Pages API

Pages represent individual pages within an artifact (document). Each page carries extracted text, compound mentions, NER tags, and a summary. Most page data is populated automatically by the extraction pipeline, but endpoints exist for manual updates and re-triggering workflows.

Get Page

GET /pages/{page_id}

Returns the full page read model including text content, compound mentions, tag mentions, summary, and metadata. Requires view permission on the parent artifact.

Response fields include:

Create Page

POST /pages/
Content-Type: application/json

{
  "artifact_id": "artifact-uuid",
  "index": 0
}

Returns 201 Created. In normal operation, pages are created automatically during artifact processing. This endpoint exists for manual page management.

Update Text

PATCH /pages/{page_id}/text_mention
Content-Type: application/json

{
  "text": "The extracted text content of this page...",
  "model_name": "pymupdf",
  "date_extracted": "2026-03-29T00:00:00Z"
}

This fires Page.TextMentionUpdated, which triggers the embedding and NER pipelines.

Update Tags

PATCH /pages/{page_id}/tag_mentions
Content-Type: application/json

[
  {"tag": "NadD", "entity_type": "target"},
  {"tag": "SACC-111", "entity_type": "compound"},
  {"tag": "tuberculosis", "entity_type": "disease"}
]

Update Summary

PATCH /pages/{page_id}/summary_candidate
Content-Type: application/json

{
  "summary": "This page presents IC50 measurements for...",
  "model_name": "manual",
  "is_locked": true
}

Add Compound Mentions

POST /pages/{page_id}/compound_mentions
Content-Type: application/json

{
  "page_id": "page-uuid",
  "compound_mentions": [
    {
      "smiles": "CC(=O)Oc1ccccc1C(=O)O",
      "extracted_id": "SACC-111",
      "confidence": 0.95
    }
  ]
}

The page_id in the body must match the path parameter. This fires Page.CompoundMentionsUpdated, triggering SMILES embedding.

Trigger Workflows

All trigger endpoints return 202 Accepted with a WorkflowStartedResponse containing the workflow ID. Re-triggering is safe due to ALLOW_DUPLICATE reuse policies.

Embedding Generation

POST /pages/{page_id}/embeddings/generate

Starts the text embedding Temporal workflow. Requires the page to have text content.

Compound Extraction

POST /pages/{page_id}/compounds/extract

Starts the OCSR ML pipeline (YOLO + DECIMER) to extract chemical structures from the page image.

SMILES Embedding

POST /pages/{page_id}/compounds/embed

Starts the ChemBERTa embedding workflow for the page's compounds. Requires extracted compounds with valid canonical SMILES.

Summarization

POST /pages/{page_id}/summarize

Starts the LLM page summarization workflow. Locked summaries (human corrections) are preserved.

NER Extraction

POST /pages/{page_id}/ner/extract

Starts the named-entity recognition workflow. Extracts compounds, targets, diseases, genes, and other entities from the page text.

Get Summary

GET /pages/{page_id}/summary

Returns the page summary with metadata:

{
  "entity_id": "page-uuid",
  "summary": "This page describes...",
  "model_name": "gemma3:27b",
  "date_extracted": "2026-03-29T12:00:00",
  "is_locked": false,
  "hil_correction": null
}

Returns 404 if the page has no summary yet.

Get Workflow Statuses

GET /pages/{page_id}/workflows

Proxies to Temporal to return the current status of all page-level workflows: embedding, compound extraction, SMILES embedding, summarization, and NER extraction.

{
  "entity_id": "page-uuid",
  "workflows": {
    "text-embedding-{page_id}": {"status": "COMPLETED", "...": "..."},
    "compound-extraction-{page_id}": {"status": "RUNNING", "...": "..."}
  }
}

Delete Page

DELETE /pages/{page_id}

Returns 204 No Content. Requires edit permission on the parent artifact.