TrustGraph
Back to News
Release 2.8

TrustGraph 2.8: Async at Scale, Hybrid Retrieval, Structured AI Output, and a Composable New Workbench

September 8, 2026
8 min read

San Francisco, CA — TrustGraph 2.8 is here. This release delivers a foundational platform upgrade for teams deploying knowledge-driven AI systems at enterprise scale: fully asynchronous messaging infrastructure, hybrid Document RAG retrieval, native LLM structured output, image understanding, comprehensive auditability, and a redesigned extensible UI platform.

TrustGraph 2.8 removes a major scaling constraint in multi-workspace deployments while expanding the ways developers can ingest, retrieve, govern, observe, and present contextual AI applications.

Whether you are operating GraphRAG across dozens of workspaces, building multimodal enterprise search, enforcing reliable JSON outputs from LLMs, or tailoring an investigation interface to your domain, TrustGraph 2.8 provides a faster and more composable foundation.

Major Features & Platform Upgrades

Async Pub/Sub Infrastructure for Enterprise Scale

TrustGraph has completed its migration from a thread-per-consumer messaging model to an asynchronous pub/sub architecture.

The previous model could create hundreds of threads as workspace and flow counts increased. TrustGraph 2.8 replaces that approach with configurable async receive and send pools, dramatically reducing resource pressure while supporting larger deployments.

  • Async implementations are now available across Pulsar, RabbitMQ, and Kafka.
  • The API gateway and reverse gateway now run on async infrastructure.
  • The migration removes approximately 7,600 lines of synchronous pub/sub code.
  • Deployments that previously faced thread exhaustion at 40 or more workspaces can now use pooled async consumers and producers instead.

This is a foundational change for organizations operating TrustGraph as a multi-tenant or high-workspace platform.

Hybrid Retrieval for Document RAG

Document RAG now combines semantic vector search with BM25 keyword search through Reciprocal Rank Fusion (RRF).

Vector retrieval is strong at finding conceptually similar material. Keyword retrieval is strong when exact terms, identifiers, error codes, names, and specialist vocabulary matter. Hybrid retrieval combines both signals so that retrieval can better handle real enterprise content.

TrustGraph 2.8 introduces:

  • A pluggable KeywordIndexService.
  • SQLite FTS5 as the initial keyword-index backend.
  • Index scoping per workspace and collection.
  • Text sanitisation for safe and reliable BM25 indexing.
  • Graceful fallback when the keyword-retrieval path is unavailable.

The result is more resilient document retrieval across both semantic questions and exact-term queries.

LLM-Native Structured Output

TrustGraph 2.8 brings JSON-schema enforcement directly into the LLM generation path.

Schemas can now flow from prompt definitions through the text-completion service to supported providers’ native structured-output APIs. A runtime compatibility check determines whether a prompt and backend can use native schema enforcement.

Structured output support now spans:

  • OpenAI, including the vLLM variant.
  • Ollama.
  • Anthropic Claude.
  • Mistral.
  • Llamafile and LM Studio.
  • Azure OpenAI and Azure serverless.
  • Text Generation Inference.
  • Vertex AI Gemini and Google AI Studio.
  • Amazon Bedrock.

This makes it easier to build dependable agents, extraction pipelines, and user interfaces that consume typed LLM responses rather than parsing unpredictable free-form text.

Image-to-Text Service

TrustGraph now includes a pluggable image description service for multimodal applications.

The first implementation includes an OpenAI vision backend, gateway dispatch, REST and WebSocket clients, IAM capability integration, and a new tg-describe-image command-line tool.

Teams can use the service to turn visual content into text that can be stored, indexed, queried, or incorporated into downstream knowledge and AI workflows.

Docling Document Decoder

TrustGraph 2.8 adds an alternative document-processing backend powered by IBM Docling.

The new trustgraph-docling package supports PDF, DOCX, XLSX, PPTX, HTML, Markdown, and CSV files. It supports page-oriented processing by default, plus hybrid chunking modes for teams that need more flexible document segmentation.

Docling can be used as a drop-in alternative to the existing Unstructured-based decoder.

Structured Audit Events

Every gateway request and IAM decision can now produce structured audit events.

The new audit pipeline includes:

  • An AuditEvent data model and AuditPublisher utility.
  • Request-ID and client-IP correlation across gateway and IAM activity.
  • Non-persistent pub/sub delivery.
  • Consumer-managed event storage and retention.

This gives security and platform teams a consistent event stream for governance, operational review, compliance workflows, and downstream SIEM integration.

End-to-End Observability

TrustGraph 2.8 significantly expands metrics across infrastructure, services, and applications.

All metrics now use the tg_ namespace. New coverage includes chunking, metering, LLM services, image-to-text, tools, embeddings, reranking, query services, store writes, agent orchestration, gateway authentication, configuration, knowledge extraction, and IAM.

The release also reduces unnecessary metric-cardinality growth by removing workspace and flow labels from infrastructure histograms.

Gateway Passthrough for Enterprise Extensions

A new generic gateway passthrough path lets organizations integrate enterprise services that are not part of the TrustGraph core platform.

The passthrough interface supports opaque JSON request and response payloads, dynamic service lookup, and dispatcher integration. This enables extensions such as attestation engines and organization-specific services without requiring changes to core TrustGraph code.

Stronger Multilingual RDF Support

TrustGraph 2.8 preserves RDF literal language tags and XSD datatypes throughout ingestion and query workflows.

The enhanced RDF model supports multilingual knowledge graphs and enables queries such as FILTER(LANG()) in SPARQL. It also prevents non-English entity-name collisions by preserving Unicode characters, including CJK characters, during ontology URI normalisation.

Performance, Reliability, and Data Improvements

TrustGraph 2.8 includes major operational improvements across configuration startup, retrieval, structured data, and storage.

  • Faster configuration startup: Per-workspace configuration fetches now run in parallel, addressing startup delays that could exceed 10 minutes in deployments with 50 or more workspaces.
  • Better structured-data imports: Row imports are batched by default, reducing embedding-service calls. Schema indexes are now explicitly divided into query-indexes for exact matching and vector-indexes for semantic search.
  • More reliable reranking: Long labels are truncated only for reranker scoring, avoiding timeouts while preserving full labels in selected graph results.
  • Correct collection deletion: Cassandra deletion paths now paginate correctly for collections larger than the configured fetch size.
  • Improved query filtering: Multi-field GraphQL row queries now apply all filters after index lookup and apply limits after filtering.
  • Self-healing authentication: JWT signing-key retrieval can retry when IAM becomes available after gateway startup.
  • Literal-object indexing: Cassandra now indexes literal RDF objects, allowing graph queries by values such as rdfs:label.

TrustGraph Workbench 2.0.3

TrustGraph 2.8 also includes Workbench 2.0.3, a major evolution of the TrustGraph user interface.

Plugin-Based UI Architecture

The Workbench is now built around a manifest-driven plugin system.

More than 10 demo explorers have moved from hardcoded built-ins to dynamically loaded plugin packages. Teams can support remote plugins, multi-component bundles, and customer-specific UI configurations using a mounted plugins.json file.

The UI architecture is now distributed through a monorepo with public packages including:

  • @trustgraph/client
  • @trustgraph/react-provider
  • @trustgraph/react-state
  • @trustgraph/trustkit

Runtime Themes and Reusable Components

The Workbench now supports five runtime theme presets:

  • Dark
  • Midnight
  • Light
  • High Contrast
  • Warm

A theme context, token-based styling system, scalable typography controls, and persistent user preferences make the interface easier to adapt to different users and deployment environments.

The release also introduces a shared library of themed components, reducing duplicated UI logic across explorers and applications.

Threat Explorer and 3D Graph Visualisation

TrustGraph Workbench 2.0.3 adds a Threat Explorer for OCSF-based insider-threat investigations.

The explorer combines SPARQL relationship analysis with GraphQL event drill-down, supports 12 risk-event subtypes, and includes contextual AI analysis, investigation finding chains, incident reports, and response-plan generation.

A new GraphCanvas3D component also brings perspective graph visualisation to the Workbench. Users can switch between 2D and 3D views, navigate clustered domains, and inspect graph structure using camera controls and depth cues.

Breaking Changes

Please review these changes before upgrading to TrustGraph 2.8:

  • Async pub/sub migration: The synchronous pub/sub layer has been removed. Custom processors must migrate from Publisher and synchronous add() methods to the new async producer and register() model.
  • Default API port: The default API port changes from 8088 to 8888.
  • Vertex AI region: The default Vertex AI region changes from us-central1 to global. Set VERTEXAI_REGION or use --region to override it.
  • Knowledge manager: The librarian now uses per-workspace clients. KnowledgeResponse.ids is now optional.
  • Health-check CLI: tg-verify-system-status no longer accepts --api-url or --pulsar-url.
  • RDF re-ingestion: Existing data containing literal language tags, datatypes, or literal objects that need object-value lookup should be re-ingested.
  • UI extension model: Workbench explorers now use the plugin architecture rather than hardcoded built-in registration.

Get Started

TrustGraph 2.8 is available now on GitHub.

Explore the new async architecture, enable hybrid retrieval for Document RAG, add structured-output schemas to prompts, process documents through Docling, or build a tailored Workbench deployment with plugins and runtime themes.

For more information:

View all news articles