Add changelog for 2025-08-11

fern/changelog/2025-08-01.mdx

@@ -0,0 +1,35 @@
+# New Features & Enhancements
+
+1. **Enhanced Call [`Artifacts`](https://api.vapi.ai/api#:~:text=Artifact)**: You can now store detailed information about call workflows and outcomes using [`Artifact`](https://api.vapi.ai/api#:~:text=Artifact) objects. Key properties include:
+    - `nodes`: History of workflow nodes executed during the call.
+    - `messages`: All messages spoken during the call.
+    - `logUrl`: **New!** Direct URL to detailed call logs for debugging.
+    - `pcapUrl`: Packet capture URL for phone calls (provider: `vapi` or `byo-phone-number`).
+    - `recording`: Recording URL (requires `assistant.artifactPlan.recordingEnabled`).
+    - `transcript`: Convenient full call transcript.
+    - `variableValues`: Final workflow variable states.
+    - `messagesOpenAIFormatted`: Spoken messages, formatted for OpenAI.
+
+<Warning>
+    `recordingUrl`, `videoRecordingUrl`, `stereoRecordingUrl`, and `videoRecordingStartDelaySeconds` are now deprecated in favor of the new `recording` and related properties within the `Artifact` object.
+</Warning>
+
+2. **Improved Azure Speech Segmentation Tuning**: You can now fine-tune speech segmentation in your [`Azure Speech Transcriber`](https://api.vapi.ai/api#:~:text=AzureSpeechTranscriber) using `segmentationStrategy`, `segmentationMaximumTimeMs`, and `segmentationSilenceTimeoutMs` properties for better transcription control. This applies to both [`Azure Speech Transcriber`](https://api.vapi.ai/api#:~:text=AzureSpeechTranscriber) and [`Fallback Azure Speech Transcriber`](https://api.vapi.ai/api#:~:text=FallbackAzureSpeechTranscriber).
+
+<Warning>
+    The `segmentationMaxTimeMs` property in [`AzureSpeechTranscriber`](https://api.vapi.ai/api#:~:text=AzureSpeechTranscriber) has been replaced by `segmentationMaximumTimeMs` for consistency.
+</Warning>
+
+## Deprecations
+<CardGroup cols={3}>
+  <Card title="Assistant & Related Objects" icon="cog">
+    <code>messagePlan</code> and <code>backgroundDenoisingEnabled</code> are now part of <code>Assistant.backgroundSpeechDenoisingPlan</code> instead of <code>Assistant</code> and <code>AssistantOverrides</code> directly.
+  </Card>
+  <Card title="Org" icon="building">
+    <code>stripeCustomerId</code> is now part of <code>Subscription</code> instead of <code>Org</code>.
+  </Card>
+  <Card title="AzureSpeechTranscriber" icon="brands microsoft">
+    <code>segmentationMaxTimeMs</code> property has been renamed to <code>segmentationMaximumTimeMs</code> for consistency.
+  </Card>
+</CardGroup>
+

fern/changelog/2025-08-02.mdx

@@ -0,0 +1,3 @@
+# Azure Speech Segmentation Strategy Tuning
+
+1. You can now tune `Assistant.transcriber[provider="AzureSpeechTranscriber"].segmentationStrategy` using `"Default"`, `"Time"`, and `"Semantic"` strategies, offering more control over phrase boundary detection.

fern/changelog/2025-08-03.mdx

@@ -0,0 +1,28 @@
+
+#### 🎤 New Gladia Transcription Provider Support
+1. **Custom vocabulary support**: Enable a custom vocabulary with [`Gladia`](https://api.vapi.ai/api#:~:text=GladiaTranscriber) using `Assistant.transcriber[provider="GladiaTranscriber"].customVocabularyEnabled`. You can also specify simple strings or detailed objects with fields for value, language, intensity, and alternative pronunciations using `Assistant.transcriber[provider="GladiaTranscriber"].customVocabularyConfig` - letting you fine-tune recognition of domain-specific terms.
+
+2. **Endpointing & Speech Threshold**: Configure endpointing time (wait time before considering speech ended) and speech sensitivity, enabling more accurate and responsive transcription with `Assistant.transcriber[provider="GladiaTranscriber"].endpointing` and `Assistant.transcriber[provider="GladiaTranscriber"].speechThreshold`.
+
+3. **Prosody & Audio Enhancer**: Optionally enable prosody (for transcribing non-verbal cues like laughter) and audio enhancement for improved accuracy with `Assistant.transcriber[provider="GladiaTranscriber"].prosodyEnabled` and `Assistant.transcriber[provider="GladiaTranscriber"].audioEnhancerEnabled`.
+
+4. **Flexible Language Detection**: Choose between manual and automatic language detection modes with `Assistant.transcriber[provider="GladiaTranscriber"].languageDetectionMode`.
+
+5. **Confidence Thresholds & Hints**: Discard low-confidence transcripts and provide context hints for improved accuracy with `Assistant.transcriber[provider="GladiaTranscriber"].confidenceThreshold` and `Assistant.transcriber[provider="GladiaTranscriber"].hints`.
+
+## 💳 Subscription Updates
+
+<CardGroup cols={3}>
+  <Card title="RBAC" icon="user-shield">
+    <b>Role-based access control (RBAC):</b><br />
+    Enable RBAC for your subscription using <code>Subscription.rbacEnabled</code>.
+  </Card>
+  <Card title="Call & Chat Retention" icon="clock-rotate-left">
+    <b>Retention settings:</b><br />
+    Configure how long calls and chats are stored with <code>Subscription.callRetentionDays</code> and <code>Subscription.chatRetentionDays</code>.
+  </Card>
+  <Card title="Minutes Included Reset Frequency" icon="rotate">
+    <b>Reset frequency:</b><br />
+    Set how often included minutes reset using <code>Subscription.minutesIncludedResetFrequency</code>.
+  </Card>
+</CardGroup>

fern/changelog/2025-08-08.mdx

@@ -0,0 +1,29 @@
+# New: Smarter Conditions & Security Filters
+
+1. **New Condition & Filter Types**: You can now use the following new condition and filter types to build more robust rejection plans and security filter plans:
+- **MessageTarget**: Target specific messages by role and position for conditions using `Assistant.hooks.do[type=tool].tool.rejectionPlan.conditions[type=regex].target`.
+- **GroupCondition**: Combine multiple conditions using AND/OR logic, with support for recursive nesting using `Assistant.hooks.do[type=tool].tool.rejectionPlan.conditions[type=group]`.
+- **RegexCondition**: Flexible pattern matching, with full support for JavaScript regex and negation using `Assistant.hooks.do[type=tool].tool.rejectionPlan.conditions[type=regex]`. 
+- **LiquidCondition**: Use Liquid templates for complex, context-aware logic using `Assistant.hooks.do[type=tool].tool.rejectionPlan.conditions[type=liquid]`.
+- **Security Filters**: New filter types for RCE, XSS, SSRF, SQL injection, prompt injection, and regex-based filtering using `Assistant.compliancePlan.securityFilterPlan.filters`.
+
+2. **Tool Rejection Plans**: You can now use [`Assistant.hooks.do[type=tool].tool.rejectionPlan`](https://api.vapi.ai/api#:~:text=ToolRejectionPlan) in all tool calls to prevent accidental tool execution, enforce confirmation steps, and build more robust conversation flows. This helps you to define complex logic for when a tool call should be rejected, enhancing both safety and call experience. Rejection plans can be built using regex conditions, [Liquid templates](https://liquidjs.com/), or logical groups (AND/OR). For example, you can prevent an `endCall` tool from executing unless the user says goodbye, or block a transfer if the user is actually asking a question.
+
+**Example:**
+```json
+{
+  "conditions": [
+    {
+      "type": "regex",
+      "regex": "(?i)\\b(bye|goodbye|farewell|see you later|take care)\\b",
+      "target": { "position": -1, "role": "user" },
+      "negate": true
+    }
+  ]
+}
+```
+
+3. **Security Filter Plans for Transcripts and Messages**: You can now use [`Assistant.compliancePlan.securityFilterPlan`](https://api.vapi.ai/api#:~:text=SecurityFilterPlan) to define how transcripts and messages are filtered against threats like SQL injection, XSS, prompt injection, and more. Choose between `sanitize`, `reject`, or `replace` when threats are detected, and specify custom replacement text. User messages and transcript objects now include:
+  - `isFiltered`: Indicates if content was filtered for security.
+  - `detectedThreats`: Lists detected threats.
+  - `originalMessage` / `originalTranscript`: Preserves original content if filtering occurred.

fern/changelog/2025-08-09.mdx

@@ -0,0 +1,57 @@
+## New: Call Metrics & Artifact Improvements
+
+You can now access detailed call performance metrics and structured output IDs directly from your call artifacts.
+
+<CardGroup cols={3}>
+  <Card title="Turn Latencies" icon="stopwatch">
+    <b>Each conversation turn's latency</b>
+    <br/>
+    <code>Call.artifact.performanceMetrics.turnLatencies</code>
+  </Card>
+  <Card title="Model Latency (avg)" icon="robot">
+    <b>Average time for the model to generate a response</b>
+    <br/>
+    <code>Call.artifact.performanceMetrics.modelLatencyAverage</code>
+  </Card>
+  <Card title="Voice Latency (avg)" icon="volume-up">
+    <b>Average time to synthesize voice</b>
+    <br/>
+    <code>Call.artifact.performanceMetrics.voiceLatencyAverage</code>
+  </Card>
+</CardGroup>
+
+<CardGroup cols={3}>
+  <Card title="Transcriber Latency (avg)" icon="wave-square">
+    <b>Average time to transcribe voice</b>
+    <br/>
+    <code>Call.artifact.performanceMetrics.transcriberLatencyAverage</code>
+  </Card>
+  <Card title="Endpointing Latency (avg)" icon="flag-checkered">
+    <b>Time to detect end of a conversation turn</b>
+    <br/>
+    <code>Call.artifact.performanceMetrics.endpointingLatencyAverage</code>
+  </Card>
+  <Card title="Turn Latency (avg)" icon="clock">
+    <b>Average latency to complete a conversation turn</b>
+    <br/>
+    <code>Call.artifact.performanceMetrics.turnLatencyAverage</code>
+  </Card>
+</CardGroup>
+
+<Card title="Structured Output IDs & Results" icon="list-ol" size="large">
+  <b>Track and extract structured outputs from your calls</b>
+  <ul>
+    <li>
+      <b>During call:</b> Access array of output IDs<br/>
+      <code>Call.artifactPlan.structuredOutputIds</code>
+    </li>
+    <li>
+      <b>After call:</b> Extracted outputs are stored here<br/>
+      <code>Call.artifact.structuredOutputs</code>
+    </li>
+  </ul>
+</Card>
+
+<Info>
+  These improvements help you monitor, debug, and analyze your calls with greater detail.
+</Info>