Changelog

v0.3.0 — 2026-03-20

Dashboard view — new top-level "Dashboard" nav item in the React dashboard with:
- Stacked bar chart showing request counts over time, broken out by success (HTTP 1xx-3xx) and failure (HTTP 4xx-5xx)
- Selectable request type filter: All Requests, Embeddings, or Inference
- Selectable timeframe: Last Hour (per-minute), Last 24 Hours (15-minute), Last 7 Days (hourly), Last 30 Days (4-hour)
- Optional endpoint URL substring filter
- Summary cards showing total successful, failed, and total request counts
- Quick actions section with shortcuts to Manage Embedding Endpoints, Manage Inference Endpoints, View Request History, and Endpoint Explorer
Request statistics API — POST /v1.0/requests/statistics endpoint returning aggregated request counts grouped by time bucket with success/failure breakdown
- Supported across all database providers: SQLite, PostgreSQL, MySQL, SQL Server
Dashboard is now the default landing page after login

Gemini and vLLM endpoint support as first-class API formats
PolyPrompt NuGet integration for upstream provider calls
Summarization pipeline step — optional LLM-powered summarization of semantic cells before chunking and embedding
Hierarchical semantic cells — SemanticCellRequest now supports parent-child relationships via GUID, ParentGUID, and Children
Completion endpoints — new CRUD resource type for managing LLM completion/inference API endpoints (Ollama, OpenAI, Gemini, vLLM), with full health check support
SummarizationConfiguration — inline configuration supporting TopDown and BottomUp strategies, customizable prompts, parallel processing, and retry logic
Summary cell type — new AtomTypeEnum.Summary for cells generated by summarization
Dashboard endpoints restructure — "Endpoints" navigation split into "Embeddings" and "Inference" sub-sections, each with CRUD, health status, and health histograms
Dashboard summarization UI — updated processing view with summarization configuration
SDK support — all three SDKs (C#, Python, JavaScript) updated with completion endpoint methods and summarization models
Default inference endpoint — new tenants are automatically provisioned with a default Ollama inference endpoint
Chunking strategies: RegexBased, Row, RowWithHeaders, RowGroupWithHeaders, KeyValuePairs, WholeTable
Table-type chunking strategies for structured data

Route restructure: Embedding endpoint routes moved from /v1.0/endpoints to /v1.0/endpoints/embedding
Route restructure: Processing routes moved from /v1.0/endpoints/{id}/process to /v1.0/process (embedding endpoint ID now in request body via EmbeddingConfiguration.EmbeddingEndpointId)
EmbeddingConfiguration schema changed: added EmbeddingEndpointId (required)
SemanticCellRequest schema changed: added GUID, ParentGUID, Children, SummarizationConfiguration
SemanticCellResponse schema changed: added GUID, ParentGUID, Type, Children
ChunkResult schema changed: added CellGUID
AtomTypeEnum extended with Summary value
Dashboard "Endpoints" navigation restructured into "Embeddings" and "Inference" sub-sections

Initial release of Partio
Multi-tenant REST API with bearer token authentication
Semantic cell processing with chunking and embedding
Chunking strategies: FixedTokenCount, SentenceBased, ParagraphBased, WholeList, ListEntry
Overlap strategies: SlidingWindow, SentenceBoundaryAware, SemanticBoundaryAware
Embedding clients: Ollama, OpenAI, Gemini, vLLM-compatible
Database support: SQLite, PostgreSQL, MySQL, SQL Server
Admin CRUD endpoints for tenants, users, credentials, and embedding endpoints
Request history with filesystem body persistence and automatic cleanup
React dashboard (Vite) with full admin UI
SDKs: C#, Python, JavaScript
Docker support with multi-arch builds (amd64, arm64)
Automated test suite