[P2] Add Configurable LLM Privacy Controls

[barvhaim]

Roadmap Item 15: Add Configurable LLM Privacy Controls

Priority: P2
Area: LLM, privacy, enterprise readiness
References: src/threatxtension/llm/, src/threatxtension/api/main.py

The project analyzes potentially sensitive extension code and sends selected content to
LLM providers. Users need clear controls over what leaves the machine.

Scope:

• Add configuration for local-only mode, metadata-only prompts, and max code context.
• Redact secrets, tokens, and obvious credentials before LLM calls.
• Show which provider/model generated each AI result.
• Document privacy tradeoffs for each provider.

Acceptance criteria:

• Users can disable all outbound LLM calls while keeping deterministic analyzers.
• Prompt payload sizes and redaction counts are logged without leaking content.
• API and CLI clearly indicate when AI analysis is disabled or redacted.

---

Created from docs/roadmap.md.