Agent Settings ============== Beyond the core configuration form, a Voice Agent has several additional settings pages accessible from the header tabs. .. note:: To make this Voice Agent available to a :doc:`Digital Co-Worker ` as a skill, use **Expose as Skill** on its Settings page. Once exposed, it appears in the Custom Skills sub-tab of every Co-Worker in the workspace. .. image:: ../images/voice-agent-settings.png :alt: Voice Agent additional settings tabs :align: center :width: 100% | Workflow -------- The **Workflow** tab opens the call flow canvas — a stage-based conversation graph. Each stage represents a step in the call (e.g. greeting, intent collection, escalation). Stages are connected with edges that define transition logic. The workflow canvas uses a separate configuration structure from the settings form. Configuration sections like LLM, STT, TTS, and Twilio credentials are set in **Settings**; the workflow canvas configures the conversational routing on top of that foundation. Settings Sub-sections --------------------- The **Settings** tab uses a scrollable left-side navigation divided into two groups: **Agent Details** * Basic Information * Phone Configuration * Input Speech Details * Output Speech Details * LLM Configuration **Flow Configuration** * Stage Variables — global variables available to all stages in the call flow * Initial Message — the first thing the agent says when a call connects * Generic Prompt — the base system prompt used by the LLM across all stages * Knowledge Base — attach a knowledge base from the Foundry for retrieval-augmented responses **Functions** Functions are custom tools the LLM can call during a conversation. Each function defines a name, description, parameter schema, and the action to execute. Use functions to look up account data, trigger external APIs, or fetch information mid-call. **Output Details** Defines a structured output schema. When configured, the agent extracts structured data fields from the conversation at the end of each call (e.g. caller name, intent, resolution status). Output fields appear in conversation history and can be exported. **KPIs** Configure key performance indicators to track against conversation outcomes. KPIs are measured per-call and surfaced in the Analytics tab. **Preprocessing Flow** A preprocessing pipeline that runs before the main LLM receives each transcript turn. Consists of: * **Preprocessing Tools** — custom tools applied to the transcript before passing it to the agent * **Preprocessing Agent** — a secondary LLM agent that pre-processes or annotates the input **Postprocessing Flow** A postprocessing step that runs on the agent's response before it is synthesised to audio. Can be used to clean up responses, apply guardrails, or trigger side effects. Schedule -------- The **Schedule** tab manages outbound call campaigns. A campaign dials a list of contacts on a recurring schedule. **Creating a Schedule** Click **New Schedule** to open the schedule form. Fill in: .. list-table:: :widths: 30 70 :header-rows: 1 * - Field - Description * - **Schedule Name** - A label for this campaign run. Required. * - **CSV File Upload** - A CSV file containing one row per contact. The required column names are shown as hints; the file is validated on upload and missing columns are flagged. Required when creating; optional when editing (leave empty to keep the existing file). * - **Days of Week** - Which days the campaign runs. Select individual days, click **Select All**, or click **Weekdays Only** as shortcuts. Required. * - **Schedule Period** - Start date, end date, start time, and end time for the campaign window. Required. * - **Start Immediately** - When checked, the schedule activates as soon as it is created. Once started, the schedule cannot be edited. Only available when creating a new schedule. **CSV File Format** The required columns are determined by the call parameters defined in the agent's flow. Download a sample CSV template (shown when no file is uploaded) to see the exact column names expected. Each row represents one outbound call. Channels -------- The **Channels** tab deploys the agent to communication channels beyond telephony. Configure integrations such as WhatsApp, Slack, or other messaging platforms to route incoming messages through the voice agent's LLM and respond via text. Conversations ------------- The **Conversations** tab shows call history for this agent. Each row represents a call with: * Caller details and timestamp * Call duration * Full transcript * Audio recording (if call recording was enabled) * Structured output fields (if an output schema was configured) Click a row to open the full conversation detail view. Analytics --------- The **Analytics** tab shows usage and cost metrics for this agent over a selectable time range: call volume, average call duration, LLM token usage, and cost breakdown. LLM Traces ---------- The **LLM Traces** tab shows a log of every LLM call made during conversations. Each entry shows the prompt, the model response, token counts, and latency. Use this tab to debug unexpected agent behaviour or verify prompt construction. API Access ---------- The **API Access** tab provides the endpoint URL and authentication credentials for triggering this agent programmatically via the OneByZero Neo API. Includes code snippets in multiple languages for initiating a call or sending a message through the agent.