Skip to content
Portfolio

Tech Stack

LayerTechnology
LanguagePython 3.x
AI frameworksLangChain, LlamaIndex
Vector storeChromaDB or FAISS
Cloud APIsGroq (Llama 3) or DeepSeek
Networking / securityParamiko (SSH), Regex (DLP scrubber)
Notificationsntfy (self-hosted pub-sub)
RuntimeLXC container or VM on Proxmox

All agent logic runs in Python 3.x:

  • LlamaIndex — document ingestion, chunking, embedding, and retrieval
  • LangChain — agent orchestration, tool definitions, memory, and LLM API integration
  • Paramiko — programmatic SSH sessions for remediation
  • Regex (re) — DLP pattern matching for IPs, credentials, MACs, and tokens

StorePurpose
ChromaDB / FAISSLocal vector database for log and runbook embeddings
PrometheusMetrics and alert triggers (via query_prometheus tool)
Local filesystemSource logs (syslog, Docker logs) and Markdown runbooks

ServiceRoleData exposure
Groq / DeepSeek APILLM reasoningSanitized text only — no raw IPs or credentials
ntfyMobile push notificationsRoot-cause summary and action webhook

API keys for cloud LLM providers are stored locally and never included in prompts sent back to the model.


The agent runs as a dedicated workload on the homelab:

  • LXC container or VM on the Proxmox cluster (Hypervisor docs)
  • Network access to monitored nodes (SSH), ntfy server, and outbound HTTPS for cloud LLM API
  • No inbound exposure to the public internet

See Enterprise Homelab — Compute for the broader container runtime context.