Release v1.0.0

Added

• Token usage extraction — input_tokens, output_tokens, cache_read_tokens, and cache_write_tokens captured from every LLM response and stored in ClickHouse; supports Anthropic (including cache tokens), OpenAI, and SAP AI Core in both streaming and non-streaming modes
• OTel metric counters — gen_ai.usage.input_tokens and gen_ai.usage.output_tokens emitted as Int64Counter metrics with gen_ai.system, llm.agent, gen_ai.response.model, and audit.project dimensions; no-op when OTEL_EXPORTER_OTLP_ENDPOINT is unset
• OTel MeterProvider — OTLP/HTTP metric exporter set up alongside existing TracerProvider; both shut down cleanly on graceful stop
• TokenUsage shared type in parser/types.go — used by all parser result types
• Token usage span attributes — gen_ai.usage.input_tokens and gen_ai.usage.output_tokens added to llm.proxy.request spans
• Migration 007_add_token_usage.sql — adds input_tokens, output_tokens, cache_read_tokens, cache_write_tokens columns to existing installations
• Dashboard design documentation (docs/dashboard/) — comprehensive design spec and implementation guide for a Nuxt 4 + Vue 3 web UI covering session browsing, event inspection, token usage analytics, and proxy health monitoring
• Parser tests for token usage extraction (Anthropic full + streaming, OpenAI full + streaming)
• Handler test for end-to-end token usage flow

Changed

• Release pipeline now extracts release notes from CHANGELOG.md instead of auto-generating from git log; falls back to git log if no entry exists for the tagged version