InspeQtor

The inspeQtor agent acts as a “senior software quality engineer” responsible for comprehensive code review after construQtor generates code. It produces a reQap (report) for the gateQeeper to review.

How it works

1. Input Gathering:

Takes summary_path, changed_files_path, and reqap_output_path
Reads construQtor’s summary and list of changed files
Loads config.yaml for AI settings and use_qontextor flag

2. Architectural Context:

If use_qontextor is enabled, collects all .q.yaml files from qontext.d
These contain rich contextual information (symbols, purposes, dependencies)
Falls back to qodeyard if qontextor is not used

3. Changed Code Inclusion:

Reads actual content of changed files from qodeyard
Allows direct inspection of newly generated code

4. AI Prompt Construction (Reviewer Prompt):

Assigns the AI the role of “inspeQtor” with specific review criteria:
- Correctness: Logical soundness, absence of bugs
- Completeness: Whether briqs were fully addressed
- Consistency: ⚠️ Most important — cross-file dependency checks
Example: If a function signature changed, were all its calls updated?
Provides full architectural context from .q.yaml files

5. AI Review and Report Generation:

AI response follows strict Markdown format:
- Assessment: [SUCCESS], [PARTIAL], or [FAILURE]
- Summary: Brief justification
- Suggestions: Actionable bullet points for next cycle

6. ReQap Output:

Writes AI-generated review to reqap_output_path
Includes fallback mechanism for AI failures

What’s configurable

config.yaml (under agents.inspeqtor):

provider (string): AI provider. Default: ‘openai’
model (string): AI model. Default: ‘gpt-4o’

config.yaml (under options):

use_qontextor (boolean): Enables rich context from .q.yaml files (recommended)

Environment Variables:

CYCLE_NUM: Used for logging and context within the reQap