stim-app

Web UI for exploring .stim trace files from Solace Agent Mesh (SAM).

Stack

• Backend: FastAPI + PyYAML + Anthropic SDK (Claude chat over stim data)
• Frontend: Vite + React + TypeScript + Tailwind + shadcn/ui

Quick start

# Backend
cd backend
python -m venv .venv && source .venv/bin/activate
pip install -e .

# Option A: native Anthropic
export ANTHROPIC_API_KEY=sk-ant-...
# Option B: OpenAI-compatible proxy (e.g. litellm)
# export LLM_SERVICE_ENDPOINT="https://lite-llm.example.com/"
# export LLM_SERVICE_API_KEY="sk-..."
# export LLM_SERVICE_PLANNING_MODEL_NAME="openai/vertex-claude-4-5-sonnet"

uvicorn stim_app.main:app --reload --port 8787

# Frontend (separate terminal)
cd frontend
npm install
npm run dev

Open http://localhost:5173.

Features

• Upload .stim YAML traces — parsed server-side.
• Summary: wall time, LLM vs tool time, per-agent token totals, slowest spans.
• Timeline: Gantt-style per-agent swimlanes, color-coded by event type.
• Events: sortable table + JSON payload inspector.
• Chat: ask Claude questions about the loaded trace. Claude has tools to query events, payloads, timings, token usage.