Documentation: algorithm / Pipeline Algorithm
Basic CRUD Operations:
Create:
1. The frontend:
- Opens the Prompt Registry pane
- Validates the payload before sending it to the server
- Sends a POST request to the backend
2. The backend:
- receives the request
- Validates the payload
- Stores the record in the database
- Sends a response to the frontend
3. The frontend:
- Receives the response from the backend
- Notifies the user of succes or failure
Read:
1. The frontend:
- Opens the Prompt Registry pane
- Sends a GET request to the backend, requesting data
2. The backend:
- Receives the request
- Validates the payload
- Fetches the record from the database
- Sends a response to the frontend
3. The frontend:
- Receives the response from the backend
- Hydrates the Promp Registry
Update:
1. The frontend:
- Opens the Prompt Registry pane
- Makes local changes to an existing record
- Validates the payload before sending it to the server
- Sends an UPDATE request to the backend
2. The backend:
- Receives the request
- Validates the payload
- Updates the record in the database
- Sends a response to the frontend
3. The frontend:
- Receives the response from the backend
- Notifies the user of success or failure
Delete:
1. The frontend:
- Opens the Prompt Registry pane
- Locates an existing record
- Validates the payload before sending it to the server
- Sends a DELETE request to the backend
2. The backend:
- Receives the request
- Validates the payload
- Deletes the record from the database
- Sends a response to the frontend
3. The frontend:
- Receives the response from the backend
- Notifies the user of success or failure
Compiler Job:
1. Frontend Initialization:
- Opens the compiler configuration pane and establishes settings.
- Opens the terminal and selects the target script.
- Validates the payload data structure.
2. Job Dispatch:
- Frontend transmits the payload to the server via HTTP POST.
- Server receives and validates the incoming payload.
- Server responds with a unique Job ID and begins asynchronous execution.
- Frontend subscribes to the task by initiating a polling loop (HTTP GET) using the Job ID.
3. Native Compilation:
- Server passes the validated payload to the C++ compiler via Pybind11.
- Compiler executes the pipeline and saves output artifacts to disk.
- Compiler returns the execution manifest back across the Pybind11 boundary.
4. State Synchronization:
- Backend receives the manifest and persists it to the database.
- Frontend polling loop detects the job status transition from "building" to "success".
- Frontend issues a final GET request to fetch the manifest and hydrate the interface.
Neural Request:
1. Frontend Initialization:
- Opens the prompt dashboard.
- Selects a designated PromptPair and target Python script.
- Validates the payload data structure.
2. Backend Dispatch:
- Frontend transmits the payload to the server via HTTP POST.
- Backend receives and validates the incoming request.
- Backend persists the initial payload state in the database.
3. LLM Execution & Persistence:
- Backend routes the payload to the designated local LLM server via HTTP POST.
- LLM server processes the prompt and returns the generated inference.
- Backend receives the inference result and updates the database record.
4. State Synchronization:
- Backend returns the final inference response to the frontend
- Frontend renders the comparative results within the dual-pane Inference Window.