Vapi Studio

@guidify-ai/vapi-studio — a code-driven toolkit for Vapi.

Website: vapi-studio.guidify.ca — coming soon

This repo is the framework (npm package). Your bot is a separate NestJS app — start from the public starter below.

Create a new project

yarn start alone is not a full setup — fill credentials for your path first. Full checklist: Quick start.

git clone [email protected]:guidify-ai/vapi-studio-project.git my-bot
cd my-bot
cp .env.example .env
# Required for live voice: VAPI_API_KEY, POC_ASSISTANT_ID, VAPI_PHONE_NUMBER_ID
# Optional: TWILIO_* (SMS / Twilio-imported number), OPENAI_API_KEY|ANTHROPIC_… (leave MockBrain otherwise)
# Always: PROJECT_NAME=…  (PROJECT_SLUG optional)
yarn install
yarn start   # Docker + Postgres + ngrok → prints URLs (fails closed without Vapi trio)

Wire Vapi to the printed URLs:

| Setting | URL | | --- | --- | | Webhook | {PUBLIC_BASE_URL}/vapi/webhook | | Custom LLM | {PUBLIC_BASE_URL}/vapi/chat/completions |

• Starter: guidify-ai/vapi-studio-project
• Walkthrough: Creating an app
• Quick start: docs/getting-started/quick-start.md

Requirements: Node 22+, Yarn or npm, Docker Compose, ngrok, a Vapi account for live calls. Twilio and an LLM key only when you leave dry-run SMS / MockBrain.

Showcase (not a blank starter): vapi-studio-landing-page-sample-model — Planner LLM demo.

Repos

| Repo | Role | | --- | --- | | This package | guidify-ai/vapi-studio → @guidify-ai/vapi-studio on npm | | Starter | guidify-ai/vapi-studio-project — fork this to ship a bot | | Showcase | guidify-ai/vapi-studio-landing-page-sample-model — full Planner sample |

Self-hosted and Dockerized. No Guidify-hosted runtime. Vapi is the voice channel; your app is the Custom LLM + webhook endpoint.

Tools

| # | Tool | Status | Summary | | --- | --- | --- | --- | | 1 | Vapi Studio | Shipped | Deterministic agents — typed steps + flow.yaml. Supervisor routes every turn; Brain only at listen boundaries. | | 2 | Conversation events | Shipped | Event-driven bus — onStudioEvent + Nest eventListeners. Build any custom logic on the hooks; Guidify tools use the same surface. | | 3 | Vapi Custom Transcriber | Coming soon | Controllable ASR for digit listens, names, and constrained extracts — fewer junk transcripts as PII. | | 4 | Vapi Voice Profiles | Coming soon | Reusable voice + persona packs for Vapi assistants — consistent brand sound across flows. | | 5 | Vapi Integrations | Coming soon | Signed outbound hooks to CRM and tools — BYOK wiring without leaking provider protocols into Nodes. | | — | And more | Coming soon | More controllable surfaces for Vapi — same self-hosted stack, same graph-owned philosophy. |

Vapi Studio — Introduction · Concepts · Runtime API.

Conversation events — Events & logging · Extending events.

How it fits with Vapi

Vapi Studio assistants are Custom LLM + webhook endpoints — fully compatible with ordinary Vapi assistants:

| # | Mode | When to use | | --- | --- | --- | | 1 | Single-assistant flow | New bots. One Vapi assistant; the graph lives in Studio (flow.yaml + nodes). | | 2 | Multi-assistant Squad | Several Studio-backed assistants in one Squad. See Workflow & Squad. | | 3 | Inject into an existing Squad | Mix Studio and native Vapi members in Squads you already run. |

Most greenfield work starts with (1) via the starter clone above.

How multi-intention routing works

One listen can surface several competing intentions. Main path left → right; portal nodes on a row below.

Supervisor picks the next agent step each turn (scores intentions, checks portals, follows flow.yaml).

| Shape | Meaning | | --- | --- | | NODE | Agent step — speaks, listens, writes memory | | PORTAL NODE | Global interrupt (portal: true in flow.yaml) | | INTENTION | Routing signal on an edge |

Contribute to the framework

Only when changing this library (not when starting a bot):

git clone [email protected]:guidify-ai/vapi-studio.git
cd vapi-studio
yarn install && yarn build && yarn test

See CONTRIBUTING.md — architectural PRs require an ARD under docs/ards/ (SpecKit under .specify/).

Documentation

| | | | --- | --- | | Create a new project | Starter · Creating an app | | Showcase | landing-page-sample-model · Example apps | | Doc site index | docs/README.md | | Runtime handbook | docs/reference/runtime-api.md | | Best practices | docs/best-practices/ | | ARDs | docs/ards/ | | SpecKit | .specify/ · docs/guides/speckit.md | | Env presets | .env.example |

License

MIT — see LICENSE.