botWebWars/trollagent_gear
Isaac Johnson 862db5e8c3 added and tested trolls 2026-09-13 14:30:16 -05:00
..
INSTALL.md added and tested trolls 2026-09-13 14:30:16 -05:00
README.md added and tested trolls 2026-09-13 14:30:16 -05:00
SETUP.md added and tested trolls 2026-09-13 14:30:16 -05:00
bot.py added and tested trolls 2026-09-13 14:30:16 -05:00
requirements.txt added and tested trolls 2026-09-13 14:30:16 -05:00

README.md

trollagent_gear - Google Cloud Vertex AI (Gemini) Troll Agent

trollagent_gear delegates autonomous tactical reasoning to Google Cloud Vertex AI and Gemini models (gemini-2.5-flash, gemini-1.5-flash, etc.) to drive a solitary Troll character in botWebWars.


👹 Troll Strategy & Gemini Reasoning

  • Hunter Instinct: Evaluates radar minimaps to track and corner player bots and squads.
  • Troll Truce: Instinctively ignores other trolls.
  • Sleep & Regeneration Strategy: Leverages Gemini reasoning to evaluate current HP against target proximity—choosing when to safely rest and sleep (+0.1 HP regeneration) vs. when to strike.
  • Obstacle Navigation: Balances Euclidean and Chebyshev distance against the -0.1 strength diagonal squeeze penalty.
  • No Wizard Challenges: Never interacts with Gary the Wizard.

⚙️ Configuration & CLI Options

Flag Short Default Description
--url -u http://localhost:8000/api botWebWars API base endpoint
--name -n GeminiTroll Troll avatar name
--color -c #047857 Troll hex color
--strength -s 3.0 Initial troll strength
--health -H 10.0 Initial troll health points
--project-id None (auto-detected) Google Cloud Project ID
--location global Vertex AI region (e.g. global, us-central1)
--model gemini-2.5-flash Gemini model name
--api-key None Google AI Studio or Vertex API key
--loop-delay 1.0 Turn polling loop delay in seconds

🚀 Invocation Example

# Using gcloud ADC authentication
gcloud auth application-default login
export VERTEX_PROJECT_ID="my-gcp-project"

python3 trollagent_gear/bot.py \
  --name ApexTroll \
  --color "#059669" \
  --strength 5 \
  --health 12 \
  --url http://localhost:8000/api