How ivancontreras-m/asistente works
Avril Lab’s prototype accepts marketing briefings through a CLI or REST API, researches the web, generates campaign copy, checks claims, and saves materials and an execution report for human review; it does not publish automatically. The README documents the four-stage workflow and optional search providers. The sampled implementation is one file, so the stages are conceptual responsibilities within agente.py, not separately verified modules; unsampled API and chat wiring is not expanded.
Main entry points: agente.py Briefing runner · output_marketing Markdown and JSON files
Components 13
User interfaces
- Briefing CLI
agente.py - REST API
agente.py - Interactive chat
agente.py
Campaign workflow
- Briefing runner
agente.py - Web research
agente.py - Copy generation
agente.py - Claim verification
agente.py - Material preparation
agente.py
External services
- NVIDIA LLM
- Brave / DuckDuckGo
Review materials
- Markdown and JSON files
output_marketing
Other components
- Marketing user
- Human reviewer
Connections 15
- Marketing user → Briefing CLI: submits briefing
README.md - Marketing user → REST API: posts briefing
README.md - Marketing user → Interactive chat: requests content
README.md - Briefing CLI → Briefing runner: runs briefing
README.md - REST API → Briefing runner: submits briefing
agente.py - Interactive chat → Briefing runner: requests generation
agente.py - Briefing runner → Web research: starts research
README.md - Web research → Brave / DuckDuckGo: searches web
agente.py - Briefing runner → Copy generation: generates copy
README.md - Copy generation → NVIDIA LLM: requests completion
README.md - Briefing runner → Claim verification: checks claims
README.md - Claim verification → NVIDIA LLM: requests verification
README.md - Briefing runner → Material preparation: prepares materials
README.md - Material preparation → Markdown and JSON files: writes files
README.md - Markdown and JSON files → Human reviewer: awaits approval
README.md