download-table fails with NOT_FOUND on dev branch tables (get_file_info missing branch prefix)

[frantisekrehor]

Summary

kbagent storage download-table fails with NOT_FOUND: File not found for any table on a dev branch. Root cause: the async export flow creates a Storage File scoped to the branch, but the follow-up get_file_info call queries the main project scope without a branch prefix.

Reproduction

Tested on kbagent v0.18.6, project databased (891), branch 15894.

# 1. Create branch (auto-activates)
kbagent --json branch create --project databased --name test-export-2026-04-14
# -> branch_id: 15894

# 2. Create bucket + table + upload data
kbagent --json storage create-bucket --project databased --stage in --name test-export-branch
kbagent --json storage create-table --project databased --bucket-id in.c-test-export-branch \
  --name test_data --column "id:INTEGER" --column "name:STRING" --primary-key id
kbagent --json storage upload-table --project databased \
  --table-id in.c-test-export-branch.test_data --file /tmp/test_data.csv
# -> imported_rows: 3, OK

# 3. Verify table exists with data
kbagent --json storage table-detail --project databased \
  --table-id in.c-test-export-branch.test_data
# -> rows_count: 3, columns listed OK

# 4. Download -> FAILS
kbagent --json storage download-table --project databased \
  --table-id in.c-test-export-branch.test_data --branch 15894 --output /tmp/out.csv

Returns:

{
  "status": "error",
  "error": {
    "code": "NOT_FOUND",
    "message": "Resource not found: File not found."
  }
}

Same failure reproduces with storage unload-table on a branch table.

Control: download-table on a production (main) table works fine:

kbagent --json storage download-table --project databased \
  --table-id in.c-data-based-ex-freelo-186733604.users --limit 5 --output /tmp/users.csv
# -> OK, file exported

Root cause

In keboola_agent_cli/client.py:

1. export_table_async (line 1167) correctly prefixes the URL with the branch:

   prefix = f"/v2/storage/branch/{branch_id}" if branch_id else "/v2/storage"
   # POST {prefix}/tables/{safe_id}/export-async

   The Storage File is created in the branch scope.

2. get_file_info (line 1199) does not accept branch_id and always queries main:

   def get_file_info(self, file_id: int) -> dict[str, Any]:
       response = self._request(
           "GET",
           f"/v2/storage/files/{file_id}",
           params={"federationToken": "1"},
       )

   → Storage API returns 404 because the file was created in the branch scope, not main.

3. services/storage_service.py:604 calls client.get_file_info(file_id) without passing branch_id, so even if the client supported it, the service wouldn't use it.

Proposed fix

client.py:1199 — accept and use branch_id:

def get_file_info(self, file_id: int, branch_id: int | None = None) -> dict[str, Any]:
    prefix = f"/v2/storage/branch/{branch_id}" if branch_id else "/v2/storage"
    response = self._request(
        "GET",
        f"{prefix}/files/{file_id}",
        params={"federationToken": "1"},
    )
    return response.json()

services/storage_service.py:604 — pass branch through:

file_detail = client.get_file_info(file_id, branch_id=branch_id)

Other file endpoints in client.py (delete_file, tag_file, untag_file, list_files) likely have the same bug — worth auditing in the same PR.

Impact

Any workflow that tries to read/export data from a dev branch via kbagent CLI fails. Currently the only workaround is to merge the branch to main first, which defeats the purpose of dev branch isolation.

Environment

• kbagent: v0.18.6
• keboola-mcp-server: v1.49.1
• Stack: connection.eu-central-1.keboola.com
• Project: 891 (data-based internal)
• Branch: 15894 (test-export-2026-04-14)

[padak]

Thanks @frantisekrehor for the bug report! The branch-scoped file operations fix is included in v0.19.0.