documents
- Cross-references between documents
- Index with quick lookup
- Source code location references
- Related topics links

### Usage Guidelines:
- **API Users**: Start with `01-rest-api.md`, `02-types.md`
- **Developers**: Start with `00-index.md`, then specialized docs
- **Operations**: Start with `03-configuration.md`, `07-advanced-topics.md`
- **Architects**: Start with `00-index.md`, `05-workflows-and-handlers.md`, `07-advanced-topics.md`
```

--------------------------------

### Get Service Readiness Status JSON

Source: https://github.com/microsoft/call-center-ai/blob/main/_autodocs/01-rest-api.md

This JSON response indicates the status of various services checked during the readiness probe.

```json
{
  "status": "ok",
  "checks": [
    {
      "id": "cache",
      "status": "ok"
    },
    {
      "id": "store",
      "status": "ok"
    },
    {
      "id": "startup",
      "status": "ok"
    },
    {
      "id": "search",
      "status": "ok"
    },
    {
      "id": "sms",
      "status": "ok"
    }
  ]
}
```

--------------------------------

### Validating Incoming Claim Data

Source: https://github.com/microsoft/call-center-ai/blob/main/_autodocs/07-advanced-topics.md

Example of validating incoming claim data against a dynamically generated Pydantic model.

```python
claim_entry = claim_model.model_validate(data)
```

--------------------------------

### GET /health/readiness

Source: https://github.com/microsoft/call-center-ai/blob/main/_autodocs/01-rest-api.md

Checks if the service is ready to serve requests. This includes validating the status of cache, database, search, and SMS services.

```APIDOC
## GET /health/readiness

### Description
Check if the service is ready to serve requests, including validation of cache, database, search, and SMS services.

### Method
GET

### Endpoint
/health/readiness

### Parameters
None

### Response
#### Success Response (200 OK)
- **status** (string) - Indicates the overall status of the readiness check.
- **checks** (array) - An array of objects, each detailing the status of a specific service check.
  - **id** (string) - The identifier of the service being checked.
  - **status** (string) - The status of the service check ('ok' or other).

#### Error Response (503 Service Unavailable)
Indicates that the service is not ready to serve requests.

### Request Example
```bash
curl http://localhost:8080/health/readiness
```

### Response Example
```json
{
  "status": "ok",
  "checks": [
    {
      "id": "cache",
      "status": "ok"
    },
    {
      "id": "store",
      "status": "ok"
    },
    {
      "id": "startup",
      "status": "ok"
    },
    {
      "id": "search",
      "status": "ok"
    },
    {
      "id": "sms",
      "status": "ok"
    }
  ]
}
```
```

--------------------------------

### Override Per-Call Claim Schema

Source: https://github.com/microsoft/call-center-ai/blob/main/_autodocs/07-advanced-topics.md

Example of initiating a call via a POST request with a custom, per-call claim schema defined in the request body.

```bash
curl -X POST http://localhost:8080/call \
  -H "Content-Type: application/json" \
  -d '{ \
    "phone_number": "+33612345678", \
    "bot_name": "Support", \
    "bot_company": "Contoso", \
    "agent_phone_number": "+33698765432", \
    "claim": [ \
      { \
        "name": "order_id", \
        "type": "text", \
        "description": "Order ID" \
      }, \
      { \
        "name": "issue_type", \
        "type": "text", \
        "description": "Type of issue" \
      } \
    ] \
  }'
```

--------------------------------

### Create Azure Blob Storage Container

Source: https://github.com/microsoft/call-center-ai/blob/main/_autodocs/07-advanced-topics.md

Command to create a new blob container in Azure Storage for storing call recordings. Ensure you have the Azure CLI installed and are logged in.

```bash
az storage container create \
  --name recordings \
  --account-name myaccount

```

--------------------------------

### Get or Create Shared Async HTTP Session

Source: https://github.com/microsoft/call-center-ai/blob/main/_autodocs/06-helpers-and-utilities.md

This function provides a shared `aiohttp.ClientSession` for making HTTP requests. It ensures connection pooling and graceful shutdown, beneficial for SDK clients.

```python
import aiohttp

async def get_session() -> aiohttp.ClientSession:
    # Implementation details for getting or creating the session
    pass

# Usage example:
session = await get_session()
```

--------------------------------

### CosmosDbStore Initialization

Source: https://github.com/microsoft/call-center-ai/blob/main/_autodocs/04-persistence-api.md

Illustrates the initialization of the CosmosDbStore, which requires an ICache instance and CosmosDbModel configuration.

```python
class CosmosDbStore(IStore):
    def __init__(self, cache: ICache, config: CosmosDbModel):
        ...
```

--------------------------------

### Initialize Persistence Backends

Source: https://github.com/microsoft/call-center-ai/blob/main/_autodocs/04-persistence-api.md

Instantiate backends like cache, database, search, SMS, and call queue from the application configuration. This pattern allows for easy swapping of implementations via configuration without code modifications.

```python
from app.helpers.config import CONFIG

# Instantiate backends from config
cache = CONFIG.cache.instance
db = CONFIG.database.instance
search = CONFIG.ai_search.instance
sms = CONFIG.sms.instance
call_queue = CONFIG.queue.call
```

--------------------------------

### CONFIG

Source: https://github.com/microsoft/call-center-ai/blob/main/_autodocs/06-helpers-and-utilities.md

Global configuration singleton that is loaded on startup. Provides access to application settings.

```APIDOC
## CONFIG

### Description
Global configuration singleton loaded on startup.

### Type
`RootModel`

### Access
```python
from app.helpers.config import CONFIG

endpoint = CONFIG.communication_services.endpoint
```
```

--------------------------------

### GET /call

Source: https://github.com/microsoft/call-center-ai/blob/main/_autodocs/01-rest-api.md

Retrieves a list of all calls, with an option to filter by a specific phone number.

```APIDOC
## GET /call

### Description
List all calls (optionally filtered by phone number).

### Method
GET

### Endpoint
/call

### Parameters
#### Query Parameters
- **phone_number** (string) - Optional - Filter by phone number (URL-encoded, e.g., `%2B33612345678`)

### Response
#### Success Response (200 OK)
- **Array of CallGetModel objects**

### Request Example
```bash
curl 'http://localhost:8080/call?phone_number=%2B33612345678'
```
```

--------------------------------

### Configure System with Azure Credentials

Source: https://github.com/microsoft/call-center-ai/blob/main/_autodocs/00-index.md

Create a `config.yaml` file to store your Azure resource credentials and endpoints. Ensure all required fields are populated for the services to function correctly.

```yaml
public_domain: "https://yourname.com"
ai_search:
  endpoint: "https://xxx.search.windows.net"
  api_key: "xxx"
  index_name: "documents"
database:
  endpoint: "https://xxx.documents.azure.com"
  api_key: "xxx"
  database: "call-center"
  container: "calls"
communication_services:
  endpoint: "https://xxx.communication.azure.com"
  access_key: "xxx"
  phone_number: "+33612345678"
  resource_id: "/subscriptions/.../providers/Microsoft.Communication/communicationServices/xxx"
# ... (see Configuration Reference for full spec)
```

--------------------------------

### GET /health/liveness

Source: https://github.com/microsoft/call-center-ai/blob/main/_autodocs/01-rest-api.md

Checks the basic liveness of the service. It's a simple probe to determine if the service is running.

```APIDOC
## GET /health/liveness

### Description
Check if the service is running (basic liveness probe).

### Method
GET

### Endpoint
/health/liveness

### Parameters
None

### Response
#### Success Response (200 OK)
No body

### Request Example
```bash
curl http://localhost:8080/health/liveness
```
```

--------------------------------

### Access Global Configuration

Source: https://github.com/microsoft/call-center-ai/blob/main/_autodocs/06-helpers-and-utilities.md

Access the global configuration singleton, which is loaded on startup. Import the CONFIG object to retrieve settings.

```python
from app.helpers.config import CONFIG

endpoint = CONFIG.communication_services.endpoint
```

--------------------------------

### GET /report

Source: https://github.com/microsoft/call-center-ai/blob/main/_autodocs/01-rest-api.md

Retrieves a list of all calls, with an option to filter by phone number. The output is provided as an HTML web interface.

```APIDOC
## GET /report

### Description
Display a list of all calls in an HTML web interface (optionally filtered by phone number).

### Method
GET

### Endpoint
/report

#### Query Parameters
- **phone_number** (string) - Optional - Filter by phone number

### Response
#### Success Response (200 OK)
**Content-Type:** text/html

### Request Example
```bash
curl 'http://localhost:8080/report?phone_number=%2B33612345678'
```
```

--------------------------------

### GET /call/{call_id_or_phone_number}

Source: https://github.com/microsoft/call-center-ai/blob/main/_autodocs/01-rest-api.md

Retrieves details of a specific call using its unique call ID or the associated phone number.

```APIDOC
## GET /call/{call_id_or_phone_number}

### Description
Retrieve a specific call by call ID or phone number.

### Method
GET

### Endpoint
/call/{call_id_or_phone_number}

### Parameters
#### Path Parameters
- **call_id_or_phone_number** (string) - The UUID of the call or the phone number (URL-encoded) to identify the call.

### Response
#### Success Response (200 OK)
- **CallGetModel** (object) - See [Call Model](#call-model) for details.

### Request Example
```bash
curl 'http://localhost:8080/call/123e4567-e89b-12d3-a456-426614174000'
curl 'http://localhost:8080/call/%2B33612345678'
```
```

--------------------------------

### Make a Test Call via API

Source: https://github.com/microsoft/call-center-ai/blob/main/_autodocs/00-index.md

Use `curl` to send a POST request to the `/call` endpoint to initiate a test call. Provide the necessary parameters such as phone numbers, bot name, and company.

```bash
curl -X POST http://localhost:8080/call \
  -H "Content-Type: application/json" \
  -d '{
    "phone_number": "+33612345678",
    "bot_name": "Amélie",
    "bot_company": "Contoso",
    "agent_phone_number": "+33698765432"
  }'
```

--------------------------------

### Check Service Readiness

Source: https://github.com/microsoft/call-center-ai/blob/main/_autodocs/01-rest-api.md

Use this endpoint to verify if the service is ready to handle requests, including checks for cache, database, search, and SMS services.

```bash
curl http://localhost:8080/health/readiness
```

--------------------------------

### Resources Configuration Settings

Source: https://github.com/microsoft/call-center-ai/blob/main/_autodocs/03-configuration.md

Configure resource paths for blob storage containers. Specifies URLs for sounds and call recordings.

```yaml
resources:
  sounds_container_url: "https://xxx.blob.core.windows.net/sounds"
  recordings_container_url: "https://xxx.blob.core.windows.net/recordings"
```

--------------------------------

### Get Resource Directory Path

Source: https://github.com/microsoft/call-center-ai/blob/main/_autodocs/06-helpers-and-utilities.md

Retrieves the absolute path to a specified resource directory. Useful for accessing templates or data files.

```python
from app.helpers.resources import resources_dir

template_dir = resources_dir("public_website")
```

--------------------------------

### GET /report/{call_id}

Source: https://github.com/microsoft/call-center-ai/blob/main/_autodocs/01-rest-api.md

Retrieves a detailed report for a specific call, identified by its UUID. The report is presented in an HTML web interface.

```APIDOC
## GET /report/{call_id}

### Description
Display a detailed call report in an HTML web interface.

### Method
GET

### Endpoint
/report/{call_id}

#### Path Parameters
- **call_id** (string) - Required - UUID of the call

### Response
#### Success Response (200 OK)
**Content-Type:** text/html
#### Error Response (404 Not Found)

### Request Example
```bash
curl 'http://localhost:8080/report/123e4567-e89b-12d3-a456-426614174000'
```
```

--------------------------------

### TrainingModel Excluded Fields Method

Source: https://github.com/microsoft/call-center-ai/blob/main/_autodocs/02-types.md

Instance method for TrainingModel to get a set of fields that should be excluded from LLM processing. Supports sorting by score.

```python
def excluded_fields_for_llm() -> set[str]:
    # Get fields to exclude from LLM (id, score)
    return {"id", "score"}
```

--------------------------------

### Load and Validate Configuration

Source: https://github.com/microsoft/call-center-ai/blob/main/_autodocs/06-helpers-and-utilities.md

Load configuration from environment variables or a YAML file. The configuration is validated against a RootModel. Raises ValueError if invalid or not found.

```python
from app.helpers.config import load_config

config = load_config()
```

--------------------------------

### load_config()

Source: https://github.com/microsoft/call-center-ai/blob/main/_autodocs/06-helpers-and-utilities.md

Loads and validates configuration from environment variables or a configuration file. Raises an error if the configuration is invalid or not found.

```APIDOC
## load_config()

### Description
Load and validate configuration from environment or file. Returns validated configuration.

### Signature
```python
def load_config() -> RootModel
```

### Returns
Validated `RootModel`

### Sources (in order)
1. `CONFIG_JSON` environment variable (JSON string)
2. `config.yaml` file (YAML)

### Raises
`ValueError` if config is invalid or not found
```

--------------------------------

### Load JSON Configuration from Environment Variable

Source: https://github.com/microsoft/call-center-ai/blob/main/_autodocs/03-configuration.md

Sets the entire application configuration using a single environment variable `CONFIG_JSON` containing a JSON string. This method offers the highest precedence for configuration loading.

```python
# Single CONFIG_JSON env var with entire config as JSON
CONFIG_JSON='{"ai_search": {...}, "database": {...}}'
```

--------------------------------

### start_audio_streaming

Source: https://github.com/microsoft/call-center-ai/blob/main/_autodocs/06-helpers-and-utilities.md

Initiates the capture of audio from an ongoing call. Requires the call automation client and the call connection ID.

```APIDOC
## start_audio_streaming()

Begin capturing audio from call.

### Method Signature
```python
async def start_audio_streaming(
    client: CallAutomationClient,
    call_connection_id: str,
) -> None
```

### Parameters
- **client** (CallAutomationClient) - The client for call automation.
- **call_connection_id** (str) - The ID of the call connection.
```

--------------------------------

### Get All Calls Report (Filtered)

Source: https://github.com/microsoft/call-center-ai/blob/main/_autodocs/01-rest-api.md

Use this endpoint to retrieve an HTML report of all calls, optionally filtered by a specific phone number. The phone number should be URL-encoded.

```bash
curl 'http://localhost:8080/report?phone_number=%2B33612345678'
```

--------------------------------

### Load YAML Configuration from File

Source: https://github.com/microsoft/call-center-ai/blob/main/_autodocs/03-configuration.md

Loads configuration from a `config.yaml` file located in the project root or current directory. This method is used when environment variables are not sufficient or for more structured configuration.

```yaml
# config.yaml
ai_search:
  endpoint: https://example.search.windows.net
  api_key: xxx
database:
  ...
```

--------------------------------

### Initialize Content Safety Blocklist Client

Source: https://github.com/microsoft/call-center-ai/blob/main/examples/blocklist.ipynb

Initializes the BlocklistClient with your Azure Content Safety endpoint and API key. Ensure you have the necessary Azure credentials.

```python
import pandas as pd
from azure.ai.contentsafety import BlocklistClient
from azure.ai.contentsafety.models import (
    AddOrUpdateTextBlocklistItemsOptions,
    TextBlocklist,
    TextBlocklistItem,
)
from azure.core.credentials import AzureKeyCredential

key = AzureKeyCredential("xxx")
client = BlocklistClient(
    "https://ccai-clesne-contentsafety.cognitiveservices.azure.com/", key
)
```

--------------------------------

### Configure Azure Storage Queues

Source: https://github.com/microsoft/call-center-ai/blob/main/_autodocs/03-configuration.md

Set up Azure Storage Queues for asynchronous processing of call events, post-call tasks, SMS events, and training data.

```yaml
queue:
  connection_string: "DefaultEndpointsProtocol=https;..."
  call_queue_name: "call-events"
  post_queue_name: "post-events"
  sms_queue_name: "sms-events"
  training_queue_name: "training-events"
```

--------------------------------

### Get Specific Call Report

Source: https://github.com/microsoft/call-center-ai/blob/main/_autodocs/01-rest-api.md

Retrieve a detailed HTML report for a specific call using its UUID. This endpoint returns a 404 Not Found if the call ID is invalid.

```bash
curl 'http://localhost:8080/report/123e4567-e89b-12d3-a456-426614174000'
```

--------------------------------

### Sync Local Configuration from Azure

Source: https://github.com/microsoft/call-center-ai/blob/main/README.md

Copies the application configuration from a remote Azure deployment to your local machine. Replace 'my-rg-name' with your resource group name.

```zsh
make name=my-rg-name sync-local-config
```

--------------------------------

### Fallback LLM Model Strategy

Source: https://github.com/microsoft/call-center-ai/blob/main/_autodocs/07-advanced-topics.md

This illustrates a fallback strategy where a faster, less capable LLM is tried first with a short timeout. If it fails, a slower, more capable LLM is attempted with a longer timeout. If both fail, an error message is returned.

```text
Try fast LLM (gpt-4.1-mini) with 4-second timeout
  ↓ fails
Try slow LLM (gpt-4.1) with 15-second timeout
  ↓ fails
Return error message to caller
```

--------------------------------

### Deploy Azure Resources Locally

Source: https://github.com/microsoft/call-center-ai/blob/main/README.md

Deploys the necessary Azure resources for the solution without the API server, enabling local testing. Replace 'my-rg-name' with your resource group name.

```zsh
make deploy-bicep deploy-post name=my-rg-name
```

--------------------------------

### Call Get Model

Source: https://github.com/microsoft/call-center-ai/blob/main/_autodocs/01-rest-api.md

Represents the complete call data returned by the API. Includes call ID, timestamps, status, initiation parameters, gathered claims, messages, and potential next actions or reminders.

```python
class CallGetModel(BaseModel):
    call_id: UUID  # Immutable
    created_at: datetime  # Immutable
    in_progress: bool  # Whether the call is actively ongoing
    initiate: CallInitiateModel  # Call initiation parameters
    claim: dict[str, Any]  # Gathered claim data
    messages: list[MessageModel]  # Conversation messages
    next: NextModel | None  # Recommended action after call
    reminders: list[ReminderModel]  # Generated reminders
    synthesis: SynthesisModel | None  # Call summary
```

--------------------------------

### Check Service Liveness

Source: https://github.com/microsoft/call-center-ai/blob/main/_autodocs/01-rest-api.md

Use this endpoint to perform a basic liveness probe and check if the service is running.

```bash
curl http://localhost:8080/health/liveness
```

--------------------------------

### Call Model Instance Methods

Source: https://github.com/microsoft/call-center-ai/blob/main/_autodocs/01-rest-api.md

Provides utility methods for interacting with call data, such as retrieving RAG documents, determining timezone, identifying assistant voice style, and checking for human interaction.

```python
    trainings(cache_only: bool)
    tz()
    last_assistant_style()
    had_interaction()
```

--------------------------------

### Call Center AI Data Structure Example

Source: https://github.com/microsoft/call-center-ai/blob/main/README.md

This JSON object represents the data stored during a call, including claim details, conversation messages, and reminders. It is used to track incident information and customer interactions.

```json
{
  "claim": {
    "incident_description": "Collision avec un autre véhicule, voiture dans le fossé, pas de blessés",
    "incident_location": "Nationale 17",
    "involved_parties": "Dujardin, Madame Lesné",
    "policy_number": "DEC1748"
  },
  "messages": [
    {
      "created_at": "2024-12-10T15:51:04.566727Z",
      "action": "talk",
      "content": "Non, je pense que c'est pas mal. Vous avez répondu à mes questions et là j'attends la dépaneuse. Merci beaucoup.",
      "persona": "human",
      "style": "none",
      "tool_calls": []
    },
    {
      "created_at": "2024-12-10T15:51:06.040451Z",
      "action": "talk",
      "content": "Je suis ravi d'avoir pu vous aider! Si vous avez besoin de quoi que ce soit d'autre, n'hésitez pas à nous contacter. Je vous souhaite une bonne journée et j'espère que tout se passera bien avec la dépanneuse. Au revoir!",
      "persona": "assistant",
      "style": "none",
      "tool_calls": []
    }
  ],
  "next": {
    "action": "case_closed",
    "justification": "The customer has provided all necessary information for the insurance claim, and a reminder has been set for a follow-up call. The customer is satisfied with the assistance provided and is waiting for the tow truck. The case can be closed for now."
  },
  "reminders": [
    {
      "created_at": "2024-12-10T15:50:09.507903Z",
      "description": "Rappeler le client pour faire le point sur l'accident et l'avancement du dossier.",
      "due_date_time": "2024-12-11T14:30:00",
      "owner": "assistant",
      "title": "Rappel client sur l'accident"
    }
  ],
  "synthesis": {
    "long": "During our call, you reported an accident involving your vehicle on the Nationale 17. You mentioned that there were no injuries, but both your car and the other vehicle ended up in a ditch. The other party involved is named Dujardin, and your vehicle is a 4x4 Ford. I have updated your claim with these details, including the license plates: yours is U837GE and the other vehicle's is GA837IA. A reminder has been set for a follow-up call tomorrow at 14:30 to discuss the progress of your claim. If you need further assistance, please feel free to reach out.",
    "satisfaction": "high",
    "short": "the accident on Nationale 17",
    "improvement_suggestions": "To improve the customer experience, it would be beneficial to ensure that the call connection is stable to avoid interruptions. Additionally, providing a clear step-by-step guide on what information is needed for the claim could help streamline the process and reduce any confusion for the customer."
  }
  ...
}
```

--------------------------------

### Async Job Scheduler

Source: https://github.com/microsoft/call-center-ai/blob/main/_autodocs/06-helpers-and-utilities.md

Obtain or create an asynchronous job scheduler for the current context. Use it within an async context manager to spawn background tasks.

```python
from app.helpers.scheduler import get_scheduler

async with get_scheduler() as scheduler:
    await scheduler.spawn(background_task())
```

--------------------------------

### Automatic LLM Request Retries with Exponential Backoff

Source: https://github.com/microsoft/call-center-ai/blob/main/_autodocs/07-advanced-topics.md

This Python snippet demonstrates how to automatically retry LLM requests that fail. It uses exponential backoff to increase wait times between retries and stops after a specified number of attempts. Ensure the `retry` library is installed.

```python
from tenacity import retry, wait_exponential, stop_after_attempt

@retry(
    wait=wait_exponential(multiplier=1, min=2, max=10),
    stop=stop_after_attempt(3),
    reraise=True,
)
async def call_llm():
    ...
```

--------------------------------

### Configure TTS and LLM Prompts

Source: https://github.com/microsoft/call-center-ai/blob/main/_autodocs/03-configuration.md

Customize Text-to-Speech (TTS) and Large Language Model (LLM) prompts using templates. Supports various placeholders for dynamic content.

```yaml
prompts:
  tts:
    hello_tpl:
      - "Hello, I'm {bot_name}!"
      - "Hi, I'm {bot_name} from {bot_company}"
  llm:
    default_system_tpl: |
      Assistant is called {bot_name}...
    chat_system_tpl: |
      # Objective
      Provide support...
```

--------------------------------

### IStore.readiness

Source: https://github.com/microsoft/call-center-ai/blob/main/_autodocs/04-persistence-api.md

Validates the health of the database by performing a CRUD test. Returns ReadinessEnum.OK if healthy, otherwise ReadinessEnum.FAIL.

```APIDOC
## readiness()

### Description
Validate that the database is healthy (performs CRUD test).

### Returns
`ReadinessEnum.OK` or `ReadinessEnum.FAIL`

### Example
```python
status = await db.readiness()
assert status == ReadinessEnum.OK
```
```

--------------------------------

### Run Local Tests

Source: https://github.com/microsoft/call-center-ai/blob/main/_autodocs/00-index.md

Execute local tests for the LLM and workflows without requiring Azure Communication Services.

```bash
python -m tests.local
```

--------------------------------

### Audio Processing Pipeline Flowchart

Source: https://github.com/microsoft/call-center-ai/blob/main/_autodocs/05-workflows-and-handlers.md

Details the steps involved in processing audio from raw input to streamed output.

```text
Raw audio (WebSocket)
  ↓
Speech-to-text (Azure Cognitive Services)
  ↓
Text → Message model
  ↓
LLM chat completion (Azure OpenAI)
  ↓
Extract style (voice tone) and tool calls
  ↓
Text-to-speech (Azure Cognitive Services)
  ↓
Audio → WebSocket (back to caller)
```

--------------------------------

### TTS Hello Prompt Templates

Source: https://github.com/microsoft/call-center-ai/blob/main/README.md

Customize the initial greeting prompts for Text-to-Speech (TTS). These templates use placeholders like {bot_name} and {bot_company} and are provided as a list to offer variety.

```yaml
prompts:
  tts:
    hello_tpl:
      - : |
        Hello, I'm {bot_name}, from {bot_company}! I'm an IT support specialist.

        Here's how I work: when I'm working, you'll hear a little music; then, at the beep, it's your turn to speak. You can speak to me naturally, I'll understand.

        What's your problem?
      - : |
        Hi, I'm {bot_name} from {bot_company}. I'm here to help.

        You'll hear music, then a beep. Speak naturally, I'll understand.

        What's the issue?
```

--------------------------------

### Azure CLI Login

Source: https://github.com/microsoft/call-center-ai/blob/main/README.md

Log in to your Azure environment using the Azure CLI. This is a prerequisite for deploying Azure resources.

```zsh
az login
```

--------------------------------

### Initiate Outbound Call

Source: https://github.com/microsoft/call-center-ai/blob/main/_autodocs/01-rest-api.md

Initiate a new outbound call by providing the target phone number, bot details, agent transfer number, and optional task or claim schema.

```bash
curl -X POST http://localhost:8080/call \
  -H "Content-Type: application/json" \
  -d '{
    "phone_number": "+33612345678",
    "bot_name": "Amélie",
    "bot_company": "Contoso",
    "agent_phone_number": "+33698765432",
    "task": "Help the customer with their insurance claim",
    "claim": [
      {
        "name": "policy_number",
        "type": "text",
        "description": "Policy number"
      }
    ]
  }'
```

--------------------------------

### POST /call

Source: https://github.com/microsoft/call-center-ai/blob/main/_autodocs/01-rest-api.md

Initiates a new outbound call to a specified phone number, configuring the AI assistant and call objective.

```APIDOC
## POST /call

### Description
Initiate a new outbound call to a phone number.

### Method
POST

### Endpoint
/call

### Parameters
#### Request Body
- **phone_number** (string (E164)) - Required - Target phone number (E164 format, e.g., `+33612345678`)
- **bot_name** (string) - Required - Name of the AI assistant
- **bot_company** (string) - Required - Company name for the assistant
- **agent_phone_number** (string (E164)) - Required - Phone number for transferring to human agent
- **task** (string) - Optional - Description of the call objective
- **claim** (array) - Optional - Custom claim schema for this call (array of field definitions)

### Request Example
```bash
curl -X POST http://localhost:8080/call \
  -H "Content-Type: application/json" \
  -d 
  {
    "phone_number": "+33612345678",
    "bot_name": "Amélie",
    "bot_company": "Contoso",
    "agent_phone_number": "+33698765432",
    "task": "Help the customer with their insurance claim",
    "claim": [
      {
        "name": "policy_number",
        "type": "text",
        "description": "Policy number"
      }
    ]
  }
```

### Response
#### Success Response (201 Created)
- **CallGetModel** (object) - See [Call Model](#call-model) for details.
```

--------------------------------

### Handle Text-to-Speech via Azure Automation

Source: https://github.com/microsoft/call-center-ai/blob/main/_autodocs/06-helpers-and-utilities.md

Plays pre-synthesized speech using the Azure Automation API. Supports custom voice styles and languages.

```python
async def handle_automation_tts(
    client: CallAutomationClient,
    call_connection_id: str,
    text: str,
    style: StyleEnum = StyleEnum.NONE,
    language: str = "fr-FR",
) -> None:
    ...
```

--------------------------------

### Deploy Call Center AI

Source: https://github.com/microsoft/call-center-ai/blob/main/README.md

Run the deployment automation for the Call Center AI project. Specify the resource group name. The image version can also be specified.

```zsh
make deploy name=my-rg-name
```

--------------------------------

### LLM Configuration for Azure OpenAI

Source: https://github.com/microsoft/call-center-ai/blob/main/_autodocs/03-configuration.md

Configure primary and fallback Azure OpenAI endpoints, including API keys, deployment names, and token limits. Use the 'fast' configuration for general use and 'slow' for fallback scenarios.

```yaml
llm:
  fast:
    api_key: "xxxx"
    deployment: "gpt-4-1-mini"
    endpoint: "https://xxx.openai.azure.com"
    max_tokens: 2000
  slow:
    api_key: "xxxx"
    deployment: "gpt-4-1"
    endpoint: "https://yyy.openai.azure.com"
    max_tokens: 4000
```

--------------------------------

### RAG Query Expansion Logic

Source: https://github.com/microsoft/call-center-ai/blob/main/_autodocs/07-advanced-topics.md

Illustrates the process of expanding queries for RAG by using recent messages, combining results, and filtering by relevance.

```text
Last 3 messages: ["Help with claim", "Policy number?", "ABC123"]
  ↓
Search AI Search with each message as query
  ↓
Combine and deduplicate results
  ↓
Sort by relevance score
  ↓
Filter by strictness threshold (e.g., 0.5)
  ↓
Inject top K results (default: 5) into LLM context
```

--------------------------------

### Root Configuration Model Definition

Source: https://github.com/microsoft/call-center-ai/blob/main/_autodocs/03-configuration.md

Defines the main `RootModel` for application configuration, inheriting from `BaseSettings`. It specifies all required and optional configuration fields with their types and default values.

```python
class RootModel(BaseSettings):
    public_domain: str  # Required
    version: str  # Default: "0.0.0-unknown"
    ai_search: AiSearchModel
    ai_translation: AiTranslationModel
    cache: CacheModel
    cognitive_service: CognitiveServiceModel
    communication_services: CommunicationServicesModel
    database: DatabaseModel
    llm: LlmModel
    monitoring: MonitoringModel
    prompts: PromptsModel
    resources: ResourcesModel
    sms: SmsModel
    conversation: ConversationModel
    app_configuration: AppConfigurationModel
    queue: QueueModel
```

--------------------------------

### CallInitiateModel Definition

Source: https://github.com/microsoft/call-center-ai/blob/main/_autodocs/02-types.md

Defines the parameters required to initiate a phone call. It includes the target phone number and inherits other initiation details from WorkflowInitiateModel.

```python
class CallInitiateModel(WorkflowInitiateModel):
    phone_number: PhoneNumber
```

--------------------------------

### ReadinessModel Definition

Source: https://github.com/microsoft/call-center-ai/blob/main/_autodocs/02-types.md

Defines the overall readiness status of the application, including a list of individual check results. Use this model to represent the health status of the entire system.

```python
class ReadinessModel(BaseModel):
    status: ReadinessEnum
    checks: list[ReadinessCheckModel]
```

--------------------------------

### Create and Update Text Blocklist from CSV

Source: https://github.com/microsoft/call-center-ai/blob/main/examples/blocklist.ipynb

Reads words from a 'blocklist.csv' file and uses them to create or update a text blocklist. It first creates the blocklist if it doesn't exist, then adds the items. Assumes the CSV has 'word' and 'blocklist' columns.

```python
name = "competitors"
description = "Competitors blocklist."

df = pd.read_csv("blocklist.csv")

block_items = {}
for index, row in df.iterrows():
    text = TextBlocklistItem(text=row["word"])
    if row["blocklist"] not in block_items:
        block_items[row["blocklist"]] = []
    block_items[row["blocklist"]].append(text)

for blocklist, words in block_items.items():
    print(f"Creating blocklist {blocklist} with {len(words)} words")  # noqa: T201
    client.create_or_update_text_blocklist(
        blocklist_name=blocklist,
        options=TextBlocklist(blocklist_name=blocklist),
    )
    client.add_or_update_blocklist_items(
        blocklist_name=blocklist,
        options=AddOrUpdateTextBlocklistItemsOptions(blocklist_items=words),
    )
```

--------------------------------

### Set Nested Environment Variables for Configuration

Source: https://github.com/microsoft/call-center-ai/blob/main/_autodocs/03-configuration.md

Configures nested settings using environment variables with a double underscore (`__`) delimiter. This allows for granular control over hierarchical configuration parameters.

```bash
export LLM__FAST__ENDPOINT=https://xxx.openai.azure.com
export LLM__SLOW__ENDPOINT=https://yyy.openai.azure.com
```

--------------------------------

### Check Database Readiness

Source: https://github.com/microsoft/call-center-ai/blob/main/_autodocs/04-persistence-api.md

Validates database health by performing a CRUD test. Returns ReadinessEnum.OK or ReadinessEnum.FAIL.

```python
status = await db.readiness()
assert status == ReadinessEnum.OK
```

--------------------------------

### Feature Flag Functions

Source: https://github.com/microsoft/call-center-ai/blob/main/_autodocs/06-helpers-and-utilities.md

Provides access to dynamic configuration settings from Azure App Configuration. These functions return integer, float, or boolean values and have a cache TTL of 60 seconds.

```APIDOC
## Feature Flag Functions

Provides access to dynamic configuration settings from Azure App Configuration.

### Functions
- `answer_hard_timeout_sec()`
- `answer_soft_timeout_sec()`
- `callback_timeout_hour()`
- `phone_silence_timeout_sec()`
- `recognition_retry_max()`
- `recording_enabled()`
- `slow_llm_for_chat()`

### Return Type
All functions return `int`, `float`, or `bool` depending on the flag's configuration.

### Cache
Features flags are cached with a Time-To-Live (TTL) of 60 seconds, refreshing every minute.
```

--------------------------------

### IStore.call_create

Source: https://github.com/microsoft/call-center-ai/blob/main/_autodocs/04-persistence-api.md

Creates a new call record in the database. The returned CallStateModel may include server-assigned identifiers.

```APIDOC
## call_create(call: CallStateModel)

### Description
Create a new call.

### Parameters
#### Request Body
- **call** (CallStateModel) - Required - New call data

### Returns
Created `CallStateModel` (may include server-assigned IDs)

### Example
```python
call = CallStateModel(
    initiate=CallInitiateModel(
        phone_number="+33612345678",
        bot_name="Amélie",
        bot_company="Contoso",
        agent_phone_number="+33698765432",
    )
)
created = await db.call_create(call)
print(created.call_id)
```
```

--------------------------------

### Call Lifecycle Flowchart

Source: https://github.com/microsoft/call-center-ai/blob/main/_autodocs/05-workflows-and-handlers.md

Illustrates the sequence of events from an incoming call to post-call processing.

```text
User dials → Incoming call event → Answer call → Media streaming starts
  ↓
Audio captured → Speech-to-text → LLM processing → Text-to-speech → Play audio
  ↓
[Loop until hangup]
  ↓
Call ends → Post-call processing (summary, reminders, next action) → Database update
```

--------------------------------

### Define Language Model

Source: https://github.com/microsoft/call-center-ai/blob/main/_autodocs/02-types.md

Use this model to configure overall language settings, specifying the default language and a list of available languages.

```python
class LanguageModel(BaseModel):
    default_short_code: str
    availables: list[LanguageEntryModel]
```

--------------------------------

### Map Nested Environment Variables to YAML Structure

Source: https://github.com/microsoft/call-center-ai/blob/main/_autodocs/03-configuration.md

Illustrates how nested environment variables with `__` delimiters are mapped to a hierarchical YAML configuration structure. This demonstrates the interpretation of the environment variables set previously.

```yaml
llm:
  fast:
    endpoint: https://xxx.openai.azure.com
  slow:
    endpoint: https://yyy.openai.azure.com
```

--------------------------------

### Monitoring Configuration Settings

Source: https://github.com/microsoft/call-center-ai/blob/main/_autodocs/03-configuration.md

Configure Azure Monitor/Application Insights for telemetry. Includes enablement and sampling rate. Requires APPLICATIONINSIGHTS_CONNECTION_STRING environment variable.

```yaml
monitoring:
  enabled: true
  sample_rate: 1.0
```

--------------------------------

### App Configuration Settings

Source: https://github.com/microsoft/call-center-ai/blob/main/_autodocs/03-configuration.md

Configure Azure App Configuration service for feature flags. Includes connection string and TTL for feature flag refresh.

```yaml
app_configuration:
  connection_string: "Endpoint=https://xxx.azconfig.io;..."
  ttl_sec: 60
```

--------------------------------

### Processing Call Data from Training Queue

Source: https://github.com/microsoft/call-center-ai/blob/main/_autodocs/07-advanced-topics.md

Shows how to process call data from an Azure Queue Storage message for RAG training. It deserializes the JSON content and initiates the training process.

```python
# training_event() processes it
async def training_event(training: AzureQueueStorageMessage):
    call = CallStateModel.model_validate_json(training.content)
    await call.trainings(cache_only=False)  # Pre-compute embeddings
```

--------------------------------

### Configure SMS Provider (Twilio)

Source: https://github.com/microsoft/call-center-ai/blob/main/_autodocs/03-configuration.md

Set the SMS mode to 'twilio' and provide Twilio account credentials, including Account SID, Auth Token, and the phone number.

```yaml
sms:
  mode: "twilio"
  twilio:
    account_sid: "ACxxx"
    auth_token: "xxx"
    phone_number: "+33612345678"
```

--------------------------------

### Implement Call Transfer to Agent in Python

Source: https://github.com/microsoft/call-center-ai/blob/main/_autodocs/07-advanced-topics.md

Implement the 'call_transfer' function to handle playing a message and transferring the call to a specified phone number. Requires CallAutomationClient and CallStateModel.

```python
async def call_transfer(
    message: str,
    call: CallStateModel,
    client: CallAutomationClient,
) -> bool:
    # Play message
    await handle_automation_tts(client, call.voice_id, message)
    
    # Transfer to agent
    transfer_result = await client.transfer_call_to_participant(
        target_participant=PhoneNumberIdentifier(call.initiate.agent_phone_number),
    )
    return True
```

--------------------------------

### Initiate Call to AI Agent

Source: https://github.com/microsoft/call-center-ai/blob/main/README.md

Use this cURL command to ask the AI bot to call a specific phone number. Ensure to replace 'xxx' with your actual API endpoint and provide all necessary data fields.

```bash
data='{
  "bot_company": "Contoso",
  "bot_name": "Amélie",
  "phone_number": "+11234567890",
  "task": "Help the customer with their digital workplace. Assistant is working for the IT support department. The objective is to help the customer with their issue and gather information in the claim.",
  "agent_phone_number": "+33612345678",
  "claim": [
    {
      "name": "hardware_info",
      "type": "text"
    },
    {
      "name": "first_seen",
      "type": "datetime"
    },
    {
      "name": "building_location",
      "type": "text"
    }
  ]
}'

curl \
  --header 'Content-Type: application/json' \
  --request POST \
  --url https://xxx/call \
  --data $data
```

--------------------------------

### MemoryCache Implementation

Source: https://github.com/microsoft/call-center-ai/blob/main/_autodocs/04-persistence-api.md

In-memory implementation of the ICache interface, suitable for local development. Not recommended for multi-process deployments.

```python
class MemoryCache(ICache):
    ...
```

--------------------------------

### Sending Call Data to Training Queue

Source: https://github.com/microsoft/call-center-ai/blob/main/_autodocs/07-advanced-topics.md

Demonstrates sending completed call data to an asynchronous training queue for RAG training. The data is serialized to JSON.

```python
# on_end_call() sends to training queue
await training_queue.send_message(call.model_dump_json())
```

--------------------------------

### AiSearchImpl Implementation

Source: https://github.com/microsoft/call-center-ai/blob/main/_autodocs/04-persistence-api.md

Implementation of the ISearch interface using Azure AI Search. It requires configuration for endpoint, API key, and index name.

```python
class AiSearchImpl(ISearch):
    def __init__(self, cache: ICache, config: AiSearchModel):
        ...
```