Skip to content

Add AI SDK 6 memory tools and zero-config Cloud client - #79

Closed
cuevaio wants to merge 1 commit into
mainfrom
sync/monorepo-e390e49
Closed

cuevaio wants to merge 1 commit into
mainfrom
sync/monorepo-e390e49

Conversation

@cuevaio

@cuevaio cuevaio commented Oct 6, 2026

Copy link
Copy Markdown

Add user-scoped memory_search and memory_ingest tools for AI SDK 6 through @atomicmemory/vercel-ai-sdk/tools. Rename the Vercel adapter to @atomicmemory/vercel-ai-sdk and support new MemoryClient() with Cloud defaults and the server's ATOMICMEMORY_API_KEY.

The application binds user identity outside model arguments. Tools validate arguments, bound search output while retaining retrieval evidence, sanitize failures, forward cancellation, and reject pending HTTP 202 ingest responses. The framework-free root entry keeps AI SDK and Zod optional.

Include a runnable workspace example and packed-consumer CI for AI SDK 6.0.0/6.0.300 with Zod 3.25.76/4.1.8/4.6.5. Source prepares SDK 1.1.2 and adapter 0.1.2; registry publication remains pending.

Validation: source CI passes, including the restarted merged-head run. The separate terminal showcase passes four tests, process-restart recall smoke, TypeScript and formatting checks against the combined implementation. Public PR CI must pass before merge. Local fixtures do not verify hosted persistence or actual engine correction.

Add user-scoped `memory_search` and `memory_ingest` tools for AI SDK 6 through `@atomicmemory/vercel-ai-sdk/tools`. Rename the Vercel adapter to `@atomicmemory/vercel-ai-sdk` and support `new MemoryClient()` with Cloud defaults and the server's `ATOMICMEMORY_API_KEY`.

The application binds user identity outside model arguments. Tools validate arguments, bound search output while retaining retrieval evidence, sanitize failures, forward cancellation, and reject pending HTTP 202 ingest responses. The framework-free root entry keeps AI SDK and Zod optional.

Include a runnable workspace example and packed-consumer CI for AI SDK 6.0.0/6.0.300 with Zod 3.25.76/4.1.8/4.6.5. Source prepares SDK 1.1.2 and adapter 0.1.2; registry publication remains pending.

Validation: source CI passes, including the restarted merged-head run. The separate terminal showcase passes four tests, process-restart recall smoke, TypeScript and formatting checks against the combined implementation. Public PR CI must pass before merge. Local fixtures do not verify hosted persistence or actual engine correction.
@chatgpt-codex-connector

chatgpt-codex-connector Bot commented Oct 6, 2026 •

Copy link
Copy Markdown

Codex Review Summary

This comment shows the latest Codex review activity on this pull request.

Review Status Commit Review trigger
📝 Code Review ✅ Completed 2026-10-06T04:52:38.878355Z 43b166d PR opened
ℹ️ About Codex in GitHub

Your team has set up Codex to review pull requests in this repo. Reviews are triggered when you

  • Open a pull request for review
  • Mark a draft as ready
  • Comment "@codex review" or "@codex security review".

Codex reacts with 👀 while any review is running, comments if it has suggestions, and reacts with 👍 once all reviews finish with no findings.

@chatgpt-codex-connector chatgpt-codex-connector Bot left a comment

Copy link
Copy Markdown

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

💡 Codex Review

Here are some automated review suggestions for this pull request.

Reviewed commit: 43b166dca6

ℹ️ About Codex in GitHub

Your team has set up Codex to review pull requests in this repo. Reviews are triggered when you

  • Open a pull request for review
  • Mark a draft as ready
  • Comment "@codex review".

If Codex has suggestions, it will comment; otherwise it will react with 👍.

Codex can also answer questions or update the PR. Try commenting "@codex address that feedback".

}

function projectHit(hit: z.infer<typeof hitSchema>, budget: number): MemorySearchHit {
const content = cutText(hit.memory.content, Math.min(MAX_HIT_CONTENT, budget));

Copy link
Copy Markdown

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

P1 Badge Sanitize retrieved content before returning it to the model

When a stored memory contains prompt-control text or model-facing markup, this path only truncates it and then returns it verbatim in the memory_search tool result, so the model receives the hostile content as part of its next prompt and may follow it despite the tool description calling it reference data. Route every retrieved hit through the shared memory→model sanitizer and add an adversarial test against the AI SDK model prompt rather than testing only field allowlisting and length bounds.

AGENTS.md reference: AGENTS.md:L49-L58

Useful? React with 👍 / 👎.

@cuevaio cuevaio closed this Oct 6, 2026
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

1 participant