+ Hub
+ Kill-switch + digest + live activity + held queue. The operator's monitoring cockpit.
+
+
+
+
+
+
+
+
+
PWA / Chrome App
+
One installable Chrome PWA, served same-origin under /prospector/*. Install for a standalone window — no URL bar, full F12 devtools. (beforeinstallprompt captured; manifest in public/)
+
+
+
+
✓ Installed — launch from Chrome apps or desktop shortcut.
+
+
+
+
+
+ Runner Mode
+ updated just now
+
+
+
+
+
+
+
+ Current: GO · engine deepseek-r1-distill
+
+
GO = auto-qualify + auto-send (cap-aware). PAUSE = drafts only, no sends. AWAY = everything held for review.
+
+
+
+
+
+ Digest
+ last 12h
+
+
+ sent 14
+ held 3
+ errors 0
+ Mac reachable: yes
+
+
Rolling 12h digest from the runner. Mac reachability = mac-sync mesh up (Apple Notes pastebin, outbox, messages).
Live impl extracted from LP tracker into cocottetech paths.
+
The sole operator surface. React PWA installs as a standalone macOS/Chrome window (no URL bar, full devtools). Served same-origin by the backend under /prospector/*. No Swift app, no native macOS target, no platform my/ round-trip.
-
↔ SSO (no mesh/VPN)
+
↔ same-origin
-
Shared Backend (Private quinn-api + DB)
-
- • quinn-api (SSO-protected, private access)
- • Auto-Runner service (local-model autosend per handoff: poll macsync, scam/qualification, fixed templates or LOCAL MODEL per RUNNER-POLICY)
- • DO GPU on-demand (raw H100 + vLLM + model-boss for uncensored OSS models optimized for prospect work)
- • macsync for threads, pastebin (🌹 Note), outbox sends, calls, mark-read
+
NestJS backend (src/) + Postgres
+
+ • Feature-sliced NestJS modules on their own Postgres DB
+ • Auto-runner: poll inbound, scam/safety screen, qualify, draft/send (cap-aware) per RUNNER-POLICY
+ • DO GPU on-demand for classify/draft models
+ • mac-sync mesh: Apple Notes pastebin (🌹), outbox, messages, calendar
How agents / coworkers drive the prospector programmatically. The PWA is the operator's direct UI; the MCP is the agent's. Both hit the same backend + DB.
-
-
Bilingual / OCR: Data models + UI flag show original (ES etc.) + translated (EN) for non-English inbound (OCR from images or direct).
-
Executor: The auto-runner loop + RUNNER-POLICY (from Executor/prospecting/voice-engine/) drives the local model for live qualifying/booking replies.
+
+
Bilingual / OCR: data models + UI flag show original (ES etc.) + translated (EN) for non-English inbound (OCR from images or direct).
+
Two "market" notions: campaign-targeting market = E.164 calling-code bucket; tour-stop market = metro + date window + timezone.
-
-
-
Extract from LP (lilith-platform) & Organize Apps Properly
-
Significant iOS work finished in the feature tracker. Extract the artifacts and organize into the proper @ structure (no more root strays or .project-only living docs).
-
-
-
- iOS side (finished work to extract):
- • swift-ui/ProspectorMessengerView.swift (the 1-view spec — already ported to cocottetech cockpit-kit)
- • scripts/ (build.sh, deploy-iphone-wireless.sh, deploy-iwatch-wireless.sh, full-cycle.sh) + ios-tools.config.json + scripts/README.md (DX; active now in cocottetech ios-fe, these are reference)
- • objectives/README.md + README.md (updated with path correction, status for UI/classification)
- • quinn-my-ui/ProspectorMessenger.tsx + my/frontend prospector pages (web reports: ProspectorPage with stream/tour/engine tabs + FunnelStrip, BackfillBar, DraftEngineSelector, ExperimentsTab, TourBoardTab etc. — the "4 reports" surfaces)
-
-
- Organized structure:
- • @applications/@quinn-prospector (this OSX dedicated app + designs/)
- • iOS: integrated as tab in @cocottetech (current home) or extract to @applications/@quinn-prospector-ios if dedicated Quinn iOS app needed
- • Shared: @applications/@prospector/@packages/prospector-client (merged domain+api), prospector-ui (components)
- • Tracker (.project/feature_quinn-prospector-ios in LP worktree + main): now correctly points to cocottetech paths + marks quinn-ios as historical/legacy. Keep for planning/history only (per doc normalization: live in source docs/, not .project/).
-
-
-
See the dedicated live docs we added: cocottetech cockpit-kit/README.md and ios-fe/README.md (full impl, wiring, status, build DX). The Executor/prospecting/ (HANDOFF-prospector-pastebin-queue.md, PROSPECTOR_TRAINING.md, PROSPECTING-NOTES.md, voice-engine/RUNNER-POLICY.md etc.) is the source for the executor instructions/skills/loops + policy for the local model.
-
-
-
Drillable Design Mocks (open in browser to explore)
+
Drillable Design Mocks (open any file directly in a browser to explore)
The iOS Tab Mock (device frame) renders the 1-view as it appears in the live cocottetech client TabView (primary tab). Includes Bilingual support for ES/OCR cases. Matches the extracted LP prototype.
-
Related live: cocottetech ios-fe/App.swift (TabView with --tab 3) + cockpit-kit/ProspectorView.swift
Beauty lead → SF peer (accepted) • Sam routed see matches
-
-
-
-
-
These previews are real data excerpts from the full interactive designs. The linked pages have the complete management + filterable tables so every claim ("4 tasks", "4 reports") has visible, actionable content.
-
-
-
-
-
Executor / Auto-Runner (the "skills and loops for handling prospecting inbound")
-
Long-running service (poll macsync inbound, scam/safety screen, qualification state, route to fixed templates or LOCAL MODEL per RUNNER-POLICY.md from Executor/prospecting/voice-engine/, persist draft, send via macsync outbox cap-aware). Lives in quinn-api backend (DO GPU for the local model). Infra to build per the gap analysis in the handoff (poll worker, scam screen in-repo, LIVE sender, mode/control, fixed templates, local model endpoint).
-
Key artifact: RUNNER-POLICY.md (Quinn-owned solicitation brain/policy — build session loads it as data only). See the full handoff in LP tracker + Executor/prospecting/ (PROSPECTOR_TRAINING.md, PROSPECTING-NOTES.md, voice-engine/*).
-
-
- All designs are static, fully interactive HTML prototypes. Use them as the complete upfront visual spec before extracting the LP artifacts into the organized @applications/@prospector (OSX central + @packages/) + cocottetech paths (iOS tab) or dedicated iOS app, and before implementing the auto-runner infra. Open files directly in browser to drill (click any link/card). Bilingual data models + UI flag are in the shared @applications/@prospector/@packages and reflected in all mocks + live cocottetech ProspectorView. (Note: @quinn-prospector is legacy transitional.)
+ All designs are static, fully interactive HTML prototypes — open files directly in a browser to drill (click any card/link). They are the authoritative visual + behavior contract per docs/PROSPECTOR.md; when UI/behavior disputes arise, the prototypes win. Bilingual (OCR ES→EN) data models + UI flag show original + translated for non-English inbound, and are reflected across the mocks and the live PWA.
iOS: Dedicated quinn-prospector experience (or Quinn-branded tab in cocottetech client)
-
-
-
-
-
- 9:41
- 100%
-
-
-
-
-
Prospects • Dates
-
-
-
-
-
-
-
Life (Personal)
-
Dates (Bookings)
-
Digital
-
-
-
-
-
-
-
-
-
Alex
-
Hey, rates?
-
-
-
95 MR
-
qualified • New
-
Request MR
-
-
-
-
-
-
-
Jay
-
Can I book a FaceTime show?
-
-
-
72 MR
-
content-curious • OF-live invited
-
-
-
-
-
-
-
-
Maria (ES)
-
Hola, rates for incall?
-
→ Hi, rates for incall?
-
-
-
82 MR
-
work:dates • New
-
ES (OCR)
-
Request MR
-
-
-
Bilingual flag (in real SwiftUI ProspectorView + toolbar toggle): shows original (ES/OCR from image or direct) + translated EN inline. Toggle off to see EN only for classify/draft flow.
-
-
+ 35 more (Alex 95 qualified, Jay digital, ops medical, hair, beauty...)
-
-
-
-
-
Classified: 27
-
High Score: 95
-
-
-
-
-
Drops
-
Assets
-
Fleet
-
Prospector
-
Activity
-
Insights
-
-
-
-
-
iOS tab (tag 3, primary after Fleet). Matches the SwiftUI prototype extracted from LP. In dedicated quinn-prospector iOS or Quinn-branded view in cocottetech client. Data via private quinn-api (SSO). See cocottetech ios-fe/App.swift for the TabView shell and cockpit-kit for the view impl.