v0.1.26

What's Changed

• Fix batch status compatibility for LiteLLM by @deltawi in #154
• Add OpenAI-compatible Batch response fields by @deltawi in #157
• Fix MCP streamable HTTP sessions by @deltawi in #160
• Split batch workers into a dedicated Helm deployment by @deltawi in #162
• Batch scheduler phase 1: instrumentation and normalization by @deltawi in #169
• Batch scheduler phase 2: work-slice claims by @deltawi in #170
• Add model capacity-aware batch scheduling by @deltawi in #171
• Implement batch tenant fair-share scheduling by @deltawi in #172
• Implement batch size and aging scheduling by @deltawi in #173
• Batch scheduler phase 6 rollout operationalization by @deltawi in #177
• Add batch item claim fencing by @deltawi in #178
• Batch scheduler: harden fair-share lock scalability by @deltawi in #179
• Batch scheduler: add stale lease sweeper and convergence metrics by @deltawi in #180
• Document batch production setup by @deltawi in #181
• Harden release image frontend install by @deltawi in #182
• Add notification dispatcher with Slack budget alerts by @deltawi in #183

Full Changelog: v0.1.24...v0.1.26