Tirith adds AI API observability

This is a pragmatic answer to a real pain point: teams want spend visibility without sending prompts to a SaaS observability layer. The product is still early, but the local-first design and one-import integration make it easy to see why devs would try it.

• –The main pitch is low-friction cost tracking: one import or proxy config, then usage metadata flows into a local dashboard
• –Privacy is a real differentiator here because the README says prompts and responses are never logged, only metadata
• –The tagging model is useful for attributing spend to features, users, sessions, or environments instead of just showing a bill total
• –It sits in the same category as LLM observability tools like Helicone, but with a more lightweight, self-hosted posture
• –The current scope is narrower than the headline implies: Anthropic is the clear first-class path, while OpenAI and Google are still marked planned