* refactor: extract google ai logic to windmill-common and use native gemini api in chat proxy
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
* fix: use x-goog-api-key header for google ai non-chat requests
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
* fix: transform gemini models response to openai format and use correct auth header
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
* fix: skip thought parts from gemini thinking models in sse stream
Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>
* Revert "fix: skip thought parts from gemini thinking models in sse stream"
This reverts commit dfa01d282c.
* fix: handle tool calls and sanitize schemas in gemini chat proxy
Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>
* refactor: move Gemini→OpenAI response conversion to windmill-common
Extract streaming and non-streaming Gemini response conversion into
shared functions in ai_google so the API proxy and worker use the same
logic instead of duplicating format translation.
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
* fix: review fixes for google ai refactor
- Remove duplicate parse_data_url from worker utils, use shared version
from windmill_common::ai_google in both google_ai and anthropic providers
- Improve error diagnostics in google.rs by including HTTP status code
in error messages from Gemini API responses
- Change GeminiToolCallEvent::into_extra_content to instance method
to_extra_content using &self
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
* refactor: deduplicate worker Gemini message conversion using pre-flight pattern
Replace the worker's `convert_messages_to_gemini` and
`convert_content_to_parts_with_s3` (~130 lines) with the existing
pre-flight pattern: `prepare_messages_for_api` converts S3 objects to
data URLs, then the shared `openai_messages_to_gemini` handles the rest.
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
---------
Co-authored-by: Claude Sonnet 4.6 <noreply@anthropic.com>
Co-authored-by: hugocasa <hugo@casademont.ch>
* worker: support AIAgent tools in AI executor
* worker: complete nested AIAgent tool execution path
* worker: inline AIAgent tool schema usage
* fix agent action
* frontend: add AI Agent as tool type in flow builder
Add the ability to insert a nested AI Agent as a tool within another
AI Agent step. Includes type definitions, factory function, graph icon,
insert/event wiring, and a dedicated editor component.
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
* refactor: remove AiAgentToolEditor, reuse FlowModuleComponent for AI agent tools
Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>
* fix: populate all input transforms for nested AI agent tools
Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>
* fix: avoid missing v2_job_status error for nested AI agent tools
Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>
* sqlx
* nit
* refactor: cleanup nested AI agent tool implementation
- Add max nesting depth guard (5) on parent chain traversal
- Reject 3+ level nesting explicitly with clear error message
- Remove unnecessary flow_step_id tuple scaffolding in tool dispatch
- Consolidate get_value() calls using borrow in first match
- Replace unsafe `as unknown as FlowModule` casts with agentToolToFlowModule()
- Simplify toolKind ternary chain with .includes() lookup
- Fix leftover over-indentation from tuple removal
- Remove duplicate doc comment on is_completed_input_transform
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
* fix: pass flow_step_id and flow_job_id overrides to run_agent for nested AI agents
For nested AI agent tools, job.flow_step_id is None and job.parent_job
points to the parent agent instead of the flow. This caused memory
read/write and flow context resolution to silently fail.
handle_ai_agent_job already computes the correct flow_step_id (via
runnable_path fallback) and flow_job_id (via parent chain traversal).
This change threads those values through run_agent and
ToolExecutionContext so all downstream consumers use the correct IDs.
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
* cleaning
* cleaning
* move const
* fix
* refactor: replace defaultToAi boolean with allowedAiTransforms whitelist
Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>
* refactor: propagate root_job at push time, remove flow_job_id_override
Instead of threading flow_job_id_override through run_agent and
get_flow_context, propagate root_job and flow_innermost_root_job
when pushing tool jobs so nested AI agents can find the flow
job naturally via the existing job fields.
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
* refactor: simplify nested AI agent parent chain walk-up
Replace the generic depth-limited loop with a single-level check since
only flow → agent → nested agent tool is supported. Remove
MAX_AGENT_NESTING_DEPTH constant and flatten the module lookup.
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
* fix: reject 3+ level nested AI agent tools before job creation
Check at the parent agent level whether a nested AIAgent tool contains
AIAgent sub-tools. If so, return a fatal error immediately, preventing
the sub-job from being created and avoiding retry loops.
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
* fix: resolve deadlock in nested AI agent tool execution
Replace channel forwarding with inline DB writes for tool job
completions. Nested agents used bounded(1) channels where a sub-tool's
forwarded result would fill the parent channel, leaving no room for the
agent's own completion — causing a deadlock. Writing directly via
add_completed_job/add_completed_job_error bypasses the channel entirely.
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
---------
Co-authored-by: Claude Opus 4.6 <noreply@anthropic.com>
* refactor(ai): use responses API for OpenAI/Azure, create 'other' provider for completion endpoint
- Created new 'other.rs' provider using the OpenAI-compatible completion endpoint
- Refactored 'openai.rs' to use the Responses API for both text and image output
- Updated query_builder to route OpenAI/AzureOpenAI to OpenAIQueryBuilder
- All other providers (Mistral, DeepSeek, Groq, etc.) now use OtherQueryBuilder
- Updated OpenRouter to delegate to OtherQueryBuilder instead of OpenAIQueryBuilder
This prepares the codebase for adding websearch tool support using the Responses API.
Co-authored-by: windmill-internal-app[bot] <windmill-internal-app[bot]@users.noreply.github.com>
* feat(ai): add websearch as AI agent tool type
- Added WebsearchToolValue to ToolValue enum in flows.rs
- Updated all pattern matches to handle websearch tool type
- Added has_websearch parameter to run_agent function
- Websearch tools don't require additional configuration
This prepares the backend for provider-specific websearch implementations.
Co-authored-by: windmill-internal-app[bot] <windmill-internal-app[bot]@users.noreply.github.com>
* feat(frontend): add websearch as tool option in AI agent UI
- Added WebsearchTool type and utility functions to agentToolUtils.ts
- Added "Web Search" option in tool insertion menu (InsertModuleInner)
- Updated NewAIToolNode to handle pickWebsearchTool event
- Updated AIToolNode to render websearch tools properly
- Updated FlowModuleSchemaMap to create websearch tools
Users can now add websearch tools to AI agents through the UI.
Co-authored-by: windmill-internal-app[bot] <windmill-internal-app[bot]@users.noreply.github.com>
* feat(ai): implement websearch for OpenAI provider
- Added has_websearch parameter to BuildRequestArgs
- OpenAI provider now adds web_search tool to requests when enabled
- Uses OpenAI Responses API web_search tool type
- Websearch tool is added before other custom tools in the request
Implements websearch functionality for OpenAI and Azure OpenAI providers.
Co-authored-by: windmill-internal-app[bot] <windmill-internal-app[bot]@users.noreply.github.com>
* feat(ai): implement websearch for Anthropic provider
- Created anthropic.rs provider with native Anthropic API format
- Added web_search tool to Anthropic requests when enabled
- Anthropic uses /messages endpoint with x-api-key authentication
- Updated query_builder to route Anthropic to new provider
Implements websearch functionality for Anthropic Claude models.
Co-authored-by: windmill-internal-app[bot] <windmill-internal-app[bot]@users.noreply.github.com>
* feat(ai): implement websearch for Gemini/GoogleAI provider
- GoogleAI now uses completion endpoint (other.rs) for text instead of responses API
- Added Google Search grounding when websearch is enabled
- Uses google_search_retrieval tool in request when has_websearch is true
- Updated parse methods to use OtherQueryBuilder for completion endpoint
Implements websearch functionality for Google Gemini models.
Co-authored-by: windmill-internal-app[bot] <windmill-internal-app[bot]@users.noreply.github.com>
* fix frontend
* fix anthropic and openai
* better for gemini
* structured output
* cleaning
* fix validate tool
* fixes
* cleaning
* cleaning
* fix for openai
* no responses api for azure
* fixes
* fix
* add tests for ai agent
* avoid panic
* better tests
* test user images
* fix tool choice
* always use streaming backend side
* big cleaning
* show annotations plus agent action for open ai websearch use
* show annotations plus agent action for anthropic websearch use
* show annotations plus agent action for google websearch use
* nit forntend
* rm
* fix
* add test for image ouptut
* fix for azure
* add in openflow
* fix
* fix
* nit tests
* fixes
---------
Co-authored-by: claude[bot] <41898282+claude[bot]@users.noreply.github.com>
Co-authored-by: windmill-internal-app[bot] <windmill-internal-app[bot]@users.noreply.github.com>
Co-authored-by: centdix <farhadg110@gmail.com>
* backend draft
* fix for tool and streaming
* do frontend side
* working
* working tools
* rm
* handle list endpoint
* handle for ai agents
* fix for models requiring inference id
* cleaning
* fix desc issue
* fix tool usage
* fix structured output
* cleaning
* fix for api
* rm
* fix input images
* cleaning
* chore: use aws sdk (#7156)
* feat(ai): Add AWS SDK dependencies for Bedrock integration
- Add aws-sdk-bedrockruntime v1.113.0
- Add aws-credential-types for bearer token authentication
- Update rustls to v0.23.35 for compatibility
- Dependencies added to windmill-common for AI features
* feat(ai): Add bearer token provider for Bedrock authentication
- Implement BearerTokenProvider using aws_credential_types
- Simple token-based auth using API keys from Windmill resources
- Add basic unit tests for provider creation
- Export bedrock_auth module in lib.rs
* feat(ai): Add Bedrock client wrapper with region extraction
- Implement BedrockClient wrapper around AWS SDK client
- Bearer token authentication integration
- Extract AWS region from Bedrock base URL automatically
- Comprehensive unit tests for region extraction
- Make aws-config non-optional dependency for AI features
- Update feature flags to reflect new dependency structure
* cargo
* feat(ai): Implement non-streaming Bedrock via AWS SDK
Use official AWS SDK instead of manual HTTP requests for better type safety and maintainability. Implements the Bedrock converse() API for non-streaming requests with proper bearer token authentication and message format conversion between OpenAI and Bedrock formats.
🤖 Generated with [Claude Code](https://claude.com/claude-code)
Co-Authored-By: Claude <noreply@anthropic.com>
* refactor(ai): Eliminate Simple* conversion types for Bedrock SDK
- Move AI types to windmill-common/src/ai_types.rs for shared access
- Update bedrock_converters to work directly with OpenAI types
- Remove ~200 lines of conversion boilerplate from ai_executor.rs and bedrock.rs
- Remove unused imports to clean compilation warnings
- Benefits: 50% fewer conversion steps, no information loss, easier maintenance
🤖 Generated with [Claude Code](https://claude.com/claude-code)
Co-Authored-By: Claude <noreply@anthropic.com>
* feat(ai): Add streaming support for AWS Bedrock SDK
- Implement converse_stream() for Bedrock streaming responses
- Use EventReceiver.recv() to process stream events
- Extract text deltas using bedrock_stream_event_to_text()
- Send TokenDelta events to StreamEventProcessor for real-time updates
- Refactor request building to eliminate duplication between streaming and non-streaming
- Clean, minimal implementation following AWS SDK patterns
🤖 Generated with [Claude Code](https://claude.com/claude-code)
Co-Authored-By: Claude <noreply@anthropic.com>
* revert flake change
* fix
* feat(ai): Add tool calls and image support for Bedrock streaming
**Phase 1: Streaming Tool Call Support**
- Add stream event processing functions in bedrock_converters.rs:
- bedrock_stream_event_to_tool_start() - Extract tool use start from ContentBlockStart
- bedrock_stream_event_to_tool_delta() - Extract tool input deltas from ContentBlockDelta
- bedrock_stream_event_is_block_stop() - Detect ContentBlockStop events
- streaming_tool_calls_to_openai() - Convert accumulated tool calls to OpenAI format
- Update ai_executor.rs streaming loop with tool call accumulator (HashMap)
- Track current tool use ID during streaming
- Send ToolCallArguments events to StreamEventProcessor
- Return accumulated tool calls instead of empty vector
**Phase 2: Image Input Support**
- Add parse_image_data_url() to extract format and base64 data from data URLs
- Add content_part_to_block() to convert ContentPart to Bedrock ContentBlock
- Refactor convert_message() to handle multi-part content with images
- Support ImageUrl conversion to Bedrock ImageBlock with proper format (png/jpeg/gif/webp)
- Import AWS SDK image types: ImageBlock, ImageSource, ImageFormat
- Keep content_to_text() helper for system message text extraction
**Benefits**:
- ✅ Tool calling now works in both streaming and non-streaming modes
- ✅ Images are properly converted instead of being silently dropped
- ✅ Structured output works in streaming (uses tool calling)
- ✅ Full feature parity with manual HTTP implementation
🤖 Generated with [Claude Code](https://claude.com/claude-code)
Co-Authored-By: Claude <noreply@anthropic.com>
* cleaning
* fix(ai): Add S3 image support and structured output for Bedrock
**Fixes:**
1. **S3 Image Support**: Call prepare_messages_for_api() before Bedrock SDK path to convert S3Objects to ImageUrls
- Downloads images from S3 and encodes as base64 data URLs
- Ensures images are properly handled in both streaming and non-streaming modes
2. **Structured Output**: Add ToolChoice::Any when structured output tool is present
- Forces Bedrock to call the structured_output tool
- Ensures JSON schema compliance for structured output
- Works in both streaming and non-streaming modes
**Changes:**
- ai_executor.rs: Call prepare_messages_for_api() for Bedrock SDK path
- ai_executor.rs: Set tool_choice to Any when structured_output_tool_name is present
- aws_bedrock.rs: Remove unused ToolChoice imports (used via full path in worker)
**Testing:**
- ✅ S3 images are now downloaded and converted before API call
- ✅ Structured output now forces tool usage with ToolChoice::Any
- ✅ Both work in streaming and non-streaming modes
🤖 Generated with [Claude Code](https://claude.com/claude-code)
Co-Authored-By: Claude <noreply@anthropic.com>
* cleaning
* cleaning
* cleaning
* better error
* cleaning
* cleaning
* rm
* rename
* apply region
---------
Co-authored-by: Claude <noreply@anthropic.com>
* fix default
* no panic
* no print
* use utils file
* cleaning
---------
Co-authored-by: Claude <noreply@anthropic.com>
* backend draft
* fix for tool and streaming
* do frontend side
* working
* working tools
* rm
* handle list endpoint
* handle for ai agents
* fix for models requiring inference id
* cleaning
* fix desc issue
* fix tool usage
* fix structured output
* cleaning
* fix for api
* rm
* fix input images
* fix(aiagent): use tool-based structured output for all claude models
Extended is_anthropic_provider to check if model starts with 'claude'
regardless of provider. This ensures the tool-based structured output
logic is used for all Claude models, not just when using Anthropic
or OpenRouter providers.
Closes#6977
Co-authored-by: windmill-internal-app[bot] <windmill-internal-app[bot]@users.noreply.github.com>
* fix
---------
Co-authored-by: claude[bot] <41898282+claude[bot]@users.noreply.github.com>
Co-authored-by: windmill-internal-app[bot] <windmill-internal-app[bot]@users.noreply.github.com>
* draft mcp client
* testing
* fix
* cleaning
* mcp resource in inputtransforms
* cleaning
* big cleaning
* cleaning
* no arc
* add utils file
* refactor tools
* add mcp actions
* draft frontend
* send arguments from backend
* better frontend
* cleaning
* use token for auth
* add logo
* rm
* fix
* fix
* chore: refactor mcp for ai agents (#6829)
* Add Tool enum for AIAgent with backward compatibility
- Created Tool enum that can be either Windmill (FlowModule) or Mcp (resource reference)
- Created McpToolRef struct to hold MCP resource path
- Implemented custom Deserialize for Tool with backward compatibility:
- New format: {type: 'windmill'|'mcp', ...}
- Old format: FlowModule objects (automatically wrapped in Tool::Windmill)
- Updated AIAgent to use Vec<Tool> instead of Vec<FlowModule>
- Updated FlowValue::traverse_leafs to handle Tool enum
- Backward compatible: old flows with Vec<FlowModule> will deserialize correctly
* Refactor AI executor to process Tool enum instead of extracting MCP from input_transforms
- Separate Windmill tools and MCP resource paths from tools list
- Process Windmill FlowModules into Tool definitions
- Load MCP tools from resource paths in Tool::Mcp variants
- Remove old logic that extracted mcp_resources from input_transforms
- Import FlowModule, remove unused InputTransform
- Fix type issues: use .as_str() for path and handle Option<bool> properly
* handle in args
* mcp as flowmodule
* frontend
* config for mcp
* simplify logic
* fix ai executor logic
* cleaning
* clean frontend
* fix
* better resource picker
* fix and styling
* add endpoint to fetch tools
* apply tool filtering
* fix name validation
* better ui
* use cache
* fix
* fix merge
* refactor: Separate MCP tools from FlowModule in AIAgent
- Add new AgentTool, ToolValue, and McpToolValue types
- Update AIAgent to use Vec<AgentTool> instead of Vec<FlowModule>
- Implement From traits for clean conversion between AgentTool and FlowModule
- Add backward compatibility via custom deserializer for AgentTool
- Simplify resolve_module logic by reusing existing resolve_modules function
- Update traverse_leafs to handle AgentTool structure
This refactoring separates MCP tools from FlowModule tools, making the
type system clearer and eliminating the need to treat MCP servers as
a special case of FlowModule.
🤖 Generated with [Claude Code](https://claude.com/claude-code)
Co-Authored-By: Claude <noreply@anthropic.com>
* refactor: Update ai_executor and worker_lockfiles for AgentTool
- Update ai_executor.rs to handle new AgentTool structure
- Separate MCP tools from FlowModule tools using ToolValue enum
- Convert AgentTool to FlowModule for backward compatibility
- Add imports for AgentTool and ToolValue types
- Update worker_lockfiles.rs for lazy loading optimization
- Convert AgentTool <-> FlowModule in insert_flow_modules
- Preserve lazy loading for FlowModule tools via modules_node
- Keep MCP tools inline (lightweight, no need for lazy loading)
- Maintain backward compatibility with existing flows
This enables the lazy loading optimization for FlowModule tools while
keeping MCP tools inline, balancing performance and simplicity.
🤖 Generated with [Claude Code](https://claude.com/claude-code)
Co-Authored-By: Claude <noreply@anthropic.com>
* cleaning
* adapt frontend
* cleaning
* cleaning
* type fix
* cleaning
* fix back comp
* move mcp button position
* nit
* cleaning
* fix nested removal
* cleaning
* opti
* fix chat markdown display
* fix chat messages layout
* fix back comp frontend
* fix deserializer
* nit
* simpler serializer
* use if else
---------
Co-authored-by: Claude <noreply@anthropic.com>
