Add 6 new endpoints, streaming usage, and rate limiting headers

[jpr5]

Summary

6 new endpoints:

• Gemini embedContent — Google embedding users were getting 404
• /v1/images/edit + /v1/images/variations — multipart, closes Support for /images/edit #221
• /v1/audio/translations — reuses transcription handler
• Ollama /api/embeddings — supports prompt and input (string or array)
• Cohere /v2/embed — multi-text, configurable embedding types
• ElevenLabs /v1/text-to-speech/{voice_id} — binary audio with voice routing

3 cross-cutting features:

• Streaming usage chunks (stream_options.include_usage)
• Automatic token usage estimation (~4 chars/token heuristic)
• Rate limiting headers on 429 responses (Retry-After, x-ratelimit-*)

Housekeeping:

• Version bump to 1.25.0
• CHANGELOG updated
• Docs, README, DRIFT.md, competitive matrix updated

Test plan

• 3088 tests pass (96 new across 11 test files)
• TypeScript clean
• Lint + format + build clean
• CR converged Round 2 (8 findings fixed, 7/7 confirmed)
• HTML validated
• Drift tests (gated behind API keys)

🤖 Generated with Claude Code

[pkg-pr-new]

Open in StackBlitz

npm i https://pkg.pr.new/@copilotkit/aimock@222

commit: 9a1f6c7