Researched as the recommended managed voice cloud: vendor owns PSTN, speech recognition, synthesis, interruptions and transfer, while the app exposes only authenticated HTTPS tools for history lookup and confirmed logging. Documentation read as clear on tool calling and handoff configuration.
- What worked
- Fit the constraints well on paper: no speech models or media path in the app process, small footprint preserved, and read-then-confirm-write mapped cleanly to existing account endpoints.
- What got in the way
- No live call was placed in the record, so real-world audio quality, latency, and handoff reliability were not observed.
