creating-replay-vision-scanners

Creates and cost-sizes PostHog Replay Vision scanners that monitor session recordings on a schedule.

713|118|Updated Aug 11, 2020
One-click install
npx skills add https://github.com/PostHog/posthog-foss --skill creating-replay-vision-scanners
Or copy as Structured Prompt for Agent
Please help me install this Agent Skill.
Skill: creating-replay-vision-scanners
Source: https://github.com/PostHog/posthog-foss/tree/main/products/replay_vision/skills/creating-replay-vision-scanners
Command: npx skills add https://github.com/PostHog/posthog-foss --skill creating-replay-vision-scanners

SYSTEM DOCUMENTATION & REQUIREMENTS

What problem does it solve?

Creating a Replay Vision scanner without checking its projected spend can drain an organization's credit budget within the first few scheduled sweeps, since scanners run automatically every 5 minutes against all matching recordings. This Skill guides agents through choosing the right scanner type, shaping the session query, estimating monthly credit spend, and checking remaining budget before creating or widening a scanner.

Core Features & Use Cases

  • Scanner type selection: Configure monitor, classifier, scorer, or summarizer scanners with the correct scanner_config shape, knowing the type is locked after creation.
  • Cost estimation and budget gating: Call vision-scanners-estimate-create and vision-quota-retrieve, prorate the monthly projection against the remaining billing period, and confirm with the user before creating an expensive scanner.
  • Query and sampling tuning: Narrow the RecordingsQuery and adjust sampling_mode, sampling_rate, and model tier to control spend per observation.
  • Use Case: A user asks to flag rage-click sessions automatically. The Skill builds a monitor scanner on the $rageclick event, estimates it at 2,000 credits/month against 800 remaining, surfaces the numbers, and suggests a focused sampling_mode or a per-scanner credit_limit before creating.

Quick Start

Ask the agent to create a Replay Vision scanner that watches sessions matching your criteria and have it estimate the monthly credit cost and check the remaining budget before creating it.

Frequently Asked Questions about creating-replay-vision-scanners

High-intent search queries and answers about installing and using this skill.

FAQPage Schema
How do I create a Replay Vision scanner in PostHog?

Call vision-scanners-create with a unique name, a scanner_type (monitor, classifier, scorer, or summarizer), a scanner_config containing the prompt, and a RecordingsQuery selecting sessions. First estimate spend with vision-scanners-estimate-create and check remaining budget with vision-quota-retrieve.

How do I estimate the credit cost of a session recording scanner?

Call vision-scanners-estimate-create with the proposed query, sampling_rate, sampling_mode, and model. It returns estimated_credits_per_month and other_enabled_scanners_monthly_credits, which you prorate against the days left in the billing period and compare to the remaining quota.

When should I use an inline scan instead of creating a scanner?

Use vision-scanners-inline-scan-create when you already have specific session IDs and a one-off question about them. Scanners are standing watches over future recordings; a throwaway scanner leaves a scheduled sweep running against every future matching recording.

Can I change a scanner's type after creating it?

No, scanner_type is locked after creation. To change it you must delete the scanner and recreate it, so confirm the type and the scanner_config shape up front since wrong shapes are rejected as create errors.

Why did my Replay Vision scanner stop producing observations?

The org's credit budget may be exhausted, in which case scheduled observations are silently skipped until the period resets. Also check error_reason for ineligible outcomes like too_short or no_recording, and remember each session is observed only once per scanner.

Which models can a Replay Vision scanner use?

Only Google Gemini models are supported: gemini-3.5-flash-lite at 2 credits per observation, gemini-3-flash-preview at 5 credits (the default), and gemini-3.7-flash at 15 credits. The model choice sets the price of every observation the scanner makes.