v0.3.0

## What's Changed
* feat: configurable embedding dimensions + OpenAI Embeddings by @nicoloboschi in https://github.com/vectorize-io/hindsight/pull/101
* fix: groq llm with free tier doesn't work by @nicoloboschi in https://github.com/vectorize-io/hindsight/pull/102
* Fix Python SDK not sending Authorization header by @cdbartholomew in https://github.com/vectorize-io/hindsight/pull/106
* Load .env file automatically on startup by @cdbartholomew in https://github.com/vectorize-io/hindsight/pull/104
* fix: make retain max completion tokens configurable by @nicoloboschi in https://github.com/vectorize-io/hindsight/pull/109
* fix: improve causal links detection by @nicoloboschi in https://github.com/vectorize-io/hindsight/pull/111
* ci: pin rust lock version by @nicoloboschi in https://github.com/vectorize-io/hindsight/pull/112
* feat: backup/restore by @nicoloboschi in https://github.com/vectorize-io/hindsight/pull/110
* fix(security): fix qs - CVE-2025-15284 by @nicoloboschi in https://github.com/vectorize-io/hindsight/pull/113
* feat: run db migrations offline (optionally) by @nicoloboschi in https://github.com/vectorize-io/hindsight/pull/114
* feat(helm): add existingSecret support by @nuclon in https://github.com/vectorize-io/hindsight/pull/119
* feat: Add per-request LLM token usage metrics by @omototo in https://github.com/vectorize-io/hindsight/pull/117
* feat(mcp): add async_processing parameter to retain tool by @slayoffer in https://github.com/vectorize-io/hindsight/pull/95
* fix: ui shows only 1000 memories by @nicoloboschi in https://github.com/vectorize-io/hindsight/pull/121
* feat: add metrics for llm call latency by @nicoloboschi in https://github.com/vectorize-io/hindsight/pull/120
* feat: support cohere as embeddings and reranker by @nicoloboschi in https://github.com/vectorize-io/hindsight/pull/122
* feat: support for multilingual content by @nicoloboschi in https://github.com/vectorize-io/hindsight/pull/124
* fix(mcp): add back bank list and create_bank tools by @nicoloboschi in https://github.com/vectorize-io/hindsight/pull/123
* feat: support different provider/models per operation by @nicoloboschi in https://github.com/vectorize-io/hindsight/pull/125
* fix: duplicated causal relationships and token optimization by @nicoloboschi in https://github.com/vectorize-io/hindsight/pull/126
* feat: add configs for database connection by @nicoloboschi in https://github.com/vectorize-io/hindsight/pull/128
* feat: delete memory bank by @dcbouius in https://github.com/vectorize-io/hindsight/pull/127
* feat: add operation_id to retain response by @nicoloboschi in https://github.com/vectorize-io/hindsight/pull/129
* chore: add flag to not include ml libs in docker image by @nicoloboschi in https://github.com/vectorize-io/hindsight/pull/130
* ci: fix flak tests by @nicoloboschi in https://github.com/vectorize-io/hindsight/pull/131
* Fix stats endpoint missing tenant authentication by @cdbartholomew in https://github.com/vectorize-io/hindsight/pull/134
* Fix embedding dimension for tenant schemas by @cdbartholomew in https://github.com/vectorize-io/hindsight/pull/135
* fix: misc perf improvements by @nicoloboschi in https://github.com/vectorize-io/hindsight/pull/133
* feat: retain modes by @nicoloboschi in https://github.com/vectorize-io/hindsight/pull/136
* fix: improve tei client parameters by @nicoloboschi in https://github.com/vectorize-io/hindsight/pull/137
* fix(typescript-client): Add error handling to all API methods by @nicoloboschi in https://github.com/vectorize-io/hindsight/pull/139
* misc: performance improvements by @nicoloboschi in https://github.com/vectorize-io/hindsight/pull/140
* ci: frozen uv sync by @nicoloboschi in https://github.com/vectorize-io/hindsight/pull/138
* fix: improve graph retrieval on large memory banks by @nicoloboschi in https://github.com/vectorize-io/hindsight/pull/141
* fix: entities list only show 100 entities by @nicoloboschi in https://github.com/vectorize-io/hindsight/pull/142
* fix: improve mpfp retrieval by @nicoloboschi in https://github.com/vectorize-io/hindsight/pull/146
* fix: batch queries on recall by @nicoloboschi in https://github.com/vectorize-io/hindsight/pull/149
* feat: support custom url for openai embeddings & cohere by @nicoloboschi in https://github.com/vectorize-io/hindsight/pull/150
* feat: add tenant to metrics labels by @nicoloboschi in https://github.com/vectorize-io/hindsight/pull/151
* feat: support litellm gateway by @nicoloboschi in https://github.com/vectorize-io/hindsight/pull/154
* Fix: Load extensions in server.py for multi-worker deployments by @cdbartholomew in https://github.com/vectorize-io/hindsight/pull/155
* feat: add memory tags by @nicoloboschi in https://github.com/vectorize-io/hindsight/pull/152

## New Contributors
* @nuclon made their first contribution in https://github.com/vectorize-io/hindsight/pull/119
* @omototo made their first contribution in https://github.com/vectorize-io/hindsight/pull/117
* @slayoffer made their first contribution in https://github.com/vectorize-io/hindsight/pull/95

**Full Changelog**: https://github.com/vectorize-io/hindsight/compare/v0.2.1...v0.3.0