v1.2.0

Added

• Anthropic API support: ask2api now supports Anthropic's Claude models (Opus 4.6, Sonnet 4.5, Haiku 4.5) in addition to OpenAI
• Provider selection: New ASK2API_PROVIDER environment variable to explicitly choose between openai and anthropic
• Provider-agnostic API key: New ASK2API_API_KEY environment variable that works with any provider
• Anthropic-specific functions:
  • convert_schema_to_anthropic_tool() - Converts JSON Schema to Anthropic tool definition
  • build_anthropic_payload() - Builds API payload using tool calling format
  • build_anthropic_headers() - Sets Anthropic-specific headers (x-api-key, anthropic-version)
  • parse_anthropic_response() - Extracts structured output from tool_use response
  • prepare_anthropic_image_content() - Handles Anthropic's image format
  • convert_content_for_anthropic() - Converts multimodal content between formats
• Provider constants: OPENAI_BASE_URL, ANTHROPIC_BASE_URL, OPENAI_DEFAULT_MODEL, ANTHROPIC_DEFAULT_MODEL, ANTHROPIC_VERSION
• Intelligent API key fallback: Automatically falls back to ANTHROPIC_API_KEY or OPENAI_API_KEY based on selected provider
• Vision support for Anthropic: Image analysis works with both URL and base64-encoded local files
• Enhanced help text: CLI help now shows provider-specific environment variables

Changed

• Config class refactored: Added provider field with automatic provider-specific defaults
• Provider routing: generate_api_response() now routes between OpenAI and Anthropic based on configuration
• Image handling: prepare_image_content() now accepts provider parameter for format-specific handling
• Function naming: Renamed OpenAI-specific functions for clarity:
  • build_payload() → build_openai_payload()
  • build_headers() → build_openai_headers()
• Response parsing: Extracted OpenAI response parsing into parse_openai_response() function
• Default provider: Defaults to openai when no provider is specified (100% backward compatible)
• Documentation: Updated README with comprehensive provider support guide and examples

Technical Details

• Anthropic support uses tool calling to achieve structured output (no native json_schema format)
• Single-file architecture maintained (~330 → 501 lines)
• Zero new dependencies (still only requests)
• Clean separation between OpenAI and Anthropic code paths