FLEET
Skedgeloop outreach control plane

FLEET
outreach control plane
Outreach Work Fleet nodes System
send lock: armed
facts first - you choose mode, then approve every body Set mode for 0 shown:
plan board loading…
manual assign → slot:
plan board loading… 👁 vision …
manual assign → slot:
lane control (12h rotation cycles): —
plan board loading…
manual assign → slot:
compose service (Edge) — drafts only, never sends. Models live in Settings → chain.
PC + Kaggle work queue. Claimed rows disappear as workers finish them.
Proof gate: a draft promotes to auto at promote_min_sent sends with zero spam fails. Auto demotes anytime.
Every email per lead: drafts waiting (drafted), queued, sent/failed — and a REPLIED flag the moment a lead writes back. Click a row for the full lead, drafts and reply history.
DM review — audited leads only. Copy the DM, open the profile, paste in Messenger. Nothing auto-sends.

—


      

Review Queue

Dead Domains

, , , ,
plan board loading…
manual assign → slot:

AI providers - model registry for audit findings + form field mapping; keys are write-only (the list is masked, a secret never reaches this page)

LabelKindModelKeyAuditFormsWeightHealthActiveAction

Mailbox accounts — connection info lives here; app passwords NEVER touch the cloud (Pi vault only)

AddressProviderSMTP hostPortUserCap/dHealthStateAction
then put the 16-char app password in secrets/pi.env → pi-push creds

Compose AI chain — tried top→bottom per lead; drag-free reorder via ▲▼; Test pings the endpoint live

API keys — set where the function runs (never readable here by design)

…
supabase secrets set JINA_API_KEY=… GEMINI_API_KEY=… OPENCODE_API_KEY=… OPENROUTER_API_KEY=… CUSTOM_LLM_KEY=… PSI_API_KEY=…
supabase functions deploy lead-compose

Instabox nodes — 1) copy the one-line ssh command from your Instabox dashboard, 2) paste it below, 3) click add node. The keeper keeps each VM awake (ping every ~5 min so it never suspends) and runs the scraper 24h/day — it never rests. Board above shows each runner's live cell + progress.

LabelHostPortUserStatusConcCellsSeenAction
Delete removes the node (its VM keeps running until you stop it in Instabox). Start sets the desired run state; Conc sets how many cells it works at once. The keeper keeps one persistent SSH session per node alive and (re)starts the scraper automatically — no manual setup on the VM.