Documentation
Service architecture, tiers, modes and how things work under the hood.
Section 01
Service overview
ChiMate.ai is an AI assistant that analyses screen content. The user takes a screenshot via the desktop client, the server runs OCR and forwards the result to the selected AI model for analysis. The answer returns to the client.
Section 02
Workflow
1) Capture: the user picks a window or area and triggers a capture via the overlay button or a hotkey. 2) OCR: the image goes to the server, where a specialised model extracts text and its structure. 3) Analysis: the text and/or image are passed to the main model, which produces a structured answer.
Section 03
Model tiers
Three tiers are available — Simple (1 credit, ~70% accuracy), Advanced (2 credits, ~85%) and Top (3 credits, ~95%). The choice depends on the importance and complexity of the task.
Section 04
Work modes
Supported modes are: Auto, Tests (multiple-choice questions), Translation (between the source and the user's native language), Code (analysing or writing code). The mode is detected automatically from screen content, or can be set manually by the user.
Section 05
Security & privacy
Screenshots are stored on the server temporarily for debugging and quality control — the retention period is configurable in the admin panel. The desktop overlay runs at the OS level and never appears in screenshots or screen sharing performed by other applications.
Section 06
Limits & quotas
Request quotas come from your plan. Each request consumes the credit cost of its tier. Credits refresh at the start of each subscription period. Additional safeguards: per-IP / per-user rate limiting on the API side.