Live Ingestion Engine

Feature Flag Management & Percentage Rollouts

Toggle production LLM routing heuristics, speculative decoding pipelines, and tenant-level rule targeting.

Total Configured Flags

3 flags

Active in Production

2 active

Gradual Canary Rollouts

1 partial

Feature Flag Key	Description	Environment	Rollout %	Targeting Rules	Toggle Status
ENABLE_LLM_PROVIDER_FALLBACK Dynamic Claude to GPT-4o Automatic Fallback Mesh	Enables automatic rerouting upon Anthropic or OpenAI 5xx / 429 rate limit errors.	production	100%	3 rules
STRICT_PII_ZERO_LOG_MASKING Strict Zero-Knowledge PII Sanitization Engine	Redacts Social Security Numbers, Credit Cards, and API keys before persisting spans.	production	100%	5 rules
SPECULATIVE_DECODING_ACCELERATOR Speculative Token Decoding via Small Draft Model	Uses Llama-3-8B draft model to achieve 2.4x speedup on 70B completions.	staging	25%	1 rules