Overview | RubricHQ Docs

The RubricHQ Observability API lets you push call data from any AI platform into RubricHQ for analysis — voice calls (Vapi, Retell, ElevenLabs, custom telephony) and text/GenAI sessions (LLM chat, function-calling agents).

Base URL

https://api.rubrichq.io/api/public/v1

What you can do

Endpoint	Description
`POST /calls`	Ingest a call (recording URL and/or transcript)
`GET /calls`	List all ingested calls
`GET /calls/{id}`	Get call details and computed metrics

Call types

`observation_type`	Use for	Metrics computed
`voice` (default)	Phone calls, Vapi, Retell, ElevenLabs	Audio metrics (latency, silence, interruptions, WPM) + LLM metrics
`text`	GenAI / LLM chat sessions, function-calling agents	LLM metrics only (no audio required)

How analysis works

When you POST a call:

Transcript metrics run immediately — if you provide a transcript, LLM-based metrics run right away
Audio metrics run async — if observation_type is voice and you provide a recording_url, RubricHQ downloads the audio and runs standard audio metrics in the background
Poll for results — check GET /calls/{id} until analysis_status is "computed"

POST /calls  →  analysis_status: "pending"
                      ↓ (seconds to minutes)
GET /calls/{id}  →  analysis_status: "computed"

Transcript roles

All transcript turns must use one of these roles — any other value is rejected with 422:

Role	Meaning
`AI Assistant`	The AI system being evaluated
`User`	The caller, customer, or simulated tester
`Function Call`	A tool/function invoked by the AI
`Function Call Result`	The response returned from a tool call

IDs you need

Before ingesting calls you need two IDs from your RubricHQ dashboard:

agent_id — from AI Testing Agent → select your agent → URL shows the ID
client_id — from Settings → Workspace → shown as Workspace ID