Manage AI model assignments, cost rates, and feature flags. Changes take effect within 5 minutes (cache TTL).
AI Model Assignments
Which model each automated workflow uses. Pitch edits and personalization use the model the user selects in the Pitch Engine. Lists and rates come from the shared Semkhor AI catalog.
Journalist search — Grok
Merged media search (Grok/X leg)
Journalist search — Gemini
Merged media search (Gemini + Google Search leg)
Coverage scan — discovery
Broad coverage search (last 12 months)
Coverage scan — verify
Per-article fact-check during scan
Pitch generation
Model for composing pitch templates (Krupp + brand matrix + optional Vertex)
Personalization (fallback)
Only if the Pitch Engine model fails or is omitted
Pitch title suggestion
Auto-title when saving or naming pitches
Media list name
Suggested name after journalist search
Client brand matrix
Extract client facts from selected documents (not writing style)
Client dossier
Legacy dossier generation from documents
Campaign report
AI campaign summary reports
Campaign report (fallback)
Secondary model if primary report generation fails
Nightly discovery
Scheduled journalist discovery jobs
Pitch tracking lookup
Grok search when correlating pitch placements
Legacy Grok-only search
Standalone /api/search-grok endpoint
Cost Rates
USD per 1 million tokens — from Semkhor AI catalog (read-only)
Model
Provider
Input / 1M
Output / 1M
Loading…
Feature Flags
Toggle behaviours that apply across all clients
Hunter Auto-Lookup
Automatically attempt Hunter.io email lookup for contacts returned without an email
Default Search Count
Number of results requested per media search query
Mailgun Webhook
Webhook signature verification for open / click tracking
Checking…
Report Branding
Overrides applied to campaign coverage reports (report.html). Leave blank to use defaults.
Logo URL
Path or URL to logo shown at the top of coverage reports
Primary Brand Color
Hex code for headlines, stat numbers, accents
Tagline Override
Footer tagline shown in italic on coverage reports
Client Brand Matrix
Per-client facts for pitch composition (identity, offerings, proof points). Rebuild from selected local uploads and/or Vertex dossier documents.
Client
Matrix status
—
Last rebuilt
—
Source documents
—
Fact sections
—
Upload files
Or paste text
Select documents to include when rebuilding the brand matrix. Vertex docs need a GCS path (from harvester).
Brand matrix preview (client facts)
Krupp Matrix
Agency-wide Krupp writing style (Semkhor client 2203). Separate from per-client brand matrices.
Client ID
—
Documents
—
Matrix status
—
Matrix updated
—
Cached locally
—
Training documents
Loading…
Document title
Doc ID (optional)
Stable id for idempotency / unfeed
Upload file or paste text
Krupp matrix preview (decree + style)
Client Vertex Config
Hub-and-spoke IDs for dossier RAG. When enabled, Vertex adds query-relevant excerpts at pitch time (Krupp + brand matrix are always used).
Client
Firestore document ID: —
Engine ID
From Vertex Integration tab (e.g. serena-poon-engine_…)
Datastore ID
Client spoke data store ID
Client Archive
Archive clients to hide them from the main dashboard. Archived clients remain in the database and can be restored.
Active clients
Loading…
Archived clients
Loading…
Test Client Seeding
Create 3 realistic test clients (Acme Tech, Riverside Hotels, Heritage Foods) to demonstrate auto-population. Skips any that already exist.
Cache Control
Force all server instances to re-read config from Firestore immediately