A pipeline is a graph of blocks (layout detection, OCR models, conditions, transcription…) that the platform runs on a document or an audio file. This section describes the gateway routes.
Base URL and authentication
Base URL: https://api-endpoints.alphaedge-ai.com
Authentication: X-API-Key: <key> or Authorization: Bearer <key> header. Only the GET /pipelines/capabilities route is public.
Two ways to run a pipeline
| Saved pipeline | Raw graph | |
|---|---|---|
| When | Pipeline created or chosen in the Studio (e.g. "IDP — Quality") | You build the graph yourself |
| Route | POST /api/pipelines/{org}/{id}/run/ | POST /api/pipelines/custom/{kind}/run/ |
| Graph | Stored on the AlphaEdge side | Sent in the pipeline field |
Response format
Your choice: JSON (default) or an SSE stream (stream=true with Accept: text/event-stream). Pipeline responses do not display any price.
Limits
| Limit | Value |
|---|---|
| Nodes per graph | 40 |
| Links per graph | 80 |
| Graph size | 64 KiB |
| Pages per document | 100 |
| Execution time | 3600 s |
| Regions processed in parallel | 8 |
The effective limits are always available in GET /pipelines/capabilities (limits field).
Routes
| Route | Role |
|---|---|
POST /api/pipelines/{org}/{id}/run/ | Run a saved pipeline |
GET /pipelines/capabilities | Blocks and tests that can be used right now (public) |
POST /pipelines/{kind}/validate | Validate a graph without a document or consumption |
POST /api/pipelines/custom/{kind}/run/ | Run a custom graph (Enterprise plan) |
POST /documents/classify, POST /documents/split | Classify or split a batch of pages |
POST /pipelines/words/writing_type | Printed or handwritten, word by word |
Errors are described on the "Error codes" page (invalid_pipeline and pipeline_error families).