v3.1.0 v0.2.15
VectifyAI/PageIndexSep 3, 2026 → Sep 6, 20263 days between releases
What's New
Older release focuses on TTS; newer release introduces a new RAG indexing engine with Flash engine support.
Breaking Changes
- `extra_body` can no longer replace the SDK's own request rows (system, instructions, input, messages, tools).
New Features
- Flash engine for faster indexing with concurrent node expansion.
- Unified `PageIndexClient` supporting both local and cloud modes.
- Agent integration with MCP tools and native adapters.
- Streaming chat support for OpenAI Responses and Anthropic Messages.
- Typed configuration shapes (`IndexConfig`, `ChatConfig`).
Improvements
- Improved error handling and robust agent integration.
- Direct image support for models via MCP content.
Upgrade Notes
This represents a major architectural shift from a TTS library to a RAG indexing engine; no direct upgrade path exists. The `PageIndexClient` API is now unified for local and cloud modes, requiring configuration updates for cloud deployments. The `extra_body` parameter is deprecated for SDK-managed request fields.
The Two Releases
v3.1.0
Sep 3, 2026· 2 months ago
This release focuses on performance optimizations, including CPU-only installation support and speed improvements for H100 models. It introduces new community models for Hindi and Korean, refactors training infrastructure, and improves overall documentation and stability.
Full release notesv0.2.15
Sep 6, 2026· 2 months ago
This release introduces the 'Flash engine' for significantly faster indexing by removing LLMs from the structural generation process. It unifies local and cloud workflows into a single client interface while adding robust agent integration and improved error handling.
Full release notes