v0.26.5

What's Changed

• docs: open-core hardening roadmap — to mainstream-ready + 1.0 by @thankrandomness in #157
• docs: marathon learnings — agentic playbook, methodology, case studies by @thankrandomness in #164
• docs: FIELD-NOTES — domain lessons as predict-first anomaly cards by @thankrandomness in #165
• docs(trackers): reconcile remote-ml-surface + roadmap to 0.26.4 by @thankrandomness in #166
• fix(py-client): pin the stub generator and couple the runtime floors to it by @thankrandomness in #168
• feat(client): put the bulk inference verb on RemoteDatabase by @thankrandomness in #169
• feat(client): put the eval verb family on RemoteDatabase by @thankrandomness in #171
• fix(jammi-db): tenant-qualify the evidence-channel catalog (D1) by @thankrandomness in #170
• feat(client): put the channel verb family on RemoteDatabase by @thankrandomness in #172
• fix(eval): order-invariant bootstrap CI + f32 confidence parity contract by @thankrandomness in #173
• release: v0.26.5 — remote eval/infer/channel parity + the eval-report bugs the cookbook surfaced by @thankrandomness in #174

Full Changelog: v0.26.4...v0.26.5