Integrations

Every external dependency is swappable and degrades gracefully. With no credentials Salim runs entirely on a deterministic simulation layer; add a key and that capability goes live, no code change.

Demo mode

Integrations

Live

Simulated

Mode

Demo

Offline-first

Core platform

AI Reasoning

Vercel AI SDK · no provider

Simulated

Deterministic rules engine drives the agent

Set to go liveOPENAI_API_KEY

Database

Neon Postgres · Drizzle

Simulated

In-memory seeded store

Set to go liveDATABASE_URL

Acquisition

Phase 1

NLP & Translation

LLM · AR / RU / ZH / EN

Simulated

Rule-based intent + canned translations

Set to go liveOPENAI_API_KEY

WhatsApp Business Cloud

Simulated

Simulated channel transport

Set to go liveWHATSAPP_API_TOKEN

SMS

Twilio

Simulated

Simulated channel transport

Set to go liveTWILIO_ACCOUNT_SID

Voice

ElevenLabs TTS

Simulated

Voice playback disabled in demo

Set to go liveELEVENLABS_API_KEY

Title AI · Back-office

Phase 2

Document VLM

Vision LLM

Simulated

Deterministic field synthesis

Set to go liveOPENAI_API_KEY

OCR

Google Document AI

Simulated

Seeded passport / MRZ text

Set to go liveGOOGLE_DOCAI_PROCESSOR_ID

AML & Sanctions

Screening provider

Simulated

Simulated watchlist matcher

Set to go liveAML_PROVIDER_API_KEY

Title Registry

DLD / Oqood

Simulated

Simulated registry responses

Set to go liveDLD_API_KEY

Object Storage

Cloudflare R2

Simulated

Local demo references

Set to go liveR2_ACCESS_KEY_ID

Durable agent runtime

The conversational brain is also packaged as an isolated eve.dev durable agent under /agent, wrapping the same shared tool-core (portfolio search, Golden Visa, document request) for always-on WhatsApp & voice channels.