
Most voice agent testing at the post-build, pre-deploy stage is ad hoc, done in Excel, not formalized, and difficult to maintain or scale. ComplyLine treats the agent like a black box: you only need its phone number. Write down, in plain language, what it should always or never do. We use AssemblyAI's LLM Gateway to turn that into caller personas, situations, and pass/fail criteria. You keep the scenarios you like and drop the rest. Then we actually call the agent. Twilio dials the number, and an AssemblyAI Voice Agent plays the caller on the live line. When the call ends, another LLM Gateway pass reads the transcript, scores every criterion, and hands back a pass or fail with quotes from the conversation. You can listen live, and you can seek the recording to the moment something went wrong. ComplyLine also covers post-call analysis. Drop in an uploaded recording or a pasted transcript. You get a ranked report on TCPA consent, whether the agent said it was AI, whether it honored an opt-out, and whether PII leaked. Optional packs cover HIPAA, GLBA, FERPA, PCI, GDPR, and other state laws. Each finding cites the statute and points to a timestamp. The idea is simple. If you are buying a voice agent, run the same tests a builder would. Use your tests to set a bar that agents must clear before deployment. This can be done independent of the agent provider. If you already shipped one, catch a missing disclosure or a policy slip before a customer (or their lawyer) does. The demo path uses AssemblyAI's Voice Agent API and its LLM Gateway. You do not need extra API keys to try it.
30 Sep 2026
.png&w=828&q=75)
ThetaGate is an autonomous, paper-only options trading desk built with Alpaca and AI agents. Before the market opens, and again midday, it analyzes Alpaca’s most-active stocks, market movers, prices, options eligibility, and recent news to create a focused watchlist of 10–20 symbols. For each candidate, ThetaGate calculates volatility and breakout signals in Python, filters real option contracts by liquidity, expiration, and delta, and gives the AI proposer only verified contracts to choose from. The model cannot invent strikes, control position sizing, or access broker order tools. Every proposal must pass deterministic account, portfolio, liquidity, and maximum-loss checks before an atomic credit-spread order can reach Alpaca paper trading. ThetaGate then reconciles fills, prevents duplicate submissions, monitors profit targets and stop-losses, responds to market-regime changes, and safely recovers after restarts. A Streamlit dashboard provides watchlists, news, positions, decisions, risk status, and trading activity.
4 Sep 2026