I built a company of bots

Description and use

Botropolis shows how 20 specialized agents across 10 departments, coordinated by a CEO agent, can process requests efficiently using YAML specs, analytics, and a LoRA‑based training pipeline.

Problem

Organizing multiple specialized agents into a coherent, company‑like structure to handle diverse tasks efficiently.

Outcome

The system demonstrates that proper routing, agent specification, and analytics enable a scalable, company‑like agent ecosystem, though the demo‑scale research model is limited in factual reliability.

Workflow

• 1. User submits a request to the FastAPI server. 2. CEO agent parses the request and determines the relevant departments. 3. CEO routes the request to the appropriate specialist agents. 4. Each specialist agent processes its part and returns results. 5. CEO aggregates results into a single report. 6. Report is returned to the user via the browser UI.
• 7. Analytics track per‑agent calls, latency, and token usage for future optimization.

Evidence

Botropolis: 20 specialist agents across 10 departments. Research has Scout. Code has Coder and Reviewer. Finance, legal, health, data, marketing, ops, security, support all have their people.

Source

The CEO agent reads your request, routes it to the right departments, runs them, and compiles one report.

Source

Each agent is a YAML spec: name, title, specialty, model, tools, system prompt, example tasks.

Source

I trained botropolis-scout-tiny, a 5.5M-parameter transformer, from scratch on CPU on 300 synthetic Q&A pairs.

Source

The part I am proudest of is the training pipeline. I trained botropolis-scout-tiny, a 5.5M-parameter transformer, from scratch on CPU on 300 synthetic Q&A pairs.

Source

Usage analytics change how you build. Once you see per-agent latency and tokens, you start routing like a manager watching a budget.

Source