Release

.env.template

@@ -142,6 +142,10 @@ CONVERSATION_SHARING_ENABLED=true
 # ============================================================================
 MCP_TOOLS_HTTP_PORT=8091
 
+# LLM API connection for tool tracking and dynamic tool descriptions
+LLM_API_BASE_URL=http://llm-api:8080
+MCP_TRACKING_ENABLED=true
+
 # Search engine configuration
 SEARCH_ENGINE=serper                 # Options: serper, searxng
 SERPER_DOMAIN_FILTER=                # Optional: domain filter for search results
@@ -223,6 +227,24 @@ MEDIA_PROXY_DOWNLOAD=true
 MEDIA_RETENTION_DAYS=30
 MEDIA_REMOTE_FETCH_TIMEOUT=15s
 
+# ============================================================================
+# Realtime API (LiveKit Integration)
+# ============================================================================
+# OpenAI Realtime API compatible service using LiveKit as transport
+# Uses global auth settings (ISSUER, AUDIENCE, JWKS_URL) when AUTH_ENABLED=true
+REALTIME_API_ENABLED=true
+REALTIME_API_PORT=8186
+
+# LiveKit Configuration (required)
+LIVEKIT_WS_URL=wss://your-livekit-server.com
+LIVEKIT_API_KEY=your-livekit-api-key
+LIVEKIT_API_SECRET=your-livekit-api-secret
+LIVEKIT_TOKEN_TTL=24h
+
+# Session management
+SESSION_CLEANUP_INTERVAL=15s
+SESSION_STALE_TTL=10m
+
 # ============================================================================
 # Template API (copy when running the scaffold locally)
 # ============================================================================

.github/workflows/dev.yml

@@ -19,6 +19,7 @@ jobs:
       response-api: ${{ steps.filter.outputs.response-api }}
       memory-tools: ${{ steps.filter.outputs.memory-tools }}
       vector-store: ${{ steps.filter.outputs.vector-store }}
+      realtime-api: ${{ steps.filter.outputs.realtime-api }}
       pkg: ${{ steps.filter.outputs.pkg }}
     steps:
       - uses: actions/checkout@v4
@@ -39,6 +40,8 @@ jobs:
               - 'services/memory-tools/**'
             vector-store:
               - 'services/mcp-tools/tools/vector-store-service/**'
+            realtime-api:
+              - 'services/realtime-api/**'
             pkg:
               - 'pkg/**'
 
@@ -129,5 +132,20 @@ jobs:
       registry-url: registry.menlo.ai
       tags: registry.menlo.ai/jan-server/vector-store-service:dev-${{ github.sha }}
       is_push: ${{ github.event_name == 'push' }}
+      build-args: |
+        VERSION_TAG=dev-${{ github.sha }}
+
+  build-realtime-api:
+    needs: changes
+    if: needs.changes.outputs.realtime-api == 'true' || needs.changes.outputs.pkg == 'true'
+    uses: ./.github/workflows/template-docker.yml
+    secrets: inherit
+    with:
+      runs-on: ubuntu-24-04-docker
+      docker-file: services/realtime-api/Dockerfile
+      context: services/realtime-api
+      registry-url: registry.menlo.ai
+      tags: registry.menlo.ai/jan-server/realtime-api:dev-${{ github.sha }}
+      is_push: ${{ github.event_name == 'push' }}
       build-args: |
         VERSION_TAG=dev-${{ github.sha }}

.github/workflows/prod.yml

@@ -13,6 +13,7 @@ jobs:
       response-api: ${{ steps.filter.outputs.response-api }}
       memory-tools: ${{ steps.filter.outputs.memory-tools }}
       vector-store: ${{ steps.filter.outputs.vector-store }}
+      realtime-api: ${{ steps.filter.outputs.realtime-api }}
       pkg: ${{ steps.filter.outputs.pkg }}
     steps:
       - uses: actions/checkout@v4
@@ -34,6 +35,8 @@ jobs:
               - 'services/memory-tools/**'
             vector-store:
               - 'services/mcp-tools/tools/vector-store-service/**'
+            realtime-api:
+              - 'services/realtime-api/**'
             pkg:
               - 'pkg/**'
 
@@ -124,5 +127,20 @@ jobs:
       registry-url: registry.menlo.ai
       tags: registry.menlo.ai/jan-server/vector-store-service:prod-${{ github.sha }}
       is_push: ${{ github.event_name == 'push' }}
+      build-args: |
+        VERSION_TAG=prod-${{ github.sha }}
+
+  build-realtime-api:
+    needs: changes
+    if: needs.changes.outputs.realtime-api == 'true' || needs.changes.outputs.pkg == 'true'
+    uses: ./.github/workflows/template-docker.yml
+    secrets: inherit
+    with:
+      runs-on: ubuntu-24-04-docker
+      docker-file: services/realtime-api/Dockerfile
+      context: services/realtime-api
+      registry-url: registry.menlo.ai
+      tags: registry.menlo.ai/jan-server/realtime-api:prod-${{ github.sha }}
+      is_push: ${{ github.event_name == 'push' }}
       build-args: |
         VERSION_TAG=prod-${{ github.sha }}

.gitignore

@@ -257,6 +257,9 @@ dist/
 .DS_Store
 cmd/server/__debug_bin*
 
+# Service binaries
+services/*/bin/
+
 # Downloaded models (large files - root directory only)
 /models/
 

Makefile

@@ -126,12 +126,14 @@ install-deps:
 # SECTION 3: BUILD TARGETS
 # ============================================================================================================
 
-.PHONY: build build-api build-mcp build-memory build-all clean-build build-llm-api build-media-api build-response-api build-memory-tools
+.PHONY: build build-api build-mcp build-memory build-realtime build-all clean-build build-llm-api build-media-api build-response-api build-realtime-api build-memory-tools
 
 build: build-api build-mcp build-memory
 
 build-api: build-llm-api build-media-api build-response-api
 
+build-realtime: build-realtime-api
+
 build-memory: build-memory-tools
 
 build-llm-api:
@@ -161,6 +163,15 @@ else
 endif
 	@echo " Response API built: services/response-api/bin/response-api"
 
+build-realtime-api:
+	@echo "Building Realtime API..."
+ifeq ($(OS),Windows_NT)
+	@cd services/realtime-api && go build -o bin/realtime-api.exe ./cmd/server
+else
+	@cd services/realtime-api && go build -o bin/realtime-api ./cmd/server
+endif
+	@echo " Realtime API built: services/realtime-api/bin/realtime-api"
+
 build-mcp:
 	@echo "Building MCP Tools..."
 ifeq ($(OS),Windows_NT)
@@ -239,7 +250,7 @@ cli-clean:
 
 # --- Swagger Documentation ---
 
-.PHONY: swagger swagger-llm-api swagger-media-api swagger-mcp-tools swagger-response-api swagger-combine swagger-install
+.PHONY: swagger swagger-llm-api swagger-media-api swagger-mcp-tools swagger-response-api swagger-realtime-api swagger-combine swagger-install
 
 swagger:
 	@echo "Generating Swagger documentation for all services..."
@@ -285,8 +296,17 @@ else
 endif
 	@echo " response-api swagger generated at services/response-api/docs/swagger"
 
+swagger-realtime-api:
+	@echo "Generating Swagger for realtime-api service..."
+ifeq ($(OS),Windows_NT)
+	@powershell -ExecutionPolicy Bypass -File jan-cli.ps1 swagger generate -s realtime-api
+else
+	@bash jan-cli.sh swagger generate -s realtime-api
+endif
+	@echo " realtime-api swagger generated at services/realtime-api/docs/swagger"
+
 swagger-combine:
-	@echo \"Merging LLM API and MCP Tools swagger specs...\"
+	@echo \"Merging LLM API, MCP Tools, and Realtime API swagger specs...\"
 ifeq ($(OS),Windows_NT)
 	@powershell -ExecutionPolicy Bypass -File jan-cli.ps1 swagger combine
 else
@@ -438,6 +458,8 @@ up-full: ## Start full stack (all services in Docker)
 	@echo ""
 	@echo "Services:"
 	@echo "  - LLM API:        http://localhost:8080"
+	@echo "  - Media API:      http://localhost:8285"
+	@echo "  - Realtime API:   http://localhost:8186"
 	@echo "  - MCP Tools:      http://localhost:8091"
 	@echo "  - Vector Store:   http://localhost:3015"
 	@echo "  - vLLM (if enabled): http://localhost:8101"
@@ -775,8 +797,9 @@ ifeq ($(OS),Windows_NT)
 	@powershell -Command "try { $$null = docker compose exec -T api-db pg_isready -U jan_user 2>&1 | Out-Null; if ($$LASTEXITCODE -eq 0) { Write-Host '  PostgreSQL: healthy' } else { Write-Host '  PostgreSQL: unhealthy' } } catch { Write-Host '  PostgreSQL: unhealthy' }"
 	@echo.
 	@echo [API Services]
-	@powershell -Command "try { $$null = Invoke-WebRequest -Uri http://localhost:8080/healthz -UseBasicParsing -TimeoutSec 2 -ErrorAction Stop; Write-Host '  LLM API:    healthy' } catch { Write-Host '  LLM API:    unhealthy' }"
-	@powershell -Command "try { $$null = Invoke-WebRequest -Uri http://localhost:8285/healthz -UseBasicParsing -TimeoutSec 2 -ErrorAction Stop; Write-Host '  Media API:  healthy' } catch { try { if ($$PSItem.Exception.Response.StatusCode.Value__ -eq 401) { Write-Host '  Media API:  healthy' } else { Write-Host '  Media API:  unhealthy' } } catch { Write-Host '  Media API:  unhealthy' } }"
+	@powershell -Command "try { $$null = Invoke-WebRequest -Uri http://localhost:8080/healthz -UseBasicParsing -TimeoutSec 2 -ErrorAction Stop; Write-Host '  LLM API:      healthy' } catch { Write-Host '  LLM API:      unhealthy' }"
+	@powershell -Command "try { $$null = Invoke-WebRequest -Uri http://localhost:8285/healthz -UseBasicParsing -TimeoutSec 2 -ErrorAction Stop; Write-Host '  Media API:    healthy' } catch { try { if ($$PSItem.Exception.Response.StatusCode.Value__ -eq 401) { Write-Host '  Media API:    healthy' } else { Write-Host '  Media API:    unhealthy' } } catch { Write-Host '  Media API:    unhealthy' } }"
+	@powershell -Command "try { $$null = Invoke-WebRequest -Uri http://localhost:8186/healthz -UseBasicParsing -TimeoutSec 2 -ErrorAction Stop; Write-Host '  Realtime API: healthy' } catch { Write-Host '  Realtime API: not running' }"
 	@echo.
 	@echo [MCP Services]
 	@powershell -Command "try { $$null = Invoke-WebRequest -Uri http://localhost:8091/healthz -UseBasicParsing -TimeoutSec 2 -ErrorAction Stop; Write-Host '  MCP Tools:      healthy' } catch { Write-Host '  MCP Tools:      unhealthy' }"
@@ -798,8 +821,9 @@ else
 	@$(COMPOSE) exec -T api-db pg_isready -U jan_user >/dev/null 2>&1 && echo "  PostgreSQL: healthy" || echo "  PostgreSQL: unhealthy"
 	@echo ""
 	@echo "[API Services]"
-	@curl -sf http://localhost:8080/healthz >/dev/null && echo "  LLM API:    healthy" || echo "  LLM API:    unhealthy"
-	@curl -s http://localhost:8285/healthz >/dev/null && echo "  Media API:  healthy" || (curl -s -w "%{http_code}" -o /dev/null http://localhost:8285/healthz | grep -q "401" && echo "  Media API:  healthy" || echo "  Media API:  unhealthy")
+	@curl -sf http://localhost:8080/healthz >/dev/null && echo "  LLM API:      healthy" || echo "  LLM API:      unhealthy"
+	@curl -s http://localhost:8285/healthz >/dev/null && echo "  Media API:    healthy" || (curl -s -w "%{http_code}" -o /dev/null http://localhost:8285/healthz | grep -q "401" && echo "  Media API:    healthy" || echo "  Media API:    unhealthy")
+	@curl -sf http://localhost:8186/healthz >/dev/null && echo "  Realtime API: healthy" || echo "  Realtime API: not running"
 	@echo ""
 	@echo "[MCP Services]"
 	@curl -sf http://localhost:8091/healthz >/dev/null && echo "  MCP Tools:      healthy" || echo "  MCP Tools:      unhealthy"

README.md

@@ -31,6 +31,7 @@ The `quickstart` target wraps `jan-cli` and guides you through:
 - Selecting the LLM provider (local vLLM vs remote OpenAI-compatible endpoint)
 - Choosing the MCP search provider (Serper, SearXNG, or disabled)
 - Enabling or disabling the Media API
+- Enabling or disabling the Realtime API (LiveKit-based real-time communication)
 
 Need to rerun the wizard? Execute `make quickstart` again and accept the prompt to update your `.env`.
 
@@ -108,9 +109,10 @@ jan-server/
 |   |-- response-api/
 |   |-- media-api/
 |   |-- mcp-tools/
+|   |-- realtime-api/
 |   |-- template-api/
 |-- docs/                  # Documentation hub
-|-- docker/                # Compose profiles (infra, api, mcp, inference)
+|-- docker/                # Compose profiles (infra, api, mcp, realtime, inference)
 |-- monitoring/            # Grafana, Prometheus, OTEL configs
 |-- k8s/                   # Helm chart + setup guide
 |-- config/                # Environment templates and helpers
@@ -138,6 +140,7 @@ Key directories:
 | Response API | Multi-step orchestration using MCP tools | 8082 | `services/response-api` | `docs/api/response-api/README.md` |
 | Media API | jan_* IDs, S3 ingest, media resolution | 8285 | `services/media-api` | `docs/api/media-api/README.md` |
 | MCP Tools | Model Context Protocol tools (search, scrape, file search, python) | 8091 | `services/mcp-tools` | `docs/api/mcp-tools/README.md` |
+| Realtime API | LiveKit-based real-time audio/video sessions (optional) | 8186 | `services/realtime-api` | `services/realtime-api/README.md` |
 
 See [docs/architecture/services.md](docs/architecture/services.md) for dependency graphs and integration notes.
 
@@ -180,6 +183,7 @@ make build-llm-api        # Build LLM API
 make build-response-api   # Build Response API
 make build-media-api      # Build Media API
 make build-mcp            # Build MCP Tools
+make build-realtime-api   # Build Realtime API
 
 # Development
 make test-all             # Run all test suites
@@ -202,6 +206,7 @@ make logs-llm-api         # View LLM API logs
 make logs-response-api    # View Response API logs
 make logs-media-api       # View Media API logs
 make logs-mcp             # View MCP Tools logs
+make logs-realtime-api    # View Realtime API logs
 make health-check         # Check all services health
 
 # Database
@@ -415,10 +420,14 @@ More examples: [API Documentation ->](docs/api/)
 ### Docker Compose Profiles
 
 ```bash
-make up-full              # All services
+make up-full              # All services (including optional ones)
 make up-gpu               # With GPU inference
 make up-cpu               # CPU-only inference
 make monitor-up           # Add monitoring stack
+
+# Optional services (enabled via profiles)
+docker compose --profile realtime up -d  # Start Realtime API
+docker compose --profile memory up -d    # Start Memory Tools
 ```
 
 ### Environment Configuration
@@ -511,6 +520,7 @@ See [Monitoring Guide](docs/guides/monitoring.md) for configuration.
 - Response API: Go 1.21+ with Gin, GORM, Wire DI
 - Media API: Go 1.21+ with Gin, GORM, S3 SDK
 - MCP Tools: Go 1.21+ with JSON-RPC 2.0
+- Realtime API: Go 1.21+ with Gin, LiveKit SDK
 
 ## Contributing