Static app tokens expire after ~60 days -> every lookup failed with
401. With igdb-client-secret set, the bot fetches the token via
client-credentials OAuth itself, refreshes a day before expiry and
retries once on 401; a static igdb-access-token still works.
IGDB renamed games.category to game_type: the category = 0 filter in
search_games silently matched nothing even with a valid token.
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
## Major Features Added
- **Enhanced igdblib.py**:
* Added search_games() method with fuzzy game search
* Added get_game_details() for comprehensive game information
* Added AI-friendly data formatting with _format_game_for_ai()
* Added OpenAI function definitions via get_openai_functions()
- **OpenAI Function Calling Integration**:
* Modified OpenAIResponder to support function calling
* Added IGDB function execution with _execute_igdb_function()
* Backward compatible - gracefully falls back if IGDB unavailable
* Auto-detects gaming queries and fetches real-time data
- **Configuration & Setup**:
* Added IGDB configuration options to config.toml
* Updated system prompt to inform AI of gaming capabilities
* Added comprehensive IGDB_SETUP.md documentation
* Graceful initialization with proper error handling
## Technical Implementation
- **Function Calling**: Uses OpenAI's tools/function calling API
- **Smart Game Search**: Includes ratings, platforms, developers, genres
- **Error Handling**: Robust fallbacks and logging
- **Data Formatting**: Optimized for AI comprehension and user presentation
- **Rate Limiting**: Respects IGDB API limits
## Usage
Users can now ask natural gaming questions:
- "Tell me about Elden Ring"
- "What are good RPG games from 2023?"
- "Is Cyberpunk 2077 on PlayStation?"
The AI automatically detects gaming queries, calls IGDB API, and presents
accurate, real-time game information seamlessly.
🤖 Generated with [Claude Code](https://claude.ai/code)
Co-Authored-By: Claude <noreply@anthropic.com>
- Remove temperature, max_tokens, top_p, presence_penalty, frequency_penalty parameters
- These parameters are no longer supported in GPT-5 models
- Update all openai_chat calls to use only model and messages parameters
- Fix openai-key vs openai-token configuration key compatibility
🤖 Generated with [Claude Code](https://claude.ai/code)
Co-Authored-By: Claude <noreply@anthropic.com>
- Fix infinite retry loop in ai_responder.py that caused test_fix1 to hang
- Add missing picture_edit parameter to all AIResponse constructor calls
- Set up complete development toolchain with Black, isort, Bandit, and MyPy
- Create comprehensive Makefile for development workflows
- Add pre-commit hooks with formatting, linting, security, and type checking
- Update test mocking to provide contextual responses for different scenarios
- Configure all tools for 140 character line length and strict type checking
- Add DEVELOPMENT.md with setup instructions and workflow documentation
🤖 Generated with [Claude Code](https://claude.ai/code)
Co-Authored-By: Claude <noreply@anthropic.com>
Oleksandr Kozachuk
ok