system-prompts-and-models-o.../dealix/.cursor/rules/dealix-v3.mdc
2026-05-01 14:03:52 +03:00

65 lines
2.0 KiB
Plaintext

---
description: Dealix v3 Saudi Revenue OS architecture and implementation rules
globs:
- "**/*.py"
- "**/*.ts"
- "**/*.tsx"
- "**/*.html"
- "**/*.sql"
- "**/*.md"
alwaysApply: true
---
# Dealix v3 Rules
Dealix is a Saudi B2B Revenue OS, not a generic CRM.
## Core product direction
- Arabic-first Personal Strategic Operator for Sami and later customers.
- Revenue Memory as the event-sourced business memory.
- Supabase/Postgres/pgvector as project + strategic + revenue memory.
- Safe Agent Runtime with approval gates.
- Saudi Market Radar for why-now opportunities.
- Compliance OS for PDPL/contactability.
- Revenue Science for forecasting, impact, churn, and launch-readiness.
- Project Intelligence to understand the codebase and docs.
- Command Center and Personal Operator UI.
## Non-negotiable safety
- Never auto-send cold WhatsApp.
- Never auto-send LinkedIn DMs.
- Never send Gmail externally without explicit approval.
- Never create calendar events externally without explicit approval.
- All external actions must be draft/approval-first.
- Every AI action must be traceable, auditable, and explainable.
## Coding rules
- Keep code deterministic and import-safe.
- Prefer small modules with clear dataclasses and Pydantic schemas where appropriate.
- Add tests for all new core logic.
- Do not add heavy dependencies unless justified.
- Keep external integrations behind interfaces/mocks until credentials exist.
- Use Arabic text for user-facing copy where the feature is intended for Sami or Saudi users.
- Use English for code symbols.
## Testing rules
- Run Python import checks.
- Run pytest where possible.
- Add smoke tests for API routers.
- If tests fail because of existing unrelated issues, document them clearly and isolate new-code tests.
## Launch goal
Reach a private-beta-ready product foundation:
- v3 endpoints work.
- Personal Operator endpoints work.
- Supabase migration exists.
- Project Intelligence works locally.
- Launch readiness report generated.
- Next external integrations are clearly specified.