fix(backends): support OpenAI response format

[drewstone]

Summary

• add responseFormat to createOpenAICompatibleBackend and forward it as OpenAI response_format
• export the new OpenAIChatResponseFormat type
• bump @tangle-network/agent-runtime to 0.75.1 and regenerate API docs

Verification

• pnpm exec vitest run tests/backends-openai-tools.test.ts tests/backends-openai-usage.test.ts tests/backends-fail-loud.test.ts tests/runtime.test.ts — 52 passed
• pnpm run typecheck — passed
• pnpm run lint — passed
• pnpm run docs:check — passed
• pnpm run build — passed
• pnpm run verify:package — passed
• env -u TANGLE_INTELLIGENCE_ENDPOINT -u TANGLE_INTELLIGENCE_URL -u TANGLE_INTELLIGENCE_OTLP_BASE -u TANGLE_OBSERVE_ENDPOINT -u OTEL_EXPORTER_OTLP_ENDPOINT -u OTEL_EXPORTER_OTLP_TRACES_ENDPOINT pnpm test — 115 files, 1114 passed, 1 skipped

Notes

This is additive. It lets consumers that need JSON-mode router calls use the shared runtime backend instead of direct /chat/completions calls.

[tangletools]

✅ Auto-approved PR — 59c866ac

Blanket team auto-approval is enabled for this reviewer service.
The full PR reviewer audit still runs separately and will publish findings if it detects issues.

_{tangletools · auto-approval · reason: blanket_auto_approve · 2026-06-23T00:24:57Z}