Bosch
Enterprise AI & Data Stack Service Solutions • Sandbox
Enterprise AI & Modern Data Infrastructure

Unified AI VeloGrid Platform Service Directory

Central control and direct interface access for 49 production microservices (64 running pods) on Google Cloud GKE Autopilot (rbprj-100522). Unifying Real-Time Conversational Voice & WebRTC Telephony, Autonomous Multi-Agent Swarms, eBPF Network Mesh, Ultra-Low Latency Event Mesh (NATS), Distributed Lakehouse Federation (Hive, Iceberg, Trino, Spark), and Resilient Sagas.

49 / 49
Services Online
64 Pods
100% Healthy
🎙️ Real-Time Voice, WebRTC & SIP Telephony
6 Services
Conversational Voice UI
01 / 49

Pipecat Voice UI Kit

Interactive conversational audio assistant with dynamic waveform visualizers, push-to-talk, VAD, and real-time transcript streams.

Endpoint: voice.velogrid-sandbox...
Voice Agent Framework
02 / 49

Pipecat AI Voice Pipeline

Production multimodal voice pipeline orchestrating Silero VAD, Deepgram STT, LiteLLM reasoning, and neural TTS with sub-300ms latency.

Port: pipecat:8765 (HTTP/WS)
WebRTC Audio SFU
03 / 49

LiveKit WebRTC Server

High-throughput WebRTC SFU server delivering low-latency bidirectional audio/video channels between vehicles and AI voice agents.

RTC Port: 7880 (HTTP) / 7881 (TCP)
SIP Proxy & SBC
04 / 49

Kamailio SIP Server

Carrier-grade Session Border Controller (SBC) handling emergency eCall routing, carrier trunking, and SIP-over-WebSocket signaling.

Signaling: 5060 (SIP) / 8080 (WS)
RTP/WebRTC Transcoder
05 / 49

RTPEngine Media Proxy

In-kernel zero-copy media relay bridging legacy PSTN audio (G.711/AMR) to WebRTC DTLS-SRTP audio (OPUS).

Control: rtpengine:22222 / 8080
SIP Trunk Controller
06 / 49

dSIPRouter GUI

Central management dashboard for provisioning Kamailio PSTN carrier trunks, inbound DID routes, PBX dispatch, and voice failovers.

User: admin / dSIPAdmin2026!
🤖 Autonomous Multi-Agent Frameworks & LLM Gateways
9 Services
Core Backend & MCP
07 / 49

FastAPI Core & MCP Server

Master backend exposing Model Context Protocol (MCP) tool endpoints, telemetry ingestion, DuckDB queries, and distributed sagas.

Endpoints: /docs, /sse, /api/testdrive
Unified LLM Gateway
08 / 49

LiteLLM Proxy & Admin UI

Universal OpenAI-compatible API gateway unifying access across Gemini 3.7, Claude 3.5, and local models with spend tracking and load balancing.

Master Key: sk-litellm-master-key-2026
LLM Rate Limiting
09 / 49

MLflow AI Gateway

Enterprise LLM routing with strict rate-limiting, query caching, token quota governance, and unified secret management.

Port: mlflow-gateway:7000
Deterministic Diagnostics
10 / 49

Agno Diagnostic Agent

Autonomous agent runtime specializing in deterministic vehicle diagnostic tool execution and automotive fault tree analysis.

Service: agent-agno:8002
Collaborative Squad
11 / 49

CrewAI Multi-Agent Squad

Collaborative multi-agent squad (Lead Triage, Field Dispatcher, Fleet Logistics) orchestrating complex customer resolution plans.

Service: agent-crewai:8001
Stateful Cyclic Graphs
12 / 49

LangGraph Workbench

Cyclic state machine engine executing multi-turn workflows with human-in-the-loop approvals, checkpoints, and conditional rollbacks.

Service: agent-langgraph:8003
Composable Chains
13 / 49

LangChain LCEL & Tools

API-first LangChain Expression Language (LCEL) runtime providing streaming composable chains, vector embeddings, and RAG retrieval.

Service: agent-langchain:8004
Agent-to-Agent Mesh
14 / 49

A2A Interop Broker

Agent-to-Agent mesh communication broker allowing disparate frameworks (CrewAI, Agno, LangGraph) to exchange state and tasks.

Protocol: A2A Protocol v2.0
Embedded In-App AI
15 / 49

CopilotKit Runtime

Frontend AI copilot runtime bridging generative UI elements, agent actions, and contextual autocomplete directly into web apps.

Port: copilotkit:3000
Event Mesh, Streaming & Distributed Sagas
4 Services
Sub-Millisecond Mesh
16 / 49

NATS JetStream & UI

Ultra-low latency (<1ms) pub/sub event mesh and persistent stream storage powering real-time agent-to-agent signalling and distributed coordination.

Ports: 4222 (NATS) / 8080 (UI)
Kafka-Compatible Engine
17 / 49

Redpanda Streaming Cluster

High-throughput C++ Kafka-compatible stream broker handling telemetry ingestion, CAN bus streams, and replay without JVM overhead.

Console: redpanda-console:8080
Streaming ETL Engine
18 / 49

Redpanda Connect ETL

Declarative streaming pipeline processor that ingests vehicle sensor streams, filters payloads, and routes data into Iceberg, ClickHouse, and Postgres.

HTTP Port: redpanda-connect:4195
Durable Sagas
19 / 49

Temporal Workflow Engine

Durable execution orchestrator executing multi-step distributed business sagas (dispatching, parts ordering, roadside recovery) with guaranteed compensation.

gRPC Port: temporal:7233
🏛️ Modern Lakehouse, Query Federation & Storage
8 Services
Lakehouse Catalog
20 / 49

Apache Hive Metastore

Centralized relational metadata catalog (backed by PostgreSQL) managing table schemas, partitions, and formats for Apache Spark, Trino, and Iceberg.

Thrift Port: hive-metastore:9083
Federated SQL Query
21 / 49

Trino Distributed SQL

High-concurrency federated SQL query engine allowing querying across Apache Iceberg, Delta Lake, PostgreSQL, and ClickHouse in a single unified SQL query.

User: trino / <none>
Distributed Compute
22 / 49

Apache Spark Master

Distributed compute cluster for large-scale data transformation, batch analytics, PySpark ML pipelines, and Lakehouse compaction jobs.

Master URL: spark://spark-master:7077
ACID Table Format
23 / 49

Apache Iceberg REST

Open ACID table catalog providing snapshot isolation, schema evolution, hidden partitioning, and time-travel querying on S3 object storage.

Port: iceberg-rest:8181
Delta ACID Tables
24 / 49

Delta Lake Engine

ACID transaction layer on Parquet storage with schema enforcement, unified batch/streaming, and Delta Sharing protocol support.

Storage: s3://lakehouse/delta/
Columnar OLAP
25 / 49

ClickHouse OLAP & Tabix

Ultra-fast columnar analytical database designed for real-time aggregation and millisecond analytical queries over billions of telematics rows.

Native Port: clickhouse:9000 / 8123
Vectorized Analytical
26 / 49

DuckDB Engine & MCP

Vectorized analytical engine executing zero-copy SQL queries directly against remote Parquet files on S3 object storage.

Protocol: Anthropic MCP DuckDB Tool
Data Transformation DAG
27 / 49

dbt Core Lineage Docs

Transformation workflow compiler generating documentation, data testing suites, and dependency DAGs for modern lakehouse tables.

Service: dbt-docs:8080
🔍 Vector & Knowledge Search Engines
2 Services
Vector Search & RAG
28 / 49

Qdrant Vector DB & UI

High-performance vector similarity search engine with HNSW indexing and payload filtering for SOP documentation, repair manuals, and RAG search.

HTTP Port: qdrant:6333 / 6334 (gRPC)
Full-Text & Geospatial
29 / 49

OpenSearch & Dashboards

Distributed full-text and geospatial search engine indexing vehicle locations, workshop addresses, and service ticket histories.

User: admin / GridAdmin2026!
🏢 Enterprise Business Applications & CCaaS
3 Services
Customer 360 & SLA
30 / 49

Twenty CRM

Next-generation open-source CRM managing enterprise customer accounts, custom pipeline objects, SLA tiers, and automated account tasks.

Account: admin@velogrid.ai / GridAdmin2026!
Enterprise ERP & Assets
31 / 49

ERPNext Enterprise Suite

Comprehensive ERP suite handling asset tracking, inventory catalogs, technician dispatch work orders, and supply chain operations.

Account: Administrator / GridErpPass2026!
Omnichannel Messaging
32 / 49

Chatwoot CCaaS Platform

Omnichannel contact center platform integrating WhatsApp, SMS, web chat, and voice handoffs between AI agents and human support teams.

Account: admin@velogrid.ai / GridAdmin2026!
🔒 Security, Governance & Observability
14 Services
Identity Provider / SSO
33 / 49

Authentik Identity Provider

Unified Identity & Access Management (IAM) provider supplying single sign-on, multi-factor authentication, and RBAC token flows across stack frontends.

Account: akadmin / GridAdmin2026!
Sidecarless eBPF Mesh
34 / 49

Cilium & Hubble eBPF Mesh

Next-generation eBPF networking layer delivering sidecarless service mesh, transparent mTLS, socket-level TCP bypass, and kernel network flow monitoring.

Port: cilium-hubble-ui:80
ABAC Policy Engine
35 / 49

Open Policy Agent (OPA)

Declarative policy decision engine enforcing Attribute-Based Access Control (ABAC), tenant isolation, and field-level data masking policies.

Port: opa:8181
AI Safety & Rails
36 / 49

NeMo Guardrails

Programmable security middleware filtering AI agent prompts and LLM completions against jailbreaks, prompt injections, and unsafe schemas.

Rails: input_rails, output_rails, injection_filter
Metrics Time-Series
37 / 49

Prometheus TSDB

Time-series metrics monitoring engine scraping Kubernetes pods, NATS throughput, Trino query latencies, and cluster resource saturation.

Port: prometheus:9090
Log Aggregation
38 / 49

Grafana Loki

Scalable, multi-tenant log aggregation system capturing standard output from all 61 cluster containers with full LogQL querying in Grafana.

Port: loki:3100
Distributed Tracing
39 / 49

Grafana Tempo & OTel

Distributed tracing backend collecting OpenTelemetry spans across voice pipelines, agent tool calls, and backend microservice RPCs.

gRPC: otel-collector:4317 / tempo:3200
Enterprise Dashboards
40 / 49

Grafana Enterprise

Central dashboarding interface visualizing infrastructure metrics, cluster CPU/RAM footprints, and voice stream quality.

User: admin / GridAdmin2026!
LLM Tracing & Metrics
41 / 49

Langfuse Studio

Specialized LLM observability suite tracking prompt versions, token consumption, model latencies, and agent execution trees.

Account: admin@velogrid.ai / GridAdmin2026!
Metadata Governance
42 / 49

OpenMetadata Catalog

Universal metadata governance catalog mapping end-to-end data lineage from S3 Parquet tables through dbt DAGs to Superset BI charts.

Account: admin / admin
Self-Service BI
43 / 49

Apache Superset BI

Modern self-service enterprise business intelligence platform with interactive dashboards connected to Trino and ClickHouse.

User: admin / GridAdmin2026!
SQL Visualizations
44 / 49

Redash Visualizations

Collaborative SQL querying and dashboarding tool tailored for rapid ad-hoc analytical queries and metric sharing.

Account: admin@velogrid.ai / GridAdmin2026!
S3 Object Lakehouse
45 / 49

MinIO Lakehouse S3

High-performance Kubernetes-native S3 object storage serving as the persistent physical lakehouse data lake for Iceberg, Delta, and MLflow.

Account: minioadmin / MinioAdminPass2026!
In-Memory Cache & Lock
46 / 49

Redis In-Memory Store

Ultra-low latency in-memory data store providing session caching, LiveKit WebRTC state coordination, and agent rate-limiting.

Endpoint: redis:6379 (DB: 0)
Agent Harness & MCP
47 / 49

TrueForge Agent Harness

Self-hostable agent execution harness managing reasoning loops, MCP client bindings, session context compaction, and Daytona sandboxed micro-VM execution.

Port: 8080 (HTTP / MCP)
LLM KV Cache Engine
48 / 49

LMCache KV Engine

High-performance tiered KV cache engine (RAM → Redis → MinIO S3) delivering ~78% TTFT latency acceleration, non-prefix reuse, and PD disaggregation.

Port: 8000 / 65432 (HTTP / Proto)
Meta-Harness & Polly
49 / 49

Omnigent Meta-Harness

Open-source meta-agent orchestrator ("Polly") coordinating multi-agent delegation across coding swarms with spend caps, sandboxing, and collaborative co-driving.

Port: 8080 (HTTP / WebSocket)