AI feature in an app
FreeSystem map@rabbinatix/ai-feature-architecture by Rabbinatixupdated Oct 13, 2026
An app adds an AI assistant: the backend builds the prompt from your own data, calls a language model, checks the answer and keeps usage within limits.
5 systems and 9 APIs. Every API is declared on both sides, so every contract starts healthy. Rename or change anything once it is on your map.
7 connections · 9 APIs
- Web App → Backend API
POST /api/v1/auth/registerCreate an accountPOST /api/v1/auth/loginLog in with email and passwordPOST /api/v1/auth/logoutLog outGET /api/v1/meThe signed-in user
- AI Service → Backend API
GET /api/v1/meThe signed-in user
- Web App → AI Service
POST /api/v1/assistant/answersAsk the assistant a question
- AI Service → Vector Database
POST /collections/{name}/queryThe documents closest to the question
- Backend API → Vector Database
PUT /collections/{name}/pointsIndex a document
- AI Service → Language Model API
POST /v1/embeddingsTurn text into a vectorPOST /v1/messagesAsk the model
- Backend API → Language Model API
POST /v1/embeddingsTurn text into a vector
Systems (5)
Web App
Where people ask the assistant.
Backend API
Accounts and the product’s own data.
AI Service
Builds prompts, calls the model, checks answers and counts usage.
Vector Database
Finds the documents closest to a question.
Language Model API
Answers prompts, like the Claude API.
A system you already have under the same name is used, not copied. Outside APIs don't count towards your plan.
Versions
