feat(embedders): Gemini Embedding 2 multimodal — native PDF, image, audio, video#1627

Draft

24601 wants to merge 16 commits intoairweave-ai:mainfrom

24601:feat/gemini-embedding-2

Commits on Mar 18, 2026

feat(embedders): add Gemini Embedding 2 as 4th dense provider

24601
and
claude
committed
feat(embedders): add multimodal + audio/video embedding via Gemini Embedding 2

24601
and
claude
committed
fix(embedders): address Codex R2 blockers — OOM, sparse alignment, fallback seam

24601
and
claude
committed
fix(embedders): close remaining Codex R3 gaps — size gate, PDF catch, docs

24601
and
claude
committed
fix(chunkers): size-aware audio segment sizing for oversized short files

24601
and
claude
committed
fix(chunkers): add 5% safety margin for container overhead + regression test

24601
and
claude
committed
feat(converters): scene-based keyframe OCR for video text extraction

24601
and
claude
committed
docs(gemini-embedding-2): comprehensive update to reflect full implementation

24601
and
claude
committed
fix(tests): align media/video test assertions with actual settings defaults

24601
and
claude
committed
fix(embedders): address all cross-val findings — architecture, perf, docs

24601
and
claude
committed
test(embedders): comprehensive coverage for all cross-val fix paths

24601
and
claude
committed
fix(embedders): address cubic review — step clamp, text alignment, protocol

24601
and
claude
committed
fix(embedders): address Codex R2 + Gemini R2 findings

24601
and
claude
committed
fix(embedders): address Codex R3 — hang test, cache test, gate test, docs

24601
and
claude
committed
test(embedders): add multimodal hang-based timeout test (Codex R4)

24601
and
claude
committed
fix(embedders): address Codex R5 — entity preservation, range validation, docs

24601
and
claude
committed

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

feat(embedders): Gemini Embedding 2 multimodal — native PDF, image, audio, video#1627

feat(embedders): Gemini Embedding 2 multimodal — native PDF, image, audio, video#1627
24601 wants to merge 16 commits intoairweave-ai:mainfrom
24601:feat/gemini-embedding-2

Commits on Mar 18, 2026

feat(embedders): add Gemini Embedding 2 as 4th dense provider

feat(embedders): add multimodal + audio/video embedding via Gemini Embedding 2

fix(embedders): address Codex R2 blockers — OOM, sparse alignment, fallback seam

fix(embedders): close remaining Codex R3 gaps — size gate, PDF catch, docs

fix(chunkers): size-aware audio segment sizing for oversized short files

fix(chunkers): add 5% safety margin for container overhead + regression test

feat(converters): scene-based keyframe OCR for video text extraction

docs(gemini-embedding-2): comprehensive update to reflect full implementation

fix(tests): align media/video test assertions with actual settings defaults

fix(embedders): address all cross-val findings — architecture, perf, docs

test(embedders): comprehensive coverage for all cross-val fix paths

fix(embedders): address cubic review — step clamp, text alignment, protocol

fix(embedders): address Codex R2 + Gemini R2 findings

fix(embedders): address Codex R3 — hang test, cache test, gate test, docs

test(embedders): add multimodal hang-based timeout test (Codex R4)

fix(embedders): address Codex R5 — entity preservation, range validation, docs