inference-gateway · edenreich · Oct 14, 2025 · Oct 14, 2025 · Oct 14, 2025 · Oct 14, 2025
diff --git a/.github/workflows/cd.yml b/.github/workflows/cd.yml
diff --git a/.github/workflows/ci.yml b/.github/workflows/ci.yml
diff --git a/.releaserc.yaml b/.releaserc.yaml
diff --git a/.well-known/agent-card.json b/.well-known/agent-card.json
diff --git a/AGENTS.md b/AGENTS.md
@@ -30,7 +30,7 @@ This agent is built using the Agent Definition Language (ADL) and provides A2A c
 
 
 
-**System Prompt**: You are an expert Playwright browser automation assistant. Your primary role is to help users automate web browser tasks efficiently and reliably.
+**System Prompt**: You are an expert Playwright browser automation assistant with the ability to create downloadable artifacts. Your primary role is to help users automate web browser tasks efficiently and reliably.
 
 Your core capabilities include:
 1. **Web Navigation**: Navigate to URLs, handle redirects, and manage page loads
@@ -41,6 +41,7 @@ Your core capabilities include:
 6. **JavaScript Execution**: Run custom scripts in the browser context
 7. **Authentication Handling**: Manage various authentication methods
 8. **Synchronization**: Wait for specific conditions and handle dynamic content
+9. **Artifact Creation**: Create downloadable files for screenshots, extracted data, and CSV exports
 
 Key expertise areas:
 - Modern web technologies (SPA, dynamic content, AJAX)
@@ -63,6 +64,11 @@ When helping users:
 - Provide clear explanations of automation steps
 - Optimize for speed while maintaining reliability
 
+**IMPORTANT - Artifact Creation**:
+When users request screenshots, the take_screenshot tool automatically creates downloadable artifacts. The screenshot will be available via a download URL returned in the response.
+
+For data extraction, you can use the create_artifact tool to save extracted data as downloadable files (JSON/CSV/TXT).
+
 Your automation solutions should be maintainable, efficient, and production-ready.
 
 
@@ -75,7 +81,7 @@ Your automation solutions should be maintainable, efficient, and production-read
 ## Skills
 
 
-This agent provides 9 skills:
+This agent provides 8 skills:
 
 
 ### navigate_to_url
@@ -134,13 +140,6 @@ This agent provides 9 skills:
 - **Output Schema**: Defined in agent configuration
 
 
-### write_to_csv
-- **Description**: Write structured data to CSV files with support for custom headers and file paths
-- **Tags**: export, csv, data, file
-- **Input Schema**: Defined in agent configuration
-- **Output Schema**: Defined in agent configuration
-
-
 
 
 ## Server Configuration
@@ -246,11 +245,6 @@ curl -X POST http://localhost:8080/skills/wait_for_condition \
   -H "Content-Type: application/json" \
   -d '{"input": "your_input_here"}'
 
-# Execute write_to_csv skill
-curl -X POST http://localhost:8080/skills/write_to_csv \
-  -H "Content-Type: application/json" \
-  -d '{"input": "your_input_here"}'
-
 
 ```
 
@@ -298,8 +292,6 @@ docker run -p 8080:8080 browser-agent
 
 │   └── wait_for_condition.go   # Wait for specific conditions before proceeding with automation
 
-│   └── write_to_csv.go   # Write structured data to CSV files with support for custom headers and file paths
-
 ├── .well-known/         # Agent configuration
 │   └── agent-card.json  # Agent metadata
 ├── go.mod               # Go module definition

diff --git a/CLAUDE.md b/CLAUDE.md
@@ -10,7 +10,7 @@ browser-agent is an A2A (Agent-to-Agent) server implementing the [A2A Protocol](
 
 ### ADL-Generated Structure
 
-The codebase is generated using ADL CLI 0.23.2 and follows a strict generation pattern:
+The codebase is generated using ADL CLI 0.23.6 and follows a strict generation pattern:
 - **Generated Files**: Marked with `DO NOT EDIT` headers - manual changes will be overwritten
 - **Configuration Source**: `agent.yaml` - defines agent capabilities, skills, and metadata
 - **Server Implementation**: Built on the ADK (Agent Development Kit) framework from `github.com/inference-gateway/adk`
@@ -82,7 +82,6 @@ The following skills are currently defined:
 - **execute_script**: Execute custom JavaScript code in the browser context
 - **handle_authentication**: Handle various authentication scenarios including basic auth, OAuth, and custom login forms
 - **wait_for_condition**: Wait for specific conditions before proceeding with automation
-- **write_to_csv**: Write structured data to CSV files with support for custom headers and file paths
 
 To modify skills:
 1. Update `agent.yaml` with skill definitions
@@ -118,7 +117,7 @@ Activate with: `flox activate` (if Flox is installed)
 
 - **Generated Files**: Never manually edit files with "DO NOT EDIT" headers
 - **Configuration Changes**: Always modify `agent.yaml` and regenerate
-- **ADL Version**: Ensure ADL CLI 0.23.2 or compatible version for regeneration
+- **ADL Version**: Ensure ADL CLI 0.23.6 or compatible version for regeneration
 - **Port Configuration**: Default 8080, configurable via `A2A_PORT` or `A2A_SERVER_PORT`
 
 ## Debugging Tips

diff --git a/README.md b/README.md
@@ -49,7 +49,6 @@ docker run -p 8080:8080 browser-agent
 | `execute_script` | Execute custom JavaScript code in the browser context |args, return_value, script |
 | `handle_authentication` | Handle various authentication scenarios including basic auth, OAuth, and custom login forms |login_url, password, password_selector, submit_selector, type, username, username_selector |
 | `wait_for_condition` | Wait for specific conditions before proceeding with automation |condition, custom_function, selector, state, timeout |
-| `write_to_csv` | Write structured data to CSV files with support for custom headers and file paths |append, data, filename, headers, include_headers |
 
 ## Configuration
 

diff --git a/Taskfile.yml b/Taskfile.yml
diff --git a/agent.yaml b/agent.yaml
@@ -310,50 +310,12 @@ spec:
       inject:
         - logger
         - playwright
-    - id: write_to_csv
-      name: write_to_csv
-      description: Write structured data to CSV files with support for custom headers and file paths
-      tags:
-        - export
-        - csv
-        - data
-        - file
-      schema:
-        type: object
-        properties:
-          data:
-            type: array
-            items:
-              type: object
-            description: Array of objects to write to CSV, each object represents a row
-          filename:
-            type: string
-            description: Name of the CSV file (without path, will be saved to configured data directory)
-          headers:
-            type: array
-            items:
-              type: string
-            description: Custom column headers for the CSV file (optional, will use object keys if not provided)
-          append:
-            type: boolean
-            description: Whether to append to existing file or create new file
-            default: false
-          include_headers:
-            type: boolean
-            description: Whether to include headers in the CSV output
-            default: true
-        required:
-          - data
-          - filename
-      inject:
-        - logger
-        - playwright
   agent:
     provider: ""
     model: ""
     systemPrompt: |
-      You are an expert Playwright browser automation assistant. Your primary role is to help users automate web browser tasks efficiently and reliably.
-      
+      You are an expert Playwright browser automation assistant with the ability to create downloadable artifacts. Your primary role is to help users automate web browser tasks efficiently and reliably.
+
       Your core capabilities include:
       1. **Web Navigation**: Navigate to URLs, handle redirects, and manage page loads
       2. **Element Interaction**: Click buttons, fill forms, select dropdowns, and interact with any web element
@@ -363,7 +325,8 @@ spec:
       6. **JavaScript Execution**: Run custom scripts in the browser context
       7. **Authentication Handling**: Manage various authentication methods
       8. **Synchronization**: Wait for specific conditions and handle dynamic content
-
+      9. **Artifact Creation**: Create downloadable files for screenshots, extracted data, and CSV exports
+
       Key expertise areas:
       - Modern web technologies (SPA, dynamic content, AJAX)
       - Selector strategies (CSS, XPath, text, accessibility)
@@ -375,7 +338,7 @@ spec:
       - File uploads and downloads
       - Network interception and modification
       - Mobile and responsive testing
-      
+
       When helping users:
       - Always use robust selectors that won't break easily
       - Implement proper wait strategies for dynamic content
@@ -384,7 +347,12 @@ spec:
       - Consider accessibility and best practices
       - Provide clear explanations of automation steps
       - Optimize for speed while maintaining reliability
-
+
+      **IMPORTANT - Artifact Creation**:
+      When users request screenshots, the take_screenshot tool automatically creates downloadable artifacts. The screenshot will be available via a download URL returned in the response.
+
+      For data extraction, you can use the create_artifact tool to save extracted data as downloadable files (JSON/CSV/TXT).
+
       Your automation solutions should be maintainable, efficient, and production-ready.
   services:
     playwright:

diff --git a/config/config.go b/config/config.go
diff --git a/example/docker-compose.yaml b/example/docker-compose.yaml
@@ -53,6 +53,7 @@ services:
       A2A_AGENT_CLIENT_MAX_CHAT_COMPLETION_ITERATIONS: 20
       A2A_AGENT_CLIENT_MAX_TOKENS: 4096
       A2A_AGENT_CLIENT_TEMPERATURE: 0.7
+      A2A_AGENT_CLIENT_TOOLS_CREATE_ARTIFACT: true
       A2A_CAPABILITIES_STREAMING: true
       A2A_CAPABILITIES_PUSH_NOTIFICATIONS: false
       A2A_CAPABILITIES_STATE_TRANSITION_HISTORY: false

diff --git a/go.mod b/go.mod
diff --git a/go.sum b/go.sum
diff --git a/internal/logger/logger.go b/internal/logger/logger.go
diff --git a/main.go b/main.go