AI and LLM Usage in BlueOptima
🌐 Public Doc
Last updated on August 20, 2026 • Approximate read time: 8 minutes
This page is available without a BlueOptima account.
This page addresses common questions about how BlueOptima uses Artificial Intelligence (AI), Large Language Models (LLMs), and Generative AI (GenAI) across its products, including the Maintainability Sub-Agent and How To Fix 2.0 (HTF 2.0). It is designed to help teams, technical and non-technical, understand how AI is applied responsibly and securely.
BlueOptima processes only the minimum code content required to perform analysis or generate recommendations. AI models are configured strictly for inference, ensuring that customer data remains private, secure, and never reused for training.
📚 Related guides:
Requires a BlueOptima account.
Who Should Read This
- Legal, Security, and Procurement Teams reviewing data privacy, compliance, and vendor assurance information.
- Engineering Managers and Team Leads understanding how AI and LLMs are applied within BlueOptima's products.
- Executives and Senior Stakeholders evaluating AI-enabled capabilities for organizational use.
- Prospective Customers and Evaluators confirming how BlueOptima handles AI and code data before choosing a platform.
Why This Matters
Clients, particularly those in regulated industries, often need clear documentation to confirm that AI-powered tools are designed to operate within enterprise compliance expectations and do not introduce risks beyond standard BlueOptima platform operations. This page supports that assurance by outlining how BlueOptima's AI systems operate safely within enterprise standards.
📚 For configuration and deployment details, see Maintainability Sub-Agent and How To Fix 2.0. 🔒 Requires a BlueOptima account.
How BlueOptima Uses AI and LLMs
BlueOptima applies AI to assist with maintainability improvements by generating candidate fixes and structural enhancements. All AI-generated output is reviewed and applied manually by developers. Nothing is ever merged automatically.
- Purpose-built AI applications: AI is used to detect code maintainability issues and suggest improvements.
- Metadata-driven: AI models analyze contextual information derived from code, which may include limited code segments or file content when required for maintainability analysis or fix generation.
- Inference-only behavior: Models generate insights without any retraining from customer repositories.
- Human-in-the-loop: Developers review and approve all AI-generated recommendations.
AI features in BlueOptima are assistive, not autonomous. No code is ever automatically merged into production branches. The Maintainability Sub-Agent may generate candidate fixes, but developers always review and apply changes manually.
Data Privacy and Security
BlueOptima's AI and LLM services follow the same security and compliance standards as all platform components.
| Safeguard | Description |
|---|---|
| Data Scope | BlueOptima uses metadata whenever possible: for example, file names, diffs, and complexity metrics. For the Maintainability Sub-Agent and HTF 2.0, only the minimum required code content is processed to generate recommendations or replacement files. |
| Inference Mode | AI models are used only for inference. No data contributes to model training. |
| Secure Transport | All data in transit is encrypted using Transport Layer Security (TLS) 1.3 or higher. |
| Storage Policy | No customer source code is stored, cached, or retained post-inference. |
| Environment Control | All AI inference occurs within secured, isolated environments, configurable via the Integrator. |
The Maintainability Sub-Agent operates under SOC 2 (System and Organization Controls 2)-aligned controls, ensuring full auditability for enterprise security and compliance reviews.
Supported AI Models
Starting with Integrator 6.18.0, BlueOptima no longer ships or defaults to a hosted model for local inference. If you use local inference, you must host and manage your own OpenAI-compatible server.
The Maintainability Sub-Agent supports multiple AI providers to align with diverse enterprise security and compliance requirements.
| Provider | Deployment Type | Notes |
|---|---|---|
| AWS Bedrock | Cloud | Supports multiple foundation models via AWS API |
| OpenAI GPT | Cloud | Supports GPT-based inference through OpenAI API |
| Anthropic Claude | Cloud | Enterprise-grade privacy and safety guardrails |
| Google Gemini | Cloud | Supported through Integrator LLM configuration |
| Azure AI Foundry | Cloud | Supports Azure-hosted OpenAI and LLM models |
| BlueOptima LLM Server | On-premise | Hosts customer-configured models locally. Qwen was the default model for Integrator versions 6.8.0 through 6.17.0. As of Integrator 6.18.0, BlueOptima no longer ships or packages a Llama Server at all, and Auto-Fix no longer defaults to it. You host and manage your own OpenAI-compatible server. BlueOptima does not provide or default to a hosted model. |
Administrators can configure the Integrator to select or restrict model usage according to organizational data governance policies.
📚 Related: How To Fix 2.0: Setup. 🔒 Requires a BlueOptima account.
Model Performance
| Factor | Impact on Processing Time |
|---|---|
| File Size | Larger files take longer to analyze. |
| Model Type | Processing times vary by model, file size, and deployment mode. Cloud-hosted models generally respond faster than self-hosted models, but performance depends on configuration. |
| Deployment Mode | On-premise models may experience higher latency due to hardware performance. |
Model selection allows organizations to balance processing speed, cost, and compliance requirements.
📚 For setup and model configuration, see:
- Maintainability Sub-Agent and How To Fix 2.0
- Maintainability Sub-Agent: Frequently Asked Questions
- How To Fix 2.0: Frequently Asked Questions
Requires a BlueOptima account.
Frequently Asked Questions
Is My Source Code Used to Train or Improve Any Model?
No. BlueOptima does not use your source code to train or update any models. Because HTF 2.0 requires organizations to provide their own Large Language Model (LLM) connection, your data remains fully under your control. If you use an internally hosted server, your code never leaves your infrastructure. If you use a cloud-based LLM, your organization's agreement with that cloud provider strictly governs data usage.
Your source code and metadata are never used to retrain, fine-tune, or update any LLM.
Does BlueOptima Use Our AI Model or Its Own?
BlueOptima connects to the AI model your organization configures. There is no default model from Integrator version 6.18.0 and onwards. If you use local inference, you must host and manage your own OpenAI-compatible server. See the Supported AI Models table for details.
Which AI Models Are Supported?
The Maintainability Sub-Agent supports AWS Bedrock, OpenAI GPT, Anthropic Claude, Google Gemini, and Azure AI Foundry as cloud providers. For local inference, you must host and manage your own OpenAI-compatible server. There is no default model.
Does My Code Leave Our Environment?
For self-hosted setups, inference runs inside your environment. For cloud models, only the minimum necessary code context or metadata is sent over encrypted channels for inference.
Can We Choose or Restrict the Model?
Yes. Administrators select the preferred provider in the Integrator configuration. If external LLMs are not permitted, host and manage your own OpenAI-compatible server for local inference.
Why Do Processing Times Differ by Model?
Cloud models such as Bedrock, GPT, and Claude typically respond faster. Self-hosted models may take longer depending on local resources.
Key Takeaways
- BlueOptima uses AI responsibly, transparently, and securely.
- Customer data is never used to train or modify AI models.
- The Maintainability Sub-Agent operates entirely in inference mode.
- All AI activities occur within encrypted, controlled environments.
- The approach meets enterprise legal and compliance expectations.
📚 See: Code Author Detection. 🔒 Requires a BlueOptima account.