v3.2.0 v0.2.16
VectifyAI/PageIndexSep 8, 2026 → Sep 10, 20262 days between releases
What's New
Introduces high-performance PageIndex Flash engine and unified client surface for local and cloud operations.
Breaking Changes
- `doc_id` removed from eight agent-surface methods; use `client.document_context(doc_id)` instead.
- The `chat()` method now requires keyword arguments after `messages`.
- `chat(protocol="chat_completions")` is the new spelling for Chat Completions.
New Features
- PageIndex Flash Engine for vectorless indexing using layout statistics.
- Unified Client Surface supporting both local and cloud modes identically.
- Enhanced Agent Integration with MCP tool contracts and specific adapters.
- Robust Error Handling with explicit `PageIndexAPIError` exceptions.
- Cited Answers feature via `chat(citations=True)`.
Improvements
- Significant performance boost via Flash engine (no LLM for structure generation).
- Consolidation of local and cloud APIs into a single client surface.
- Enhanced caching and targeting via `client.document_context(doc_id)`.
Upgrade Notes
Update `chat()` calls to use keyword arguments for arguments after `messages`. Replace `doc_id` usage in agent-surface methods with `client.document_context(doc_id)`. Update `chat` protocol calls to use `protocol="chat_completions"`.
The Two Releases
v3.2.0
Sep 8, 2026· 1 month ago
Introduces support for custom Pi model configurations with registry-resolvable model IDs and distinct error codes, alongside improved visibility of startup and preflight failures.
Full release notesv0.2.16
Sep 10, 2026· 1 month ago
The v0.2.16 release introduces the high-performance PageIndex Flash engine for vectorless indexing and consolidates local and cloud APIs into a unified client surface, while enhancing agent integration and error handling.
Full release notes