Linking Tavus as a source
Let AI connect your sources for you
Skip the manual setup — run this in your project and the wizard auto-detects your databases and APIs and connects them to PostHog.
Contents
The Tavus connector syncs your AI video generation and conversational AI data into PostHog, including videos, faces, PALs, conversations, transcriptions, and canvas interactions.
Adding a data source
- Go to the Data pipeline page and select the Sources tab.
- Click + New source and select Tavus by clicking the Link button.
- Enter your API key. See Getting your API key below.
- Click Next.
- Select the tables you want to import, set the sync method and frequency, then click Import.
The data warehouse then starts syncing your Tavus data. You can see details and progress in the data pipeline sources tab.
Getting your API key
To connect Tavus to PostHog, you need a Tavus API key:
- Log in to your Tavus account at platform.tavus.io.
- Navigate to the API Keys page.
- Create a new API key or copy an existing one.
- Paste the key into PostHog when linking the source.
The API key provides read access to all your Tavus data, including videos, faces, PALs, conversations, transcriptions, and canvas interactions.
Configuration
| Option | Type | Required |
|---|---|---|
API key | password | Yes |
Supported tables
| Table | Description | Sync method | Incremental field | Primary key |
|---|---|---|---|---|
videos | A Tavus generated video — a rendered talking-head video produced from a script and a replica. | Full refresh | — | — |
replicas | A Tavus replica — a trained digital likeness used to generate personalized videos. | Full refresh | — | — |
personas | A Tavus persona — the conversational configuration (system prompt, layers, replica) for a CVI agent. | Full refresh | — | — |
conversations | A Tavus conversation — a real-time Conversational Video Interface (CVI) session with a replica. | Full refresh | — | — |
memory_stores | A Tavus memory store — the persistent memory one PAL keeps for one participant across conversations. | Full refresh | — | — |
faces | A Tavus face — a human likeness trained with Phoenix, used in conversations and videos. The current name for a replica. | Full refresh | — | — |
pals | A Tavus PAL — the agent configuration (system prompt, layers, default face) that a conversation runs with. The current name for a persona. | Full refresh | — | — |
transcriptions | A Tavus transcription — a speech-to-text job for a video file. | Full refresh | — | — |
canvas_interactions | A Magic Canvas interaction recorded in a Tavus conversation, such as a submit, skip, or dismiss on a card. | Full refresh | — | — |
Sync modes
All Tavus tables use full refresh sync. Tavus's API doesn't support incremental filtering, so each sync fetches all records.
| Table | Sync method |
|---|---|
videos | Full refresh |
replicas | Full refresh |
personas | Full refresh |
conversations | Full refresh |
faces | Full refresh |
pals | Full refresh |
transcriptions | Full refresh |
canvas_interactions | Full refresh |
Available tables
| Table | Description |
|---|---|
videos | Generated talking-head videos produced from a script and a face |
replicas | Trained digital likenesses used to generate personalized videos (legacy name for faces) |
personas | Conversational configurations for CVI agents (legacy name for PALs) |
conversations | Real-time Conversational Video Interface (CVI) sessions |
faces | Trained digital likenesses used in conversations and videos (current name for replicas) |
pals | Agent configurations (system prompt, layers, default face) for conversations (current name for personas) |
transcriptions | Speech-to-text jobs for video files |
canvas_interactions | Magic Canvas interactions recorded in conversations (submit, skip, dismiss actions on cards) |