ai-prod-demo
AI overview
Sign in and the AI will write an overview from our coverage.
Headlines · 2
- Turn Token Logs into Per-Request AI Cost Metrics
掘金 · 人工智能 presents an open-source cost_model.py script that combines structured token logs with a configurable model price table to calculate total, per-request, and per-model costs. The tutorial also covers budget limits, CI-compatible exit codes, model- and feature-level attribution, and warnings for models left unpriced. Its sample output uses synthetic logs and a configurable price table, while a separate real-link test reports roughly 0.001 yuan per request with DeepSeek-tier models.
掘金 · 人工智能 · 🔥 0 - Fallback Chain Preserves Basic AI Service Availability During Model Failures
A Juejin tutorial presents a production fallback chain that moves from the primary chat-strong model to the backup chat-cheap model and then to a fixed rule-based response. It also recommends configuration-driven feature flags for enabling or disabling the backup model and rule fallback. The companion ai-prod-demo project includes six author-described unit cases covering successful responses, backup fallback, rule fallback, and disabled paths, along with an outage drill. The article recommends degraded-status markers, fallback logs, and alerts because rule-based responses can otherwise conceal model failures.
掘金 · 人工智能 · 🔥 0
Experience and discussion from the community
Share my ai-prod-demo experienceAsk about ai-prod-demo
Nobody has shared their experience with ai-prod-demo yet.