Skip to content

feat: Add DataFrame and Data type passthrough support for RunFlow#7

Open
mihirathale98 wants to merge 14 commits intomainfrom
ma/data-fix-runflow
Open

feat: Add DataFrame and Data type passthrough support for RunFlow#7
mihirathale98 wants to merge 14 commits intomainfrom
ma/data-fix-runflow

Conversation

@mihirathale98
Copy link
Collaborator

Summary

  • Enables DataFrame and Data objects to be passed through RunFlow tweaks to subflow components
  • Detects components with unconnected required DataFrame/Data inputs and exposes them as RunFlow fields
  • Fixes ParameterHandler to process 'other' type fields when values are set via tweaks

Changes

  • run_flow.py: Add _get_data_entry_vertices() and _get_data_fields_from_vertex() methods
  • param_handler.py: Update should_skip_field() and process_field_parameters() to handle 'other' type fields

Test plan

  • Test RunFlow with DataFrame Operations subflow - verify DataFrame passes through
  • Test RunFlow with Data type inputs - verify Data objects pass through
  • Verify existing string-based inputs still work as expected

mihirathale98 and others added 14 commits January 14, 2026 11:49
Adds OAuth authentication capability for HTTP-based MCP servers:
- Added enable_oauth toggle and oauth_url input fields
- Implemented _get_oauth_token method using MCPOAuthClient
- Inject Authorization header for HTTP-based servers when OAuth is enabled
- Handle 401 errors by clearing OAuth token for re-authentication
- Improved error handling for ExceptionGroup/TaskGroup errors
- Changed from process_tweaks_on_graph to process_tweaks for proper tweak application
- Added _normalize_input_value method to handle Message/Data objects from wired components
- Updated update_input_types to ensure Message type is accepted for ChatInput/TextInput connections
- Rewrote _run_flow_with_cached_graph to apply tweaks to flow data before Graph creation
- Added _infer_input_type to detect input type from vertex ID
- Added _get_parent_input_value_by_type for auto-forwarding parent inputs to sub-flows
- Enhanced _build_inputs_from_tweaks with input type inference and parent value forwarding
- Added ChatGoogleGenAI wrapper in base/models that bridges google-genai SDK with LangChain
- Updated Vertex AI component to use new SDK wrapper instead of langchain-google-vertexai
- Added thinking_level dropdown for Gemini 3 reasoning depth configuration
- Implements feature parity with ChatVertexAI: tool calling, structured output, multimodal, streaming
- Updated default model to gemini-3-flash-preview
- Changed default location to "global" for Gemini 3 compatibility
…t-patch

feat: Add OAuth support for MCP HTTP-based servers
fix: Improve RunFlow component tweak handling and input normalization
…nent-patch

feat: Add Gemini 3 support via google-genai SDK for Vertex AI
- Add Transpose: flip rows and columns
- Add Reset Index: reset to default integer index
- Add Set Index: set a column as the DataFrame index
- Add Fill NA: fill missing values with specified value
- Add Drop NA: remove rows with missing values
- Add Sample: random sampling of rows with optional seed
- Add Slice: get rows by index range
- Add Unique Values: get unique values from a column

Uses safe DataFrame conversion pattern to avoid NaN issues with
Langflow's custom DataFrame class.
…tions-enhancement

feat: Add 8 new DataFrame operations for enhanced data manipulation
…e-id-passing

feat: add trace id to langlfow cli
- Remove input_types=[] from ChatInput to allow incoming Message connections
- Add update_input_types() call in update_build_config_from_graph so RunFlow
  dynamic inputs accept Message type
- Preserve lists in _normalize_input_value to prevent files=[] from being
  converted to string '[]' which caused ChatOutputResponse validation errors

Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>
fix: Enable ChatInput wiring to RunFlow and fix list serialization
- Add _get_data_entry_vertices() to detect components with unconnected DataFrame/Data inputs
- Add _get_data_fields_from_vertex() to extract field configs with 'other' type mapped to Any
- Update should_skip_field() to not skip 'other' type fields when they have a value
- Add handler in process_field_parameters() for 'other' type fields from tweaks

This enables DataFrame and Data objects to be passed through RunFlow tweaks
to subflow components without requiring string conversion.
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

1 participant