Core Operational • Claude 3.5 Sonnet Prompt-Cached

Autonomous L1/L2 Technical Support & Incident Root-Cause Intelligence

RRT SupportOps ingests massive 200k-token multi-server logs, correlates distributed kernel failures, and synthesizes rollback-safe remediation runbooks in under 3.5 seconds.

Request SupportOps Pilot
$ curl -fsSL https://rrt.vn/agent.sh | bash
Latency: 312ms Prompt Cached: 90.4% MTTR: -74.2%
06:58:12.102 [node-sg-01] [CRIT] kernel: TCP: request_sock_TCP: Possible SYN flooding on port 443. Sending cookies. Check SNMP counters.
06:58:12.184 [node-sg-01] [WARN] nginx-ingress: 2026/10/08 06:58:12 [alert] 1290#1290: *49821 768 worker_connections are not enough
06:58:12.240 [node-sg-02] [CRIT] dockerd: container_oom: cgroup /docker/d3b09 memory limit (1024MB) reached. Invoking OOM killer.
06:58:12.310 [node-sg-01] [INFO] rrt-agent: Ingestion complete. 148,200 tokens dispatched to Claude 3.5 Sonnet context.
// Anthropic Claude 3.5 Sonnet Execution Matrix:
• Root Cause Hypothesis: Upstream keepalive exhaustion causing TIME_WAIT socket starvation across Docker NAT boundary.
• Diagnostic Tool 1: query_prometheus_socket_count() → Result: 28,400 TIME_WAIT sockets on node-sg-01.
• Diagnostic Tool 2: inspect_sysctl_net_core() → Result: somaxconn=128 (default Linux bottleneck).
• Confidence Score: 0.984 (Definitive failure pattern verified against kernel telemetry).
• Automated Action: Synthesizing non-destructive sysctl and Nginx configuration diff.
--- /etc/sysctl.d/99-network-tuning.conf (Current) +++ /etc/sysctl.d/99-network-tuning.conf (Remediation Diff) - net.core.somaxconn = 128 + net.core.somaxconn = 65535 - net.ipv4.tcp_max_syn_backlog = 512 + net.ipv4.tcp_max_syn_backlog = 32400 + net.ipv4.tcp_tw_reuse = 1 --- /etc/nginx/nginx.conf - worker_connections 768; + worker_connections 8192;
[STATUS: Rollback-safe syntax validated • Awaiting on-call engineer approval]

Engineered for Deep Infrastructure Reasoning

Moving beyond generic chatbot replies. RRT SupportOps delivers deterministic, multi-layer diagnosis directly into your existing observability stack.

200k Context Multi-Server Telemetry

Traditional monitoring tools fragment when errors cascade across microservices. RRT SupportOps ingests complete distributed syslog buffers, Nginx error dumps, and kernel panics in a single 200,000-token reasoning window without loss of context.

Ingestion Window Size 200,000 Tokens / Incident

Anthropic Prompt Caching

Runbooks, topology specs, and cluster manifests are cached in volatile memory. Recurring diagnostic calls read from cache, slashing reasoning latency by 80% and reducing token costs by 90%.

Cache Read Hit Rate 90.4% Average

Read-Only Diagnostic Tools

Deterministic tool calling interfaces with Prometheus metrics, Docker health APIs, and journalctl in a sandboxed, read-only loop. The engine cannot perform unapproved write operations on production nodes.

Tool Security Policy Strict Read-Only Sandboxed

Human-in-the-Loop Remediation Diffs

Resolutions are synthesized as clear, rollback-safe configuration diffs and shell scripts. On-call engineers inspect the exact proposed systemd or sysctl modification and approve execution with one click.

Mean Time to Resolution (MTTR) -74.2% Reduction
Verified Infrastructure & Observability Stack
Linux Debian 12
Docker Containers
AWS Lightsail SGP
Cloudflare mTLS
Prometheus Telemetry
Claude 3.5 Sonnet
PQ

Pham Quang

Founder & Principal Systems Support Architect — 10+ years engineering Linux cloud clusters (AWS, OVH, Oracle Cloud), mTLS edge architectures, and deterministic AI troubleshooting engines.

Linux Kernel Tuning Cloud VPS Clusters Prompt Caching Specialist Zero-Retention Architect
LinkedIn [email protected]
Hanoi Engineering Lab

Red River Tech (RRT)

Linh Dam Urban Area, Hoang Mai District, Hanoi, Vietnam

System Availability: 24/7 Telemetry Triage

Asia-Pacific Deployment Node

AWS Lightsail Datacenter

Singapore Region (ap-southeast-1a) • IPv4: 18.136.181.242

Perimeter Defense: Cloudflare Anycast mTLS Pulls

Direct Enterprise Channels

Founder / Tech Lead: [email protected]

Pilot Deployment: [email protected]

Direct Escalation: (+84) 035 6010 226