The Stack · Running Now

The infrastructure
behind the advisory.

Most AI consultants show slides. Tacavar shows the metal. Every system on this page runs in production — 24/7, in real environments, with real data. When we architect your stack, we pull from here.

Autonomous Agents
12
Active across 4 workstreams
Research Streams
8
Signals ingested daily
Uptime (30d)
99.94%
Self-healing circuit breakers
Approval Gates
100%
Money-moving actions gated

Hermes — the gateway that routes everything.

All autonomous work flows through Hermes: a multi-platform agent gateway that receives inputs from Telegram, Discord, and web interfaces, routes tasks to the right LLM, enforces approval gates, and coordinates with specialized agent teams.

INPUTS Telegram Mobile / Desktop Web Dashboard Caddy + Cloudflare Scheduled Triggers Cron / Webhooks ORCHESTRATION HERMES Agent Gateway ├─ Auth & access control ├─ LLM routing (hybrid) ├─ Tool use dispatch ├─ Approval gate engine ├─ Audit & observability └─ Session & context state LLM LAYER — OPENCLAW PROXY Claude GPT Ollama (local) AGENT TEAMS Research Team 3 agents · 8-stream ingest Content Team 3 agents · video & brief Trading Team 3 agents · signals + critic Ops Team 3 agents · dev + infra PERSISTENCE & INFRASTRUCTURE PostgreSQL Domain pipeline · Audit log Vector Store Embeddings · semantic search Caddy + Cloudflare TLS · CDN · Subdomain routing DigitalOcean · Ubuntu 24.04 Docker Compose · systemd

Twelve agents. Four workstreams. One shared spine.

Each agent team operates as a specialized unit with its own context, tools, and guardrails. They share the same orchestration spine — which means capabilities compound across teams rather than siloing.

R-01 · RESEARCH
Active

Signal Harvester

Pulls from 8 source streams — HN, Reddit, arXiv, podcasts, GitHub trending, SERP, newsletters, and private feeds — on a daily cadence.

PythonAsync I/ORSS + API
R-02 · RESEARCH
Active

Synthesizer

Takes raw signal and produces thematic clusters, weekly briefs, and ranked opportunity lists. Uses long-context Claude for pattern recognition.

ClaudeEmbeddingsClustering
R-03 · RESEARCH
Active

Critic

Audits every synthesis before publication. Flags thin reasoning, unsupported claims, or recycled takes. Quality gate, not a filter.

GPTRubricVeto Gate
C-01 · CONTENT
Active

Brief Writer

Turns synthesized research into weekly briefs and blog drafts. House style enforced via retrieval-grounded prompts.

ClaudeRAGStyle Guide
C-02 · CONTENT
Active

Video Factory

Converts briefs into shortform video scripts, assembles assets, and stages for review. Human approves before publish.

Multi-modalFFmpegTTS
C-03 · CONTENT
Active

SEO Analyst

Keyword discovery, SERP analysis, and on-page optimization recommendations piped directly into content drafts.

SERP APINLPSchema
T-01 · TRADING
Active

Signal Generator

Multi-strategy signal pipeline with ATR-based dynamic stops, RSI filters, volume confirmation, and FinBERT sentiment overlay.

KrakenFinBERTTA-Lib
T-02 · TRADING
Active

Decision Engine

LLM-based trade decisioner fed by the signal correlator. Outputs position, size, and risk envelope. Paper-only until critic clears.

ClaudeCorrelatorDry-Run
T-03 · TRADING
Active

Trade Critic

Validates every proposed trade against historical patterns, risk limits, and portfolio constraints. Veto authority on live execution.

GPTBacktestRisk Model
O-01 · OPS
Active

Dev Agent

Pair-programming partner with Claude Code under the hood. Handles refactors, migrations, and new-feature scaffolding on approval.

Claude CodeGitCI
O-02 · OPS
Active

Infra Sentinel

Monitors services, alerts on anomalies, auto-restarts healthy-failed workers, and pages when human action is required.

systemdPrometheusTelegram
O-03 · OPS
Active

Approval Broker

Every money-moving or destructive action routes through here. Generates human-readable summaries, waits for chat-based confirmation.

TelegramHIL GateAudit Log

Eight streams. One knowledge base.

The research engine ingests from eight distinct sources on a daily cadence, runs them through embedding and clustering, and produces ranked signals with full provenance. Nothing is black-boxed.

Hacker News

Front page, Show HN, Ask HN. Signal on what operators and builders are paying attention to right now.

freq: 6h · avg: 120 items

Reddit

Curated subreddit set across ML, startups, trading, healthcare. Sentiment and emerging topics.

freq: 4h · avg: 340 items

arXiv

cs.AI, cs.LG, q-fin, and adjacent categories. Abstracts filtered by relevance scoring before full-text.

freq: daily · avg: 90 papers

GitHub Trending

Daily trending across languages, tracking repo momentum, release cadence, and maintainer signals.

freq: daily · avg: 80 repos

Podcast Transcripts

Auto-transcribed feeds from curated industry podcasts. Searchable, timestamped, clustered.

freq: daily · avg: 15 shows

SERP Intelligence

Targeted keyword SERPs, SGE snapshots, and ranking shifts. Early warning on content gaps.

freq: daily · avg: 200 queries

Newsletters

Curated operator and research newsletters ingested via dedicated inbox, parsed to structured signal.

freq: daily · avg: 40 issues

Private Feeds

Client-specific sources wired in per engagement — CRM signals, internal wikis, support ticket streams.

freq: varies · engagement-scoped

Built on boring, reliable metal.

The whole stack runs on infrastructure you can explain to a CTO in a single sentence. No exotic dependencies, no hosted-platform lock-in, no magic. The magic is what we put on top.

Layer 01 · Edge

Cloudflare

DNS · CDN · WAF · DDoS

Layer 02 · Proxy

Caddy

TLS automation · subdomain routing

Layer 03 · Host

DigitalOcean Droplet

Ubuntu 24.04 · dedicated IP

Layer 04 · Services

Docker Compose + systemd

12 agents · orchestrated · self-healing

Layer 05 · Data

PostgreSQL

Pipelines · audit · pgvector

Layer 06 · Intelligence

OpenClaw LLM Proxy

Hybrid routing: Claude / GPT / Ollama

Service Role Status
hermes-gateway Inbound agent routing Healthy
tacavar-bot Telegram interface Healthy
openclaw-proxy LLM routing layer Healthy
research-engine 8-stream ingest Healthy
trading-core Signal + decision pipeline Healthy
domain-empire Domain investing pipeline Healthy
approval-broker HIL gate service Healthy
postgres-14 Primary datastore Healthy
caddy Reverse proxy + TLS Healthy

Autonomous doesn't mean unsupervised.

Every system on this page was built under a core constraint: the human stays in the loop for anything that moves money, touches production data, or is difficult to reverse. Below is how that constraint shows up in the code.

Approval Gates

Every money-moving or destructive action requires explicit human confirmation in Telegram before execution. No implicit consent. No auto-accept after N seconds.

Dry-Run Enforcement

Trading and any irreversible-action subsystems default to paper / dry-run. Live modes require multi-step unlock, critic clearance, and clean backtest window.

Audit Log

Every agent action, LLM call, approval, and override is written to an append-only log with full request/response capture. Fully auditable after the fact.

Dynamic Access Control

User-level scopes for which tools, data, and actions each human can invoke. Access revocation is immediate and enforced at the gateway layer.

Circuit Breakers

Self-healing wrappers around every external call. Cost, rate, and failure budgets enforced per agent. A runaway loop stops itself before you notice.

Shell Injection Prevention

Every shell-reaching tool call goes through a command approval workflow. No raw shell strings. Parameters are allow-listed and escaped at the boundary.

The system, narrating itself.

A representative sample of the audit stream the infrastructure produces. In a client engagement, this same feed is exposed to your team with appropriate scoping.

audit.stream
↻ live

We don't show you slides. We show you the metal.

If you've seen enough, the next step is a private conversation — no pitch, no deck. Just a discussion about what your business's version of this looks like.