Skip to content
Agent ChartingDevelopers
Documentation Continue a conversation

Continue a conversation

Ask follow-up questions or revise drawings within an existing analysis session.

POST/v1/analysis/runs/RUN_ID/messages

A new run, in the same session

After a completed analysis, send a follow-up to its messages link. The service keeps its owned session, frozen market window and earlier commands. The follow-up receives a new job ID and image; it does not overwrite the earlier evidence. This is not a live market refresh—submit a new market request to analyze a new window.

instruction.promptstring
Your follow-up question or requested chart revision.
instruction.modeenum
prompted (default), professional_auto, or respond_only. respond_only preserves the chart and returns a conversational answer.
curl -X POST "$AGENT_CHARTING_URL/v1/analysis/runs/RUN_ID/messages" \
  -H "Authorization: Bearer $AGENT_CHARTING_API_KEY" \
  -H "Content-Type: application/json" \
  -H "Idempotency-Key: my-chart-request-001" \
  -d '{
  "instruction": {
    "mode": "prompted",
    "prompt": "Explain the invalidation. Keep useful drawings and revise the setup if the evidence warrants it."
  }
}'
202 · accepted (example)
{
  "id": "run_…",
  "object": "analysis.run",
  "status": "queued",
  "links": {
    "self": "/v1/analysis/runs/run_…",
    "events": "/v1/analysis/runs/run_…/events"
  }
}

Response and permissions

Analysis returns 202 immediately by default. A completed run includes analysis.text, setup (when created), commands, inputImage, image, verification and provenance. A failed job has status failed and an error; acceptance is not success.

Required scopes: market:read chart:read chart:write agent:run memory:write setup:write