Summary

When using the azure-responses provider, OpenClaw sends a reference to a prior Responses item id (rs_...) even though the model config declares compat.supportsStore=false. The Azure/OpenAI Responses API returns:

HTTP 400: Item with id 'rs_...' not found. Items are not persisted when store is set to false.

This prevents normal chat completion.

Environment

• OpenClaw: 2026.2.22-2 (45febec)
• OS: Windows 10 (x64)
• Provider: azure-responses (api=openai-responses)
• Models: gpt-5.2, gpt-5.2-codex

Config snippet (models)

From openclaw config get models --json:

{
  "providers": {
    "azure-responses": {
      "baseUrl": "https://<redacted>.cognitiveservices.azure.com/openai/v1",
      "api": "openai-responses",
      "models": [
        {
          "id": "gpt-5.2",
          "compat": { "supportsStore": false }
        },
        {
          "id": "gpt-5.2-codex",
          "compat": { "supportsStore": false }
        }
      ]
    }
  }
}

Logs

Gateway log shows the failure:

error=HTTP 400: Item with id 'rs_0741061cf7a0bf0001699d22b5bd3c819487fdc97e35e8568b' not found. Items are not persisted when `store` is set to false. Try again with `store` set to true, or remove this item from your input.

Expected

If supportsStore=false, the adapter should not send any previous_response_id / continuation reference that requires server-side persistence.

Actual

Request payload appears to include a reference to a prior rs_... id, causing 400.

Notes

openclaw config get models --json reports compat.supportsStore=false for these azure-responses models.