LPCOS Intelligence Studio
Starter-compatible orchestration. The AI operating brain runs client-side: it probes what's actually reachable, maintains the honest data chain, and routes around locked capabilities.
Capability chain
SOURCE
1758
Pets Currently Tracked by LPCOS (CONNECTED_SOURCE_COUNT, confidence: high)
ACQUIRED
200
LpcosPetReservoir (page sample — floor, not total)
VERIFIED
1837
records past acquisition in lifecycle
AVAILABLE
1839
active + adoptable
INDEXED
1834
discoverable (name+image)
DISPLAYED
24
page-size cap of 24
MATCHED
0
favorites + inquiries
ACQUIRED is never reported as SOURCE. The broad provider universe — every animal the source could list. NOT a census; LPCOS may have acquired only a subset.
Capability matrix
Objective routing
When a capability is LOCKED, the system finds a Starter-compatible alternative. If none exists, it preserves last-known-good and records a recovery path.
Understand system health
uses native.truthEngine
Run the client-side truth engine on demand. It computes the full chain and quality gaps from live Pet records — the same data the backend function would have read.
Discover & acquire pets from RescueGroups
Use InvokeLLM with add_context_from_internet (Gemini) to pull publicly available RescueGroups listings on demand, then write new Pet records client-side. This is on-demand, not scheduled — it runs when an admin opens the dashboard.
Recover missing adoption URLs
uses native.urlRecovery.deterministic
Run deterministic URL recovery client-side: for each available Pet missing an adoption_url, derive it from the source organization URL pattern + source_animal_id. Write back via base44.entities.Pet.update.
Detect & resolve duplicates
uses native.duplicateDetection
Client-side duplicate detection over the available Pet inventory: match on (source_animal_id || name+breed+province). Flag, don't auto-delete. Surface pairs for admin review.
Classify data freshness
uses native.freshnessClassifier
Recompute data_freshness from source_updated_at/last_seen_at client-side and write back to Pet records.
Score the mission & self-certify
uses native.truthEngine
Compute the mission score client-side from the truth chain + quality gaps. Write the snapshot to LpcosMissionScore entity (write works on Starter).
Guard the brand (logo/character/spelling)
uses integration.GenerateImage
Use InvokeLLM with file_urls (vision) to inspect assets client-side and flag brand/spelling/character issues. No scheduled enforcement — runs on admin demand.
Generate SEO metadata
uses native.seoGenerator
Generate SEO metadata deterministically from the route registry client-side. No AI credits needed for structural metadata.
Run autonomously on a schedule
Genuinely impossible on Starter — nothing runs when no one is on the site. Workaround: every admin dashboard visit triggers a diagnostic pass, so intelligence is on-demand rather than continuous. This is the one unavoidable limit.
Receive a webhook (Stripe/FTP)
Genuinely impossible on Starter — no public function endpoint. Workaround: use Stripe Checkout (client-side redirect) for payments; for FTP, the external bridge pattern still works if re-pointed to a hosted endpoint you control elsewhere. Not bypassable on-platform.
Acquisition (Starter-compatible)
Two honest paths to grow the pet population without the locked backend sync. Both use real source data only — no hallucinated records.
Promote acquired reservoir → Pet
Records we already acquired from RescueGroups but never canonicalized into the displayable Pet entity (the locked orchestrator's job). Real data, already in our DB.
Acquire on demand (web search)
Pull real, currently-listed RescueGroups pets via InvokeLLM web search. Each candidate is URL-verified (anti-hallucination gate) and created inactive — admin approves before it's public.