2026-04-09 Thu

LangChain/Ecosystem
- It kind of seems like LangSmith/Deployment should have asset storage.
  - Discovered LangSmith/Storage/Blob/S3
    - LangChain Support Portal - How do I resolve S3 PutObject access denied errors when enabling LangSmith blob storage with KMS encryption requirements?
      How do I resolve S3 PutObject access denied errors when enabling LangSmith blob storage with KMS encryption requirements?
      
      Context
      
      When enabling LangSmith blob storage with S3, you may encounter access denied errors if your S3 bucket has a resource-based policy that requires KMS encryption headers. The error typically appears as:
      
      AccessDenied: User: arn:aws:sts::account:assumed-role/langsmith-platform-backend-blob-storage-v2-environment/session is not authorized to perform: s3:PutObject on resource with an explicit deny in a resource-based policy
      
      This occurs when your bucket policy includes aDenyIncorrectEncryptionHeaderstatement that blocks PutObject requests without proper KMS encryption headers.
      
      Answer
      
      LangSmith now supports KMS encryption for S3 blob storage. To resolve this issue, configure KMS encryption in your LangSmith deployment:
      Add the KMS encryption configuration to your LangSmith config:
      config: blobStorage: enabled: true engine: "S3" chSearchEnabled: false bucketName: "your-bucket-name" apiURL: "https://s3.your-region.amazonaws.com" kmsEncryptionEnabled: true kmsKeyArn: "arn:aws:kms:your-region:account-id:key/your-kms-key-id"
      Ensure your IAM roles have the necessary KMS permissions in addition to S3 permissions:
      kms:Encrypt
      kms:GenerateDataKey
      kms:Decrypt
      Update your service account annotations with the appropriate IAM role ARNs for backend, platformBackend, and queue components.
      Deploy the updated configuration and verify that objects are being created in your S3 bucket with KMS encryption.
      
      With KMS encryption enabled, LangSmith will include the required encryption headers in S3 requests, satisfying bucket policies that enforce encryption requirements. You should see objects appearing in your S3 bucket once the configuration is properly applied.
    - Actually, it looks like this is not for that - it's for OLAP over traces, runs, etc
PL stuff
- thinking about trying a Commitment to primarily or only writing in Rust for a year, as a way to Concentrate the Attention
  - of course, this wouldn't be strictly possible, but perhaps there's a version of this challenge which could be articulated and followed
sec
- Frustrations with 1Password
  - learned that 1Password/Dev/ssh/agent has a setting where one can extend the timeout. I was getting fingerprinted every single time my computer unlocked, which was getting super annoying, especially considering that every time I unlocked the computer, 1Password ended up not detecting that the Mac/TouchID was available for ~1min. the result was that I had a foreground dialog on top of the focused application that I had to move out of the way until touchid was considered available by 1Password
en garde(ning)
- rulesync/Skill for proxying pages from other logseq gardens
  - Logseq/Idea/Proxy
    - I could have sworn that I had some context in this garden related to Person/codekiln/GitHub/logseq-gardener
- getting carded
  - Did some flashcards this morning. The Logseq/Flashcard/Review UI (the one in the side-panel) is a bit confusing. It has a drop down menu to select different categories of flashcards, which in theory is convenient. Currently, in this garden, though, many items appear more than once, and some of the items that appear in the menu don't have any flashcards. In addition, right now I'm not prioritizing VSCode flashcards anymore (though I do still need a few; for example, being able to see the log after rebuilding devcontainers is a good one).
  - I should probably have a better precedent for how to log Logseq/Flashcards that are not Keyshorts; I don't want to stuff CLI commands and GUI/Menu/Commands into keyshorts if it isn't actually a Keyboard Shortcut. It might make sense to somehow tie this to my nascent Logseq/Entity system.
- on my burgeoning Logseq/Entity system
  - It's been bugging me that the current Logseq/Frontmatter/Key for it, logseq-entity, implies a singular entity in a page, when really, a page may need to document more than one entity.
    - for example, something can be a term, and also a concept, and also a software-project.
  - This makes me wonder about how to more clearly identify what I'm saying a logseq entity is.
    - Taxonomy?
      - Is it a page that is intended to have certain structural elements?
        One of the motivating factors was to provide AI Knowledge Gardening with AI/Context about how to import an entity like a book or a software project. If it's the case that I'm thinking of an entity as being more like a "page template" or a "page schema" - or, and let's be more accurate here, is it more like structural typing in programming languages, except instead of describing methods and properties, the entities describe page frontmatter and headings
        A key aspect of how I use logseq is that I create a page so that I can refer to something by name.
        Usually I do this with a logseq tags:: [[Term]] frontmatter entry. Usually I place the term at a logical location in the namespaces for where I want it to be sorted on disk, lexicographically, so that when I look at the files in the Logseq/Page pages directory, it works a bit like Library/Science/Classification/Dewey or Library/Science/Classification/Library of Congress
      - Is it a type of Digital Twin
gard(en)diff

3 Linked References

Unlinked References