Skip to content

Documentation / Examples / Vertex Product Catalog

Vertex Product Catalog consumer

This portable example is based on the proven standalone prod-v4.2 migration:

  • consumer-owned prompt and synthetic catalog;
  • a closed-schema lookup_product_catalog tool;
  • Deepgram-style STT, Vertex-style reasoning, and Google speech route IDs;
  • a real public LiveKitTransportAdapter handoff;
  • separate scripted route adapters;
  • terminal transcript, tool, lifecycle, and resource evidence.

Run it:

python examples/vertex_product_catalog/main.py

Files

File Owner
prompts.py consumer persona and product rules
product_catalog.py consumer lookup behavior
product_catalog.json synthetic reviewed data
main.py consumer/VAANI composition

The route IDs are realistic, but provider behavior is scripted. No credential, SIP route, customer record, or proprietary catalog fact is included. The example does not claim live Vertex, Deepgram, Google speech, or telephony success.

What to replace in a real consumer

  • Load prompts and catalog/domain services from the consumer application.
  • Create provider adapters with supported public factories and explicit credential sources.
  • Translate a real, already-routed LiveKit job into LiveKitCallHandoff.
  • Keep model, voice, language, fallback, transfer, and frontend choices in reviewed consumer configuration.
  • Consume CallResult once after Closure.

The VAANI runtime code remains unchanged across all of those replacements.

Next: port an existing bot or configure providers.