Bias Detection Analyzer — Full UI and Functionality Launch

September 24, 2025

This release brings the Bias Detection Analyzer from prototype to fully functional. The entire interface has been rebuilt with a modern design, the underlying connections to AI providers have been repaired, and several usability features have been added including a real-time progress bar and one-click prompt copying. Analyzers are also now free for all authenticated paid users — no subscription tier upgrade required.

New Features

• Bias Detection Analyzer — A complete prompt analysis tool that evaluates your prompts for potential bias. Submit any prompt, choose your AI provider and model, and receive a detailed analysis of bias patterns, blind spots, and suggestions for more balanced prompt construction.
• Real-time progress bar — When running an analysis, a visual progress bar shows elapsed time with a percentage indicator and countdown estimate. The bar tracks a 60-second window and intelligently jumps to completion when the analysis finishes early, so you always know where things stand.
• Copy to clipboard — A copy button on the prompt input panel lets you quickly re-copy your prompt text after analysis. Visual feedback confirms the copy with a checkmark icon, and toast notifications confirm success or report errors.
• Two-column analysis layout — The analyzer uses a clean two-column design: your prompt input on the left, analysis results on the right, with provider and model selection spanning the full width above. This gives you a clear side-by-side view of what you submitted and what the analysis found.

Improvements

• Free access for all authenticated paid users — Analyzers no longer require an upgrade to your subscription plan. Any authenticated user can run prompt analyses, removing a barrier that was limiting access to these tools.
• Modern interface design — The analyzer pages have been redesigned with a cohesive visual theme, glass-effect card layouts, gradient text, hover animations, and polished button styling. The design is consistent with the rest of the EnginifyAI interface.
• Provider card layout — The provider selector now displays all four providers (OpenAI, Anthropic, Google, xAI) in a single row with connected-status badges aligned inline with provider names, making it faster to scan and select.
• Streamlined prompt input — The prompt input area has been simplified to a direct textarea, removing unnecessary tab navigation and focusing the workflow on the core task: paste your prompt, run the analysis.

Bug Fixes

• Fixed analyzer not found errors — Resolved an issue where navigating to an analyzer page would return a “not found” error due to incorrect URL parameter extraction. Analyzers now load reliably from their detail pages.
• Fixed API key decryption — The analyzer’s connection to AI providers was failing because API keys weren’t being decrypted correctly. The decryption pipeline has been repaired so that stored provider keys are properly retrieved when running analyses.
• Fixed model verification — The system was checking the wrong database location when verifying which AI models are available, causing valid models to appear unavailable. Model verification now checks the correct sources.
• Fixed instruction file loading — Analyzer instruction files weren’t loading due to a naming convention mismatch between the database and the file system. A path conversion layer now ensures instructions are found and loaded correctly.
• Fixed API field naming — The analysis request was sending provider and model identifiers with incorrect field names, causing the API to reject valid requests. Field names now match what the backend expects.