Note: This changelog tracks all releases and changes across Jan Server's microservices architecture, organized chronologically by version.
Status: ✅ Complete | Date: December 2025
Major feature release with realtime communication via LiveKit, enhanced MCP capabilities, improved automation testing, multiple vLLM instance support, and refined user experience features.
- Multi-vLLM Instance Provider:
- Round-robin load balancing across multiple vLLM instances
- Enhanced swagger documentation
- Improved provider instance management
- MCP Tools Enhancements:
- Admin API integration for MCP management and monitoring
- Model prompt administration
- URL filtering and validation
- Performance improvements for search operations (reduced latency)
- Improved error handling for invalid format handling
- Tool call tracking and verification
- User Settings & Personalization:
- User-specific settings storage and management
- Default settings for existing users
- Settings validation and improvement
- Browser user support
- Conversation & Content Management:
- Image uploading and processing
- Image content parsing improvements
- Conversation title generation improvements
- Model context size validation
- Message sharing features
- Deep research capability integration
- Project-level conversation management
- Conversation deletion (including bulk operations)
- Model tool support flag
- Better conversation naming and titling strategies
- File Search & Content Management:
- File search disabled by default for better performance
- Content filtering and optimization
- File upload and handling improvements
- Instruct Model Setup:
- Dedicated configuration for instruct-based models
- Enhanced model initialization
- Monitoring & Metrics:
- Enhanced monitoring metrics tracking
- Improved observability for system performance
- Latest prompt orchestration metrics
- Automation & Testing:
- Refactored automation test suite
- Improved test execution and logging
- Cleaner test output and reporting
- Bug Fixes & Improvements:
- Context size reload handling improvements
- Token size calculation fixes
- Conversation recreation with same name fixes
- Invalid format parsing fixes
- Browser compatibility improvements
- Settings configuration improvements
- Improved CI/CD pipelines for all services
- Enhanced swagger API documentation
- Better code organization and standardization
- Cleaner logging and debugging output
- Improved test infrastructure and reporting
Status: ✅ Complete | Date: December 2025
Major expansion of admin capabilities with comprehensive management APIs, enhanced model catalog features, user personalization system, and robust middleware stack.
- Admin API Endpoints: Full operations for user management, group management, and feature flag administration
- Model Catalog Enhancements:
- Feature flag integration for controlled model access
- Experimental model support with explicit flagging
- Advanced capability filtering (images, embeddings, reasoning, audio, video)
- Model ordering and categorization improvements
- User Personalization & Project Instructions:
- Custom project instructions per workspace/conversation
- User-specific preferences and settings storage
- Context-aware instruction injection for enhanced AI responses
- Project-level configuration management
- Database Schema Updates:
- Feature flags table with user/group assignments
- Audit logging for admin operations
- Model display names and ordering fields
- Enhanced provider model metadata
- Project instructions and user preferences tables
- Middleware Additions:
- Admin authorization middleware for protected endpoints
- Token bucket rate limiting for API protection
- Feature flag checking middleware for gated features
- Request validation and sanitization
- Platform web admin UI for model catalog management
- Comprehensive API documentation for admin endpoints
- Improved error handling and validation messages
Status: ✅ Complete | Date: November 2025
Complete architectural overhaul from Kubernetes-native monolithic platform to microservices-first architecture with Docker Compose, Kong Gateway, and comprehensive observability.
- 4 Core Microservices: LLM API (8080), Media API (8285), Response API (8082), MCP Tools (8091)
- Kong 3.5 Gateway: Centralized routing, auth, and rate limiting
- PostgreSQL 18: Upgraded from v16, simplified single-instance architecture
- 100+ Makefile Commands: Setup, testing, monitoring, deployment (400% increase)
- 6 Test Suites: Auth, conversations, responses, media, MCP, E2E with jan-cli
- Observability Stack: OpenTelemetry, Prometheus (9090), Grafana (3331), Jaeger (16686)
- Service Template: Automated microservice generator with
new-service-from-template.ps1
- One-command setup:
make quickstart - Profile-based deployment:
make up-full,make up-gpu,make monitor-up - Hybrid development mode: Native execution with hot reload
- 20+ documentation pages covering guides, API reference, architecture
- Docker Compose profiles (infra, api, mcp, full, monitoring)
- Kubernetes/Helm charts updated for all services
- S3-compatible encrypted storage
- Keycloak OIDC authentication + guest login
- Environment configs: defaults.env, development.env, testing.env, secrets.env.example
- MCP Protocol integration (mark3labs/mcp-go)
- Google Search (Serper API), web scraping (SearXNG)
- Code execution sandbox (SandboxFusion)
- Vector store integration for semantic search
- Gateway port changed:
8080→8000 - Swagger UI moved:
/api/swagger/index.html→/v1/swagger/ - Health endpoint:
/healthcheck→/healthz - Project structure:
apps/→services/ - Removed PostgreSQL replicas (now single instance)
- Removed Python inference service (replaced with vLLM)
Status: ✅ Complete | Date: Oct 2025
Foundation for microservices transition with Makefile consolidation, MCP provider integration, and initial observability stack.
- Makefile Consolidation: Merged modular Makefiles into single structured file (10 sections)
- Hybrid Development Mode: Native binary execution with hot reload for faster iteration
- MCP Provider Integration: Basic Model Context Protocol support with JSON-RPC endpoint
- Observability Foundation: Prometheus, Jaeger, Grafana, OpenTelemetry collector
- Guest Authentication: Keycloak token exchange for quick access without full registration
- PostgreSQL 16: Upgraded from earlier versions
- Kong Gateway Migration: Replaced custom gateway with Kong 3.5
- Service Restructuring: Moved from
apps/toservices/directory structure
- OpenAI-compatible endpoint:
POST /v1/chat/completions - Initial MCP tools endpoint:
POST /v1/mcp - Docker Compose organization with profiles
- Basic test collections for core functionality
Status: ✅ Complete | Date: Initial Release
First release of Jan Server with LLM API, authentication, conversation management, and Docker deployment.
- LLM API Service: OpenAI-compatible chat completions, conversation management, message history
- Authentication System: User registration, API key generation, token-based auth
- Conversation Management: Create, manage, and persist conversations with message history
- Docker Deployment: Basic Docker Compose configuration for local development
- PostgreSQL Backend: Database schema for conversations, messages, users, and authentication
| Metric | v0.0.10 | v0.0.11 | v0.0.12 |
|---|---|---|---|
| Services | 1 | 2 | 4+ |
| Deployment Methods | Docker | Docker | Docker + K8s |
| Make Commands | ~10 | ~20 | 100+ |
| Test Suites | None | Basic | 6 collections |
| Documentation Pages | ~3 | ~5 | 20+ |
| Monitoring Tools | None | 2 | 4 (full stack) |
| Auth Methods | Basic | Keycloak | Keycloak + Guest |
| Gateway | None | Kong | Kong 3.5 |