Competitive feature parity: embeddings, structured output, sequential responses, Docker

[jpr5]

Summary

Adds four high-value features that fill real gaps relative to VidaiMock, dwmkerr/mock-llm, and piyook/llm-mock:

• Embeddings endpoint (POST /v1/embeddings) — deterministic embedding generation via SHA-256 hash, fixture matching on inputText, onEmbedding() convenience, automatic fallback to deterministic embeddings when no fixture matches
• Structured output / JSON mode — responseFormat matching on response_format.type, onJsonOutput() convenience, validation that JSON-mode fixtures contain valid JSON
• Sequential/stateful responses — sequenceIndex on fixtures enables multi-turn sequences (first call → tool call, second call → text), per-fixture match counting in Journal with sibling propagation
• Docker + Helm — multi-stage Dockerfile (node:22-alpine), GHCR publish workflow (linux/amd64+arm64), basic Helm chart with deployment/service templates

Also includes:

• Documentation website — 14 static HTML pages in docs/ covering all features with unit test and integration test examples
• Updated competitive matrix — added VidaiMock, mock-llm, piyook/llm-mock as columns + 7 new feature rows
• Embeddings drift test — three-way SDK/real/mock shape comparison for OpenAI embeddings
• 109 new tests (658 total, up from 549)

Test Plan

• pnpm test — 658 tests passing
• pnpm run build — clean
• pnpm run format:check — clean
• pnpm run lint — clean
• Full 6-agent code review loop — 2 rounds, zero actionable feedback in Round 2
• pnpm test:drift — drift tests (requires API keys)
• docker build -t llmock . — Docker build verification

[pkg-pr-new]

Open in StackBlitz

npm i https://pkg.pr.new/CopilotKit/llmock/@copilotkit/llmock@42

commit: c1bb46e