feat(knowledge): add embedding model selection and Cohere reranker#4349

Merged

waleedlatif1 merged 22 commits into

stagingfrom

waleedlatif1/embedding-model-research

Apr 30, 2026

Commits on Apr 30, 2026

feat(knowledge): add embedding model selection and Cohere reranker
waleedlatif1
committed
fix(knowledge): split reranker model constants into client-safe module
waleedlatif1
committed
fix(knowledge): bill rerank on every successful API call and fix MDX docs literal
waleedlatif1
committed
test(knowledge): align embedding tests with provider abstraction changes

waleedlatif1
and
claude
committed
fix(knowledge): require explicit Azure deployment per OpenAI embedding model

waleedlatif1
and
claude
committed
fix(knowledge): skip platform reranker billing for BYOK Cohere keys

waleedlatif1
and
claude
committed
fix(knowledge): match search tokenizer to embedding provider; remove dead var

waleedlatif1
and
claude
committed
fix(knowledge): match chunk tokenizer to KB embedding provider

waleedlatif1
and
claude
committed
refactor(knowledge): centralize tokenizer mapping on EmbeddingModelInfo

waleedlatif1
and
claude
committed
refactor(knowledge): lock embedding model to KB_EMBEDDING_MODEL env var

waleedlatif1
and
claude
committed
fix(knowledge): use provider tokenizer for chunks and bound rerank indices

waleedlatif1
and
claude
committed
fix(knowledge): use .count from estimateTokenCount return value

waleedlatif1
and
claude
committed
fix(knowledge): only enforce single embedding model when query is present

waleedlatif1
and
claude
committed
fix(knowledge): use getConfiguredEmbeddingModel in copilot KB creation

waleedlatif1
and
claude
committed
fix(knowledge): make EMBEDDING_DIMENSIONS a literal type

waleedlatif1
and
claude
committed
fix(knowledge): use per-KB embedding model in v1 search route

waleedlatif1
and
claude
committed
chore(knowledge): polish embedding/reranker implementation

waleedlatif1
and
claude
committed
fix(knowledge): resolve type errors and unhandled rejection in search routes

waleedlatif1
and
claude
committed
fix(knowledge): pass Gemini API key via x-goog-api-key header

waleedlatif1
and
claude
committed
fix(knowledge): default Azure deployment name to embedding model name

waleedlatif1
and
claude
committed
fix(knowledge): cap Gemini batches at 100 items, add singular GEMINI_API_KEY fallback

waleedlatif1
and
claude
committed
fix(knowledge): prefer singular Cohere key before rotation

waleedlatif1
and
claude
committed

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

feat(knowledge): add embedding model selection and Cohere reranker#4349

feat(knowledge): add embedding model selection and Cohere reranker#4349
waleedlatif1 merged 22 commits into
stagingfrom
waleedlatif1/embedding-model-research

Commits on Apr 30, 2026

feat(knowledge): add embedding model selection and Cohere reranker

fix(knowledge): split reranker model constants into client-safe module

fix(knowledge): bill rerank on every successful API call and fix MDX docs literal

test(knowledge): align embedding tests with provider abstraction changes

fix(knowledge): require explicit Azure deployment per OpenAI embedding model

fix(knowledge): skip platform reranker billing for BYOK Cohere keys

fix(knowledge): match search tokenizer to embedding provider; remove dead var

fix(knowledge): match chunk tokenizer to KB embedding provider

refactor(knowledge): centralize tokenizer mapping on EmbeddingModelInfo

refactor(knowledge): lock embedding model to KB_EMBEDDING_MODEL env var

fix(knowledge): use provider tokenizer for chunks and bound rerank indices

fix(knowledge): use .count from estimateTokenCount return value

fix(knowledge): only enforce single embedding model when query is present

fix(knowledge): use getConfiguredEmbeddingModel in copilot KB creation

fix(knowledge): make EMBEDDING_DIMENSIONS a literal type

fix(knowledge): use per-KB embedding model in v1 search route

chore(knowledge): polish embedding/reranker implementation

fix(knowledge): resolve type errors and unhandled rejection in search routes

fix(knowledge): pass Gemini API key via x-goog-api-key header

fix(knowledge): default Azure deployment name to embedding model name

fix(knowledge): cap Gemini batches at 100 items, add singular GEMINI_API_KEY fallback

fix(knowledge): prefer singular Cohere key before rotation