ConstruQtor
The construQtor agent is the code generator of the QonQrete system. It takes detailed instructions from briq files and translates them into actual code written to the qodeyard directory.
v1.0.0 Highlight: Bulletproof Language Detection
ConstruQtor now includes 400+ language keywords for parsing AI-generated code blocks, ensuring compatibility with ALL major AI providers:
| Provider | Status |
|---|---|
| OpenAI GPT-4/GPT-4o | ✅ BULLETPROOF |
| Google Gemini | ✅ BULLETPROOF |
| Anthropic Claude | ✅ BULLETPROOF |
| DeepSeek Coder | ✅ BULLETPROOF |
| Qwen/Qwen2.5-Coder | ✅ BULLETPROOF |
How it works
1. Input and Setup:
- Takes
input_dir(briq files),summary_output, andchanged_files_output - Loads
config.yamlfor AI provider, model, anduse_qompressorsetting - Determines operational
modefrom environment variables
2. Context Provisioning:
- If
use_qompressoris enabled andbloq.dexists, uses skeleton files - Otherwise, uses full source code from
qodeyard
3. AI Prompt Construction:
- Defines the AI’s role as “construQtor” to implement briq plans
- Includes
mode_personafor coding style guidance - Provides strict “MANDATORY NAMING CONVENTIONS” for functions/methods
- Specifies required output format: Markdown code blocks with
language:path/to/file.ext
4. AI Code Generation:
- Sends briq content, context files, and prompt to configured AI model
- Handles potential failures and malformed outputs
5. Code Extraction and Writing:
- Parses AI response for Markdown code blocks
- Extracts filename and code content from each block
- Security: Validates file paths to prevent writing outside
qodeyard - Creates necessary subdirectories
- Performs basic
validate_code(e.g., Python syntax check)
6. Summary Generation:
- Tracks execution status for each briq
- Generates
summary_outputwith overall and individual statuses - Generates
changed_files_outputlisting all created/modified files
What’s configurable
config.yaml (under agents.construqtor):
provider(string): AI provider (e.g., ‘gemini’, ‘openai’). Default: ‘gemini’model(string): AI model (e.g., ‘gemini-1.5-pro’, ‘gpt-4o’). Default: ‘gemini-1.5-pro’
config.yaml (under options):
use_qompressor(boolean): Whether to use compressed “skeletons” or full code as context
Environment Variables:
QONQ_MODE: Sets the coding style personaCYCLE_NUM: Current cycle number for file naming