v0.16.0 - Native MMR Search Reranking! 🔎⚡️

Native, SIMD-accelerated MMR (Maximal Marginal Relevance) Reranking!

• Rerank Memory and Document searches using MMR
• MMR is hardware SIMD-accelerated on amd64 CPUs (Intel and AMD)

What's Changed

• feat: MMR Reranking for Document and Memory Search by @danielchalef in #232
• fix: mmr search_type json naming; use min limit for mmr by @danielchalef in #234
• fix: upgrade go toolchain 1.21.3 by @danielchalef in #237

Full Changelog: v0.15.2...v0.16.0

v0.15.2 - Bugfix: Don't fail on pgvector upgrade error

What's Changed

• fix: don't fail on pgvector extension upgrade error by @danielchalef in #231

Full Changelog: v0.15.1...v0.15.2

v0.15.1 - New K8s deployment; Bump version to fix Container builds

• Fix zep-cloud container builds.
• zep-cloud arm64 container
• new K8s deployment

What's Changed

• fix: k8 deployment; arm64 cloud build by @danielchalef in #227

Full Changelog: v0.15.0...v0.15.1

v0.15.0 - Right To Be Forgotten Support

Zep now supports executing Right To Be Forgotten requests against Zep's MemoryStore ensuring compliance with regulation such as the CCPA and GDPR.

When using Zep's UserStore functionality, a user's data may be scheduled for deletion. All known user data is marked for deletion: messages, summaries, message embeddings. Records marked for deletion are expunged when the next Purge batch run executes.

What's Changed

• Bump golang.org/x/net from 0.15.0 to 0.17.0 by @danielchalef in #224
• feat: Delete All User Data (RTBF) by @danielchalef in #225

Full Changelog: v0.14.0...v0.15.0

v0.14.0 - Improvements & Fixes, incl. MaxRequestSize and Go upgrade

What's Changed

• improv: MaxRequestSize; Additional Document Collection tests by @danielchalef in #219
• improv: Custom Model Names for Azure OpenAI APIs by @danielchalef in #220
• fix: model name validation when using custom model by @danielchalef in #214
• fix: refactor parseJSONQuery variants to single function by @danielchalef in #215
• Upgrade: Go compiler, toolchain 1.21.2 by @danielchalef in #217
• fix: parseJSONQuery for doc search; document search route test by @danielchalef in #221

Full Changelog: v0.13.0...v0.14.0

v0.13.0 - HNSW Indexing! ⚡️ Custom Prompts! 😀

What's Changed

• feat: HNSW indexing; refactor doc embeddings by @danielchalef in #209
• feat: Configurable Summarizer Prompt by @bricemacias in #211
• feat: darkmode toggle by @danielchalef in #210
• fix: Don't retry OpenAI calls on 400 errors; Better Error Messages by @danielchalef in #213

HNSW index support

Requires pgvector v0.5 or later to be installed in order to create HNSW indexes.

• Message and Document embedding columns default to using an HNSW index
• Manual Collection indexing is a noop if this is the case

Full Changelog: v0.12.4...v0.13.0

v0.13.0-beta.0: HNSW indexing; Configurable Summarizer Prompt

What's Changed

• feat: HNSW indexing; refactor doc embeddings by @danielchalef in #209
• feat: Configurable Summarizer Prompt by @bricemacias in #211
• feat: darkmode toggle by @danielchalef in #210

HNSW index support

Requires pgvector v0.5 or later to be installed in order to create HNSW indexes.

• Message and Document embedding columns default to using an HNSW index
• Manual Collection indexing is a noop if this is the case

Also includes:

• A refactoring of Collection embedding functionality

Installation or Upgrading

Ensure that your Postgres provider has pgvector 0.5 installed -or- that you're using the latest Zep Postgres container.

Modify the docker-compose.yaml file (or your Render or other hosting config) to include the correct Zep container build tags:

  zep:
    image: ghcr.io/getzep/zep:0.13.0-beta.0

-or for cloud deployments-

  zep:
    image: ghcr.io/getzep/zep-cloud:0.13.0-beta.0

and then run a pull -or- run a redeploy in your cloud environment.

docker compose pull

Your Zep log file should contain an INFO level entry as follows:

vector extension version is >= 0.5.0. hnsw indexing available

IMPORTANT: Pre-release software. May contain bugs!

This is an early release and may container bugs! Please back up your database prior to upgrading.

Full Changelog: v0.12.4...v0.13.0-beta.0

v0.13.0-alpha.0 - HNSW Support 🔎; improvements & bugfixes

What's new

HNSW index support

Requires pgvector v0.5 or later to be installed in order to create HNSW indexes.

• Message and Document embedding columns default to using an HNSW index
• Manual Collection indexing is a noop if this is the case

Also includes:

• A refactoring of Collection embedding functionality

Installation or Upgrading

Ensure that your Postgres provider has pgvector 0.5 installed -or- that you're using the latest Zep Postgres container.

Modify the docker-compose.yaml file (or your Render or other hosting config) to include the correct Zep container build tags:

  zep:
    image: ghcr.io/getzep/zep:0.13.0-alpha.0

-or for cloud deployments-

  zep:
    image: ghcr.io/getzep/zep-cloud:0.13.0-alpha.0

and then run a pull -or- run a redeploy in your cloud environment.

docker compose pull

Your Zep log file should contain an INFO level entry as follows:

vector extension version is >= 0.5.0. hnsw indexing available

IMPORTANT: Pre-release software. May contain bugs!

This is an early release and may container bugs! Please back up your database prior to upgrading.

Full Changelog: v0.12.4...v0.13.0-alpha.0

v0.12.4 - Postgres 15.4; pgvector 0.5; Bugfixes

Note: this Postgres and pgvector upgrade will only affect docker compose installs. If you're deploying to a cloud provider, you're likely using your provider's Postgres service.

What's Changed

• fix: DB ReadTimeout; Index Creation by @danielchalef in #203
• feat: upgrade pg154; pgvector 0.5 by @danielchalef in #204

Full Changelog: v0.12.3...v0.12.4

v0.12.3 - Bugfixes and Improvements 🐞🐛🪲

What's Changed

• fix: dark mode; page widths by @danielchalef in #202

Full Changelog: v0.12.2...v0.12.3