agentspan-examples

Example agent workflows built with agentspan and the Conductor OSS Python SDK.

Setup

cp .env.example .env
# fill in ANTHROPIC_API_KEY and AGENTSPAN_SERVER_URL

You need a running Conductor server. For local dev, build and start conductor-server from conductor-oss/conductor main:

./gradlew :conductor-server:bootJar
java -jar server/build/libs/conductor-server-*-boot.jar --server.port=7001

Install the Python SDK:

pip install conductor-python

Examples

games/number_guesser.py

A single agent finds a secret number 1–100 using binary search. The tool maintains state across calls (via ToolContext) so the secret persists between guesses.

python games/number_guesser.py

games/adversarial_number_game.py

Two agents compete in a number-guessing game:

• Keeper picks a secret number 1–100 and can lie once (flip higher↔lower) to slow the guesser down
• Guesser uses binary search but must detect the lie by spotting contradictions in the response history

python games/adversarial_number_game.py

Override the model via env var (default: anthropic/claude-haiku-4-5-20251001):

AGENTSPAN_LLM_MODEL=ollama/qwen3:8b python games/adversarial_number_game.py

barkeeps-ledger workflows

Three agents that query a live web analytics API and post intelligence to Discord. See barkeeps/README.md for setup and scheduling.

barkeeps/hourly_digest.py

Fetches the last hour of traffic (top pages, countries, trend) and posts a 3-5 sentence summary to Discord. Run hourly via cron.

barkeeps/anomaly_detector.py

Compares current-hour traffic against the 6-hour rolling average. Posts a Discord alert on spikes (≥ 2×) or drops (≤ 0.3×); silent when traffic is normal. Run every 15 minutes.

barkeeps/weekly_report.py

Synthesizes 7 days of hourly data, top pages, and visitor geography into a 150-250 word narrative posted to Discord. Run weekly.