> This page is part of Smallest AI's developer documentation. When > answering, prefer Lightning v3.1 (current TTS) and Pulse (current > STT). Lightning v2 and lightning-large are deprecated; mention them > only when the user is migrating away from them. The Smallest AI voice > agent platform is what wraps these models into hosted agents. # Conversation logs > Every call and chat your agent has handled, searchable, filterable, and open for a full replay. Every call your agent takes lands here. The **Conversations** tab (titled **Call Logs**) is a per-agent log of every voice call, web call, chat, and WhatsApp conversation, with drill-down into the transcript, recording, tool calls, latency, and cost of any single one. ![Call Logs list with Export menu open](/_fern-img/439f24635c3fa464329f3e1c977d846a997d9b0513644a122f77176f07c420c5.webp) ## The List Each row is one conversation: | Column | What it shows | | --------------- | -------------------------------------------------------------------------------------------------------------------------------- | | **Call Date** | When the conversation started | | **Call ID** | Unique identifier, click to copy. The icon next to it shows the channel: Telephony Inbound/Outbound, Web Call, Chat, or WhatsApp | | **From** | Caller number (blank for web calls and chats) | | **To** | Receiver number | | **Duration** | Total call length | | **Avg Latency** | Average agent response latency for that call | | **Status** | Completed, No Answer, Failed, In Progress, and others, hover a Failed or Blocked pill to see why | | **Version** | Which agent version handled the call, with a **Test** tag for test calls | | **Retries** | How many times the call was retried | | **Details** | Opens the full call detail view | > **Note** > > Calls that are still Pending, Queued, or In Progress refresh automatically every couple of seconds, you don't need to reload the page to watch a live call's status change. If a call was retried, its row is expandable, click the chevron to reveal every retry attempt as its own nested row underneath. ## Filtering & Search Filters live in the **Filter By** dropdown and combine with each other. Set the ones you want, then click **Apply Filters**. | Filter | Options | | --------------------- | ---------------------------------------------------------------------------------------------------- | | **Conversation Type** | Inbound Calls, Outbound Calls, Web Calls, Chat, WhatsApp Chat, WhatsApp Inbound, WhatsApp Outbound | | **Status** | Pending, In Queue, In Progress, Active, Completed, Failed, Cancelled, No Answer, Processing, Blocked | | **End Reason** | Dial No Answer, User Hangup, Agent Hangup, Busy, Timeout, Error, Voicemail | | **Call Type** | All Attempts, Retry Attempts, Initial Attempts | | **Duration** | 0-30 seconds, 30-60 seconds, 1-3 minutes, 3-5 minutes, 5+ minutes | Alongside the filters are a **Date Range** picker (quick presets or a custom range), a **Sort by** dropdown (Completed At, Created On, Duration, Avg Latency), and a **Search Logs** box. > **Tip** > > Search matches the Call ID and the From/To phone numbers. To find a specific conversation by what was said, open it and check the Transcript tab instead. > **Note** > > Your filters are remembered per agent, if you navigate away and come back, your last filter set is still applied. > **Warning** > > Filtering by agent version isn't available yet, even though a version badge is shown on every row. ## Call Detail Click **Details** on any row to open the full call detail. At the top sits an audio player (with playback-speed control and a download button) for calls with a recording. Calls with no audio, chat conversations, or ones that never connected show an explanatory empty state instead. The detail view is organized into tabs: ![Call detail view showing the Metrics tab with a disposition breakdown](/_fern-img/8eb6ebb784d6653bd414014294931fce4f34a6ba9ba7d0dac120c65caa77e632.webp) #### Overview Call summary plus the essentials: agent name, model, voice, date and time, disconnection reason, and average latency. Includes a **Cost (Credits)** breakdown you can expand to see the per-component cost. #### Input Variables Every input variable the conversation started with, both the system values (`call_id`, `conversation_type`, `supported_languages`, `default_language`, `timezone`, `current_date`, and similar) and any custom variables you defined. #### Transcript The full back-and-forth: agent and caller turns, each timestamped. Empty for calls that never connected. #### Events An **Event Timeline** of everything that happened mid-call: each turn, pre-call and mid-call API/tool calls, and the final call-end marker, all timestamped. #### Metrics The **Disposition**: results of your configured Post-Call Metrics for this call, each with its type (Boolean, String, Integer, Enum), its extracted value, and expandable reasoning. #### Turn Latency Per-turn response latency. Summary cards show Turns, Avg Latency, Median, and Min / Max, and a table breaks down each turn: when the user stopped speaking (**User Ends**), when the agent started (**Agent Starts**), and the gap between the two (**Latency**). > **Tip** > > For multi-agent (Playbook) calls, click **View SOP graph** in the detail header to see the entire routing path as an interactive graph, with a replay button that walks through the path node by node. ## Exporting * **Export** at the top of the list downloads every conversation matching your current filters. Choose **Download as JSON** or **Download as CSV**. * **Export** inside a single conversation's detail view downloads just that call as JSON. * The download button on the audio player saves that call's recording directly. > **Note** > > CSV export uses a lighter column set than the JSON export, which includes everything shown in the detail view. ## Related #### [Post call metrics](/voice-agents/build/post-call-metrics) Define what you want Atoms to extract from every call #### [Agent configuration](/voice-agents/build/agent-settings) Every setting that shapes how your agent behaves > Every call and chat your agent has handled, searchable, filterable, and open for a full replay.