Skip to content

chore: changelog for 2025-08-20 — major API update: handoff tool, rejection plans, security, metrics #19

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Closed
quantstruct-canvas wants to merge 1 commit into from
Closed

chore: changelog for 2025-08-20 — major API update: handoff tool, rejection plans, security, metrics #19

Changes from all commits
Commits
Show all changes
1 commit
Select commit
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
73 changes: 73 additions & 0 deletions fern/changelog/2025-08-20.mdx
View file
Open in desktop
Original file line number Diff line number Diff line change
@@ -0,0 +1,73 @@
---
title: "2025-08-20: Major API Updates — Handoff Tool, Rejection Plans, Metrics, Security, and More"
date: 2025-08-20
---

<Card>
<Check/>
<b>Major OpenAPI Release: Multi-Agent Handoff, Tool Rejection, Security, and Observability</b>
</Card>

## 🚀 New Features & Additions

### Multi-Agent Handoff Tool
- **New `HandoffTool` and related endpoints**: Seamlessly hand off conversations between agents (assistants) using the new `handoff` tool type. Handoffs can be configured for:
- <b>Single or multiple static destinations</b> (by assistant ID or name)
- <b>Dynamic destinations</b> via server webhook, including custom parameters (e.g., customer intent, sentiment)
- <b>Use case:</b> Enables advanced multi-agent orchestration, escalations, and dynamic call routing in complex workflows.

### Tool Rejection Plans
- **All tools now support a `rejectionPlan`**: Define conditions under which a tool call should be rejected (e.g., user didn't say goodbye, bot is looping, or user intent mismatch).
- <b>Flexible logic:</b> Supports regex, Liquid templates, and logical groups for advanced conversation state handling.
- <b>Use case:</b> Prevents accidental tool calls, enables smarter, context-aware flows, and reduces conversational errors.

### Expanded Voice Capabilities
- **New `MinimaxVoice` support**: Configure Minimax TTS with options for model, pitch, speed, region, volume, emotion, and caching.
- **Voice fallback plans**: Enhanced failover across voice providers, including Minimax and Inworld.

### Performance Metrics & Observability
- **`call.artifact.performanceMetrics`**: Track detailed turn-level latency for model, voice, endpointing, and transcription.
- **`AnalysisCostBreakdown`**: Now includes structured output token usage and costs.
- **Support for Langfuse and improved logging/recording paths in `ArtifactPlan`**.

### Security & Compliance
- **Security filter plans**: New `securityFilterPlan` in compliance configuration allows for transcript sanitization, rejection, or replacement on threats (e.g., SQLi, XSS, prompt injection).
- **User message filtering**: Each message now tracks `isFiltered`, `detectedThreats`, and `originalMessage`.
- **RBAC and retention controls**: Subscription objects now include `rbacEnabled`, `callRetentionDays`, and `chatRetentionDays`.

### Workflow & Assistant Enhancements
- **Workflow-level hooks, background sound, keypad input plans, and dynamic credentials** (including Minimax credentials).
- **Custom LLM models and headers**: Workflows and assistants can now specify custom model endpoints and headers.
- **More granular artifact and analysis configuration**.

## ⚠️ Breaking Changes & Deprecations

### Major Deprecations
- **Tool `function` property removed**: All custom logic for tool execution must now be handled via webhooks or rejection plans. Update your tool configurations accordingly.
- **Removed legacy support endpoints**: `/support/ticket`, `/workflow/generate`, `/campaign/paginated` and discriminator mappings for `ghl`, `make`, `output` have been removed.
- **Removed properties from core objects:**
- `Assistant`: `messagePlan`, `silenceTimeoutSeconds`, `backgroundDenoisingEnabled`
- `Artifact`/`NodeArtifact`: `variables`
- `AssemblyAITranscriber`: `enableUniversalStreamingApi`
- Tool DTOs: All `function` properties

### Migration Guidance
- <Card>
<b>Update all tool definitions to use <code>rejectionPlan</code> for conditional logic.</b><br/>
Replace removed `function` fields with webhook-based or plan-based logic.
<br/><br/>
<b>Review assistant and workflow schemas:</b><br/>
Remove deprecated properties and update to new compliance, hooks, and model options.
</Card>

## 📝 Documentation Updates Required

- **Tool configuration**: Update all tool examples and guides to use `rejectionPlan` instead of `function`.
- **Handoff flows**: Add new documentation and code samples for `handoff` tool and dynamic routing.
- **Security/compliance**: Document new `securityFilterPlan` and message filtering fields on user messages.
- **Performance metrics**: Add usage guides for turn latency and cost breakdown.
- **Assistant/workflow schemas**: Remove deprecated fields and highlight new credential, sound, and model options.

<Frame>
<b>Note:</b> This release introduces significant improvements for multi-agent orchestration, security, and observability, but also requires updates to existing tool and assistant configurations. Please review all breaking changes and update your integrations accordingly.
</Frame>
Loading