Agent Connector 4.3.3: Add create_work_items_from_dataframe action

[raivolink]

Summary

• Adds create_work_items_from_dataframe() action that creates one Work Item per row of a named dataframe for a specified agent
• Adds _fetch_dataframe() helper that handles multiple server versions (thread-scoped path for newer local servers, flat path for cloud/older servers) and both response formats (list[dict] and {"columns": [...], "rows": [...]})
• Updates README with usage example and CHANGELOG with 4.3.3 entry
• Bumps package version from 4.3.1 to 4.3.3

Test plan

• Verify create_work_items_from_dataframe creates one Work Item per dataframe row with correct payload keys/values
• Test against a newer local server (thread-scoped endpoint path)
• Test against a cloud server (flat endpoint path, thread from invocation context header)
• Confirm graceful error when dataframe name not found (404 on both paths)
• Confirm ActionError is raised (not swallowed) on non-404 server errors

🤖 Generated with Claude Code

[playerzero-ai]

PlayerZero

feat-df-to-workitem → main

Change Summary

Summary

• Added a new create_work_items_from_dataframe() action to Agent Connector that:
  • resolves the target agent by name,
  • loads a named dataframe from the current thread,
  • creates one Work Item per dataframe row by mapping column names to payload keys.
• Introduced _fetch_dataframe() as a compatibility shim for agent-server dataframe reads:
  • tries threads/{thread_id}/data-frames/{name} first, then falls back to data-frames/{name},
  • accepts both response shapes: {"columns": [...], "rows": [...]} and list[dict],
  • surfaces non-404 failures as ActionError instead of swallowing them.
• Updated release metadata and docs:
  • README now documents dataframe-driven work item creation with an example,
  • CHANGELOG adds the 4.3.3 entry,
  • package version bumped from 4.3.1 to 4.3.3.

Broader impact

• This extends the package’s existing work-item orchestration from manual JSON/batch inputs to a data-driven batch path.
• The change deliberately bypasses the higher-level get_data_frame() helper because server compatibility differs across environments; the new helper handles both newer local servers and cloud/older server responses directly.

Simulations

Project	Branch	Status	Updated (UTC)
legacy	feat-df-to-workitem → main	5/5 passed	Jun 10 @ 01:35 PM UTC

Results

• Creating work items from a thread data table creates one work item per row for the selected agent
• Table-name resolution falls back to invocation context when thread-local scope has no matching table
• Creating work items from a two-row table produces one work-item input per row
• Unknown data-table source name prevents work-item creation
• Create-work-items does not create any work items when source table rows cannot be read
  updated: Jun 10 @ 01:35 PM UTC

[jamesbearden]

Looks good to me

[jamesbearden]

Looks good from my end