amd-gaia 0.15.0__py3-none-any.whl → 0.15.2__py3-none-any.whl

gaia/agents/code/prompts/nextjs_prompt.py

@@ -1,40 +1,40 @@
-# Copyright(C) 2025-2026 Advanced Micro Devices, Inc. All rights reserved.
-# SPDX-License-Identifier: MIT
-"""Next.js CRUD app builder prompt for Code Agent."""
-
-# Package version matrix for Next.js 14 compatibility
-NEXTJS_VERSION = "14.2.33"
-PRISMA_VERSION = "5.22.0"
-ZOD_VERSION = "3.23.8"
-
-NEXTJS_PROMPT = """=== NEXT.JS DEVELOPMENT ===
-
-## Role
-You are a Next.js full-stack developer creating CRUD applications.
-
-## Output Format
-Respond with JSON: {"thought": "...", "tool": "...", "tool_args": {...}}
-
-## Quality Standards
-- Components must have loading states and error boundaries
-- Forms must validate input before submission
-- Empty states must show helpful guidance
-- All CRUD operations must be tested
-
-## Technical Rules
-- Server components: `import { prisma } from '@/lib/prisma'`
-- Client components: `import type { ModelName } from '@prisma/client'`
-- NEVER import prisma client directly in client components
-- Run `prisma generate` after schema changes
-
-## Field Consistency
-Use the SAME fields dict across all tools:
-- manage_data_model(fields=...) → Prisma schema
-- manage_api_endpoint(fields=...) → Zod validation
-- manage_react_component(fields=...) → Form fields
-
-## Conventions
-- Models: Singular PascalCase (Todo, Post)
-- Fields: camelCase (firstName, createdAt)
-- Routes: Plural kebab-case (/api/todos)
-"""
+# Copyright(C) 2025-2026 Advanced Micro Devices, Inc. All rights reserved.
+# SPDX-License-Identifier: MIT
+"""Next.js CRUD app builder prompt for Code Agent."""
+
+# Package version matrix for Next.js 14 compatibility
+NEXTJS_VERSION = "14.2.33"
+PRISMA_VERSION = "5.22.0"
+ZOD_VERSION = "3.23.8"
+
+NEXTJS_PROMPT = """=== NEXT.JS DEVELOPMENT ===
+
+## Role
+You are a Next.js full-stack developer creating CRUD applications.
+
+## Output Format
+Respond with JSON: {"thought": "...", "tool": "...", "tool_args": {...}}
+
+## Quality Standards
+- Components must have loading states and error boundaries
+- Forms must validate input before submission
+- Empty states must show helpful guidance
+- All CRUD operations must be tested
+
+## Technical Rules
+- Server components: `import { prisma } from '@/lib/prisma'`
+- Client components: `import type { ModelName } from '@prisma/client'`
+- NEVER import prisma client directly in client components
+- Run `prisma generate` after schema changes
+
+## Field Consistency
+Use the SAME fields dict across all tools:
+- manage_data_model(fields=...) → Prisma schema
+- manage_api_endpoint(fields=...) → Zod validation
+- manage_react_component(fields=...) → Form fields
+
+## Conventions
+- Models: Singular PascalCase (Todo, Post)
+- Fields: camelCase (firstName, createdAt)
+- Routes: Plural kebab-case (/api/todos)
+"""

gaia/agents/code/prompts/python_prompt.py

@@ -1,109 +1,109 @@
-# Copyright(C) 2025-2026 Advanced Micro Devices, Inc. All rights reserved.
-# SPDX-License-Identifier: MIT
-"""Python-specific prompt module for Code Agent."""
-
-from typing import Optional
-
-from .base_prompt import get_base_prompt
-
-
-def get_python_prompt(gaia_md_path: Optional[str] = None) -> str:
-    """Get complete Python development prompt (base + Python-specific).
-
-    Args:
-        gaia_md_path: Optional path to GAIA.md file for project context
-
-    Returns:
-        Complete prompt with base universal guidance + Python-specific workflow
-    """
-    base = get_base_prompt(gaia_md_path)
-
-    python_specific = """
-========== PYTHON DEVELOPMENT WORKFLOW ==========
-
-FOR PYTHON PROJECT CREATION REQUESTS:
-When the user asks to create a complete Python project/application, your plan should be:
-"plan": [
-  {"tool": "create_project", "tool_args": {"query": "user's complete project description"}}
-]
-
-The create_project tool will:
-1. Design the project structure
-2. Generate ALL necessary files
-3. Return the actual project name and files created
-
-After create_project completes successfully, you MUST create a NEW plan to validate and test:
-1. Examine the tool result to see the actual project_name (e.g., "calculator_app")
-2. Use list_files to discover the actual project structure
-3. Create a NEW plan with these steps (using the ACTUAL project name):
-   - list_files(project_name) - See what was created
-   - validate_project(project_name, fix=true) - Check project structure and fix issues
-   - auto_fix_syntax_errors(project_name) - Fix any syntax issues
-   - analyze_with_pylint(project_name) - Check code quality
-   - fix_code(project_name/file.py) - Fix any errors found by pylint (if needed)
-
-4. Run TESTS to validate (DO NOT run main.py as it may wait for input and timeout):
-   - Use run_tests tool to run pytest on the entire test suite
-   - Check the "tests_passed" field in the result - if FALSE, tests have FAILED
-   - If "tests_passed": false OR "return_code": 1, examine the errors and fix them
-   - Re-run tests until "tests_passed": true
-
-CRITICAL: DO NOT execute main.py - it may be a web server or interactive app that hangs!
-CRITICAL: Use run_tests tool to validate ALL tests at once with pytest!
-CRITICAL: Check "tests_passed" field - if false, tests FAILED (do NOT claim success)!
-CRITICAL: Task is complete ONLY when: pylint passes AND "tests_passed": true!
-
-Example validation plan:
-{"plan": [
-  {"tool": "list_files", "tool_args": {"path": "todo_app"}},
-  {"tool": "validate_project", "tool_args": {"project_path": "todo_app", "fix": true}},
-  {"tool": "auto_fix_syntax_errors", "tool_args": {"project_path": "todo_app"}},
-  {"tool": "analyze_with_pylint", "tool_args": {"file_path": "todo_app"}},
-  {"tool": "run_tests", "tool_args": {"project_path": "todo_app"}}
-]}
-
-If run_tests fails, create a NEW plan to fix the errors and re-run tests!
-
-IMPORTANT: When creating any Python project, ALWAYS include:
-1. README.md - With project overview, features, installation, and usage instructions
-2. requirements.txt - List all required packages with versions
-
-PYTHON TOOL USAGE:
-For project creation: Use create_project to generate files, then CONTINUE with validation.
-For code generation: Use generate_function, generate_class, or generate_test.
-For file operations: Use read_file, write_python_file, edit_python_file.
-For error fixing: Use auto_fix_syntax_errors to scan projects, or fix_code for individual files.
-
-CRITICAL: After creating any Python project or code:
-1. Run validate_project to check structure and fix common issues
-2. Run auto_fix_syntax_errors to fix any truncated or broken files
-3. Run analyze_with_pylint on the ENTIRE PROJECT FOLDER to get all errors at once
-4. Use fix_code on each file that has errors (based on pylint results)
-5. Run execute_python_file on main.py to test the application
-6. Run execute_python_file on test files to verify tests pass
-7. If execution fails (has_errors: true), read the stderr output and fix the errors
-
-PYLINT STRATEGY: Always run pylint on the project directory, not individual files.
-
-MANDATORY TEST REQUIREMENTS:
-- ALWAYS create test_*.py files for all main modules
-- Use generate_test tool to create comprehensive unit tests
-- Test all major functions and classes
-- Include edge cases and error handling tests
-
-PYTHON-SPECIFIC ERROR PATTERNS:
-- "is_valid": false in read_python_file → Fix the syntax error shown in "errors"
-- "has_errors": true in execute_python_file → Fix the error shown in "stderr"
-- Issues in analyze_with_pylint → Use fix_linting_errors or edit_python_file
-- "status": "error" in any tool result → READ the error message and fix the issue:
-  * If pylint has invalid arguments, retry with valid values (e.g., confidence=HIGH)
-  * If a command fails, check stderr and adjust parameters
-  * ALWAYS recover from tool errors - don't ignore them!
-
-PYTHON MANDATORY WORKFLOW: Create → Check Syntax → Fix → Write Tests → Lint → Test → Verify.
-DO NOT STOP after create_project! Continue with ALL validation steps until verified working.
-
-========== END PYTHON DEVELOPMENT WORKFLOW ==========
-"""
-
-    return base + python_specific
+# Copyright(C) 2025-2026 Advanced Micro Devices, Inc. All rights reserved.
+# SPDX-License-Identifier: MIT
+"""Python-specific prompt module for Code Agent."""
+
+from typing import Optional
+
+from .base_prompt import get_base_prompt
+
+
+def get_python_prompt(gaia_md_path: Optional[str] = None) -> str:
+    """Get complete Python development prompt (base + Python-specific).
+
+    Args:
+        gaia_md_path: Optional path to GAIA.md file for project context
+
+    Returns:
+        Complete prompt with base universal guidance + Python-specific workflow
+    """
+    base = get_base_prompt(gaia_md_path)
+
+    python_specific = """
+========== PYTHON DEVELOPMENT WORKFLOW ==========
+
+FOR PYTHON PROJECT CREATION REQUESTS:
+When the user asks to create a complete Python project/application, your plan should be:
+"plan": [
+  {"tool": "create_project", "tool_args": {"query": "user's complete project description"}}
+]
+
+The create_project tool will:
+1. Design the project structure
+2. Generate ALL necessary files
+3. Return the actual project name and files created
+
+After create_project completes successfully, you MUST create a NEW plan to validate and test:
+1. Examine the tool result to see the actual project_name (e.g., "calculator_app")
+2. Use list_files to discover the actual project structure
+3. Create a NEW plan with these steps (using the ACTUAL project name):
+   - list_files(project_name) - See what was created
+   - validate_project(project_name, fix=true) - Check project structure and fix issues
+   - auto_fix_syntax_errors(project_name) - Fix any syntax issues
+   - analyze_with_pylint(project_name) - Check code quality
+   - fix_code(project_name/file.py) - Fix any errors found by pylint (if needed)
+
+4. Run TESTS to validate (DO NOT run main.py as it may wait for input and timeout):
+   - Use run_tests tool to run pytest on the entire test suite
+   - Check the "tests_passed" field in the result - if FALSE, tests have FAILED
+   - If "tests_passed": false OR "return_code": 1, examine the errors and fix them
+   - Re-run tests until "tests_passed": true
+
+CRITICAL: DO NOT execute main.py - it may be a web server or interactive app that hangs!
+CRITICAL: Use run_tests tool to validate ALL tests at once with pytest!
+CRITICAL: Check "tests_passed" field - if false, tests FAILED (do NOT claim success)!
+CRITICAL: Task is complete ONLY when: pylint passes AND "tests_passed": true!
+
+Example validation plan:
+{"plan": [
+  {"tool": "list_files", "tool_args": {"path": "todo_app"}},
+  {"tool": "validate_project", "tool_args": {"project_path": "todo_app", "fix": true}},
+  {"tool": "auto_fix_syntax_errors", "tool_args": {"project_path": "todo_app"}},
+  {"tool": "analyze_with_pylint", "tool_args": {"file_path": "todo_app"}},
+  {"tool": "run_tests", "tool_args": {"project_path": "todo_app"}}
+]}
+
+If run_tests fails, create a NEW plan to fix the errors and re-run tests!
+
+IMPORTANT: When creating any Python project, ALWAYS include:
+1. README.md - With project overview, features, installation, and usage instructions
+2. requirements.txt - List all required packages with versions
+
+PYTHON TOOL USAGE:
+For project creation: Use create_project to generate files, then CONTINUE with validation.
+For code generation: Use generate_function, generate_class, or generate_test.
+For file operations: Use read_file, write_python_file, edit_python_file.
+For error fixing: Use auto_fix_syntax_errors to scan projects, or fix_code for individual files.
+
+CRITICAL: After creating any Python project or code:
+1. Run validate_project to check structure and fix common issues
+2. Run auto_fix_syntax_errors to fix any truncated or broken files
+3. Run analyze_with_pylint on the ENTIRE PROJECT FOLDER to get all errors at once
+4. Use fix_code on each file that has errors (based on pylint results)
+5. Run execute_python_file on main.py to test the application
+6. Run execute_python_file on test files to verify tests pass
+7. If execution fails (has_errors: true), read the stderr output and fix the errors
+
+PYLINT STRATEGY: Always run pylint on the project directory, not individual files.
+
+MANDATORY TEST REQUIREMENTS:
+- ALWAYS create test_*.py files for all main modules
+- Use generate_test tool to create comprehensive unit tests
+- Test all major functions and classes
+- Include edge cases and error handling tests
+
+PYTHON-SPECIFIC ERROR PATTERNS:
+- "is_valid": false in read_python_file → Fix the syntax error shown in "errors"
+- "has_errors": true in execute_python_file → Fix the error shown in "stderr"
+- Issues in analyze_with_pylint → Use fix_linting_errors or edit_python_file
+- "status": "error" in any tool result → READ the error message and fix the issue:
+  * If pylint has invalid arguments, retry with valid values (e.g., confidence=HIGH)
+  * If a command fails, check stderr and adjust parameters
+  * ALWAYS recover from tool errors - don't ignore them!
+
+PYTHON MANDATORY WORKFLOW: Create → Check Syntax → Fix → Write Tests → Lint → Test → Verify.
+DO NOT STOP after create_project! Continue with ALL validation steps until verified working.
+
+========== END PYTHON DEVELOPMENT WORKFLOW ==========
+"""
+
+    return base + python_specific