TurbolineTurboline Logo

Changelog

All notable changes to TurboStream will be documented in this file. The format is based on Keep a Changelog, and this project adheres to Semantic Versioning.

Version 0.3.1

Latest

Released on February 8, 2026

Added

Modern API Documentation

  • Replaced Swagger UI with Scalar for modern, interactive API documentation
  • New documentation endpoint at http://localhost:7210/docs
  • Enhanced request/response visualization with better schema examples
  • Improved browser compatibility with ad-blocker support
  • Separate Content Security Policy for documentation and API endpoints

Enhanced Topic Extraction

  • Root-level array indexing: [3] extracts element at index 3 from root arrays
  • Nested array access: data.path[-1] accesses last element in nested structures
  • Combined path expressions: field[0].subfield for complex data navigation
  • Comprehensive logging for extraction debugging and troubleshooting

API Improvements

  • Complete OpenAPI specification for API key management endpoints
  • Documented create, list, and revoke API key operations
  • Consolidated topic prompt configuration in feed model
  • Enhanced WebSocket authentication logging

Changed

Documentation Updates

  • Renamed swagger.json to openapi.json for clarity
  • Updated all documentation references and endpoints
  • Added browser compatibility notes for ad-blockers
  • Improved API reference structure and examples

Security Headers

  • Context-aware Content Security Policy (CSP) enforcement
  • Permissive CSP for documentation pages to enable API testing
  • Strict CSP for API endpoints to prevent XSS and injection attacks
  • Enhanced frame protection and content type validation

Fixed

  • Fixed Content Security Policy blocking Scalar documentation functionality
  • Resolved MongoDB index conflict errors during API key service initialization
  • Corrected security header middleware for documentation pages
  • Fixed topic extraction for root-level array data structures

Removed

  • Removed redundant topic-prompt management endpoints (replaced by topicPrompts map)
  • Consolidated deprecated route handlers

Version 0.3.0

Released on January 30, 2026

Added

Multi-Topic Architecture

  • Topic-based message routing using configurable field identifiers
  • Topic Router service for message distribution to topic-specific channels
  • Per-topic context store with independent rolling windows (50 entries default)
  • Per-topic analysis memory with conversation continuity (3 Q&A pairs default)
  • Compound key architecture (feedID:topic) for complete data isolation
  • Support for unlimited topics per feed through single WebSocket connection
  • Topic subscription filtering for selective intelligence delivery

Stateful Intelligence System

  • Continuous per-topic LLM analysis with configurable intervals (default: 10s)
  • Topic LLM Scheduler service for automated query loop management
  • State memory service with rolling window and state token compression
  • Analysis memory for conversation continuity across queries
  • TSLN (Time-Series Lean Notation) format integration for 40-60% token reduction
  • Thread-safe concurrent access to context and memory stores

Intelligence Broadcasting

  • LLM intelligence delivery to consumers (not raw feed data)
  • Topic-specific WebSocket room broadcasting
  • Real-time intelligence streaming with llm-intelligence message type
  • Callback-based architecture for flexible delivery mechanisms

API-First Foundation

  • REST API framework for feed management
  • WebSocket API for real-time intelligence streaming
  • API key authentication infrastructure
  • Multi-tenant namespace isolation
  • Usage tracking and metering foundation
  • Rate limiting framework
  • Comprehensive API documentation at https://turboline.ai/docs/api

Documentation

  • Complete architecture documentation suite
  • Per-topic implementation guides
  • Client integration guides
  • Performance optimization strategies
  • Testing and validation procedures

Changed

Repository Structure

  • Terminal UI extracted into separate standalone repository: https://github.com/turboline-ai/turbostream-tui
  • Backend reorganized as pure Intelligence-as-a-Service platform
  • Backend code consolidated in root turbostream folder
  • Clear client-server separation for independent development

Feed Model

  • Extended with topicField for topic identification
  • Added topics array for multi-topic configuration
  • Added enableTopicRouting flag for feature control
  • Backward compatible with single-topic feeds

WebSocket Manager

  • Refactored to support topic-based routing and broadcasting
  • Integrated Topic LLM Scheduler lifecycle management
  • Enhanced message handling for intelligence delivery
  • Improved connection and reconnection logic

LLM Service

  • Adapted context and memory stores for per-topic architecture
  • Implemented compound key pattern for topic isolation
  • Enhanced thread safety for concurrent topic access
  • Optimized token usage through TSLN format

Fixed

  • Thread safety issues in concurrent feed context access
  • Memory leaks in topic router channel management
  • State memory corruption in high-concurrency scenarios
  • WebSocket reconnection race conditions

Deprecated

  • Monolithic TUI-Backend coupling (superseded by client-server architecture)
  • Single-topic-only architecture (superseded by multi-topic support)
  • Direct TUI backend access (moving to API-mediated access)

Removed

Security

  • Per-topic data isolation through compound keys
  • Thread-safe concurrent access to shared resources
  • API key authentication framework
  • Customer namespace isolation foundation

Performance

  • 95% infrastructure cost reduction through connection multiplexing
  • 40-60% LLM token reduction through TSLN optimization
  • Support for 100+ concurrent topics per feed
  • 18,000 LLM queries/hour throughput (50 topics)
  • 29 KB memory footprint per topic

Version 0.2.x - Previous Versions

Prior to v0.3.0, TurboStream operated as a monolithic application with Terminal UI and backend tightly coupled. Version history for 0.2.x releases focused on single-topic feed processing and direct TUI integration.

Migration Notes

Upgrading from 0.3.0 to 0.3.1

Backend Update:

  1. Pull latest changes: git pull origin main
  2. Update dependencies: go mod tidy
  3. Build and test: go build ./... && go test ./... -v
  4. Restart server to access new documentation at http://localhost:7210/docs

Breaking Changes:

None. All changes are backwards compatible. Existing feeds and API integrations continue to work without modification.

Documentation Access:

Swagger UI endpoint (/swagger/index.html) redirects to new Scalar endpoint (/docs). If using an ad-blocker, you may need to disable it for localhost:7210 to test API calls.

Upgrading from 0.2.x to 0.3.0

Backend Services:

  1. Backup existing feed configurations
  2. Update dependencies: go mod tidy
  3. Build and test: go build ./... && go test ./... -v
  4. Deploy new binary
  5. Verify existing feeds still connect

Terminal UI:

  1. Clone standalone TUI repository: https://github.com/turboline-ai/turbostream-tui
  2. Configure backend API endpoint
  3. Build and run TUI client
  4. Verify monitoring capabilities

Single-Topic Feeds:

No changes required. Feeds continue to work with existing configuration.

Multi-Topic Migration:

Add topicField, topics, and enableTopicRouting fields to feed configuration.

Resources