chore: Update AI SDK dependencies to v2 and migrate tests to Bun#40
chore: Update AI SDK dependencies to v2 and migrate tests to Bun#40
Conversation
WalkthroughMigrates tests to Bun’s Jest-like API, updates imports to node:fs/path, refactors LLM embedding and generation options (token usage fields, maxOutputTokens), adjusts OpenRouter caching handling, modifies fragment ID generation, adds KnowledgeService getMemories/deleteMemory, and bumps package and dependencies. Changes
Sequence Diagram(s)sequenceDiagram
autonumber
actor Caller
participant LLM as LLM Module
participant Provider as AI Provider
Note over LLM,Provider: Text generation (updated params & token accounting)
Caller->>LLM: generateText(prompt, { maxOutputTokens, model, ... })
rect rgba(230,240,255,0.5)
LLM->>Provider: generate({ maxOutputTokens, ... })
Provider-->>LLM: result { usage: { inputTokens, outputTokens }, text }
end
LLM-->>Caller: { text, usage: { inputTokens, outputTokens } }
sequenceDiagram
autonumber
actor Caller
participant LLM as LLM Module
participant OpenAI as OpenAI Embed API
Note over LLM,OpenAI: Embedding (refactored embedOptions)
Caller->>LLM: generateTextEmbedding(value, { model, dimensions? })
rect rgba(230,255,230,0.5)
LLM->>OpenAI: embed({ model, value, dimensions? })
OpenAI-->>LLM: embedding[]
end
LLM-->>Caller: embedding[]
sequenceDiagram
autonumber
actor Client
participant KS as KnowledgeService
participant Store as Memory Store
Note over KS,Store: New public methods
Client->>KS: getMemories(params)
KS->>Store: query(params + agent scope)
Store-->>KS: Memory[]
KS-->>Client: Memory[]
Client->>KS: deleteMemory(memoryId)
KS->>Store: delete(memoryId)
Store-->>KS: ack
KS-->>Client: void
Estimated code review effort🎯 4 (Complex) | ⏱️ ~60 minutes Poem
Pre-merge checks and finishing touches✅ Passed checks (3 passed)
✨ Finishing touches
🧪 Generate unit tests
Tip 👮 Agentic pre-merge checks are now available in preview!Pro plan users can now enable pre-merge checks in their settings to enforce checklists before merging PRs.
Please see the documentation for more information. Example: reviews:
pre_merge_checks:
custom_checks:
- name: "Undocumented Breaking Changes"
mode: "warning"
instructions: |
Pass/fail criteria: All breaking changes to public APIs, CLI flags, environment variables, configuration keys, database schemas, or HTTP/GraphQL endpoints must be documented in the "Breaking Change" section of the PR description and in CHANGELOG.md. Exclude purely internal or private changes (e.g., code not exported from package entry points or explicitly marked as internal).Please share your feedback with us on this Discord post. Thanks for using CodeRabbit! It's free for OSS, and your support helps us grow. If you like it, consider giving us a shout-out. Comment |
![coderabbitai[bot]](https://avatars.githubusercontent.com/in/347564?s=60&v=4){kind=small}
There was a problem hiding this comment.
Actionable comments posted: 1
🧹 Nitpick comments (1)
src/actions.ts (1)
188-188: Inconsistency with file processing: missing fragment count in response.The file processing path handles the
fragmentCountfrom the result (line 157), but the text processing path doesn't use the returned value fromaddKnowledge. This inconsistency means users won't get information about how many fragments their text was split into, unlike when processing files.Consider capturing and using the result to provide consistent feedback:
- await service.addKnowledge(knowledgeOptions); + const result = await service.addKnowledge(knowledgeOptions); response = { - text: `I've added that information to my knowledge base. It has been stored and indexed for future reference.`, + text: `I've added that information to my knowledge base. It has been split into ${result?.fragmentCount || 1} searchable fragment${(result?.fragmentCount || 1) > 1 ? 's' : ''} and indexed for future reference.`, };
📜 Review details
Configuration used: CodeRabbit UI
Review profile: CHILL
Plan: Pro
Cache: Disabled due to data retention organization setting
Knowledge base: Disabled due to data retention organization setting
⛔ Files ignored due to path filters (1)
bun.lockis excluded by!**/*.lock
📒 Files selected for processing (6)
__tests__/action.test.ts(15 hunks)package.json(3 hunks)src/actions.ts(3 hunks)src/llm.ts(11 hunks)src/service.ts(2 hunks)src/tests.ts(1 hunks)
🔇 Additional comments (18)
src/tests.ts (1)
16-17: LGTM! Import path migration to Node.js conventions.The migration from
fsandpathtonode:fsandnode:pathaligns with Node.js best practices for built-in module imports. This change makes it explicit that these are core Node.js modules rather than npm packages.src/actions.ts (2)
10-11: LGTM! Import path migration to Node.js conventions.The migration from
fsandpathtonode:fsandnode:pathaligns with Node.js best practices for built-in module imports.
157-157: Remove unnecessary optional chaining for fragmentCountaddKnowledge's return type includes fragmentCount: number (see src/service.ts), so result.fragmentCount is always present; replace
${result?.fragmentCount || 0}with${result.fragmentCount}(or keep|| 0only if you intentionally want a defensive display).
Location: src/actions.ts:157Likely an incorrect or invalid review comment.
src/service.ts (3)
480-480: LGTM! Improved logging readability.Using template literal instead of string concatenation improves code readability.
876-902: LGTM! New public API methods for memory management.The addition of
getMemoriesanddeleteMemorymethods provides essential memory management capabilities that were previously missing from the public API. The methods are well-documented and properly scoped to the agent.
843-844: No change required — createUniqueUuid correctly takes runtime as first arg.
Both src/service.ts (lines 841–845) and src/routes.ts (around line 383) call createUniqueUuid(runtime, …), so passing this.runtime + fragmentIdContent is consistent with existing usage and does not change fragment-ID semantics.Likely an incorrect or invalid review comment.
__tests__/action.test.ts (2)
1-17: LGTM! Successful migration to Bun test framework.The migration from Vitest to Bun's built-in test framework is well executed, maintaining Jest-like syntax for familiarity.
9-17: LGTM! Proper module mocking for Node.js core modules.The mocking setup correctly handles the
node:fsandnode:pathimports using Bun'smock.moduleAPI.package.json (3)
4-4: Verify version bump alignment with project versioning strategy.The version has jumped from 1.2.3 to 1.5.10. Please confirm this aligns with the project's versioning strategy and that all intermediate versions have been properly released or documented.
52-52: LGTM! Addition of Bun types for test migration.The addition of
@types/bunas a devDependency properly supports the test framework migration.
30-36: Verify AI SDK v2 compatibility and ai v5 interoperabilitypackage.json now depends on @ai-sdk/* (v2) and ai@^5.0.48 — these are major-version upgrades; confirm migration and compatibility.
- Update all SDK usages to the v2 API (search for "@ai-sdk/" imports and adjust call signatures).
- Follow each provider's v2 migration guide (Anthropic, Google, OpenAI).
- Confirm ai@5 is compatible with the @ai-sdk v2 providers; run unit/integration tests and a full build.
Location: package.json (dependencies) — @ai-sdk/anthropic ^2.0.17, @ai-sdk/google ^2.0.14, @ai-sdk/openai ^2.0.32, ai ^5.0.48.
src/llm.ts (7)
126-151: LGTM! Improved embedding configuration structure.The refactor to use
embedOptionsobject with structured parameters is cleaner and more maintainable than the previous approach. The logging correctly referencesembedOptions.dimensionsand the conditional dimension parameter handling for OpenAI models is appropriate.
281-287: Parameter rename aligns with AI SDK v2.The change from
maxTokenstomaxOutputTokensand token usage field updates (inputTokens/outputTokensinstead ofpromptTokens/completionTokens) correctly align with AI SDK v2 conventions.
337-343: Token usage tracking updated consistently.The OpenAI text generation correctly uses the new token field names and
maxOutputTokensparameter, maintaining consistency with the AI SDK v2 migration.
373-379: Google provider updates aligned with SDK changes.The parameter and token usage field updates are consistent with the overall AI SDK v2 migration pattern seen throughout the file.
549-563: Caching implementation uses updated token fields correctly.The Claude caching implementation correctly uses
maxOutputTokensand the new token usage field names (inputTokens/outputTokens). The caching logic appears intact despite the parameter changes.
623-638: Gemini caching maintains functionality with updated parameters.The Gemini caching implementation correctly uses the updated parameter names and token fields while preserving the implicit caching logic for Gemini 2.5 models.
658-671: Standard OpenRouter text generation updated consistently.The standard (non-caching) OpenRouter implementation uses the correct updated parameter names and token field references, maintaining consistency with the rest of the file.
| maxTokens: number, | ||
| cacheDocument?: string, | ||
| cacheOptions?: { type: 'ephemeral' }, | ||
| _cacheOptions?: { type: 'ephemeral' }, |
There was a problem hiding this comment.
🧩 Analysis chain
Verify caching functionality still works as intended.
The parameter renaming to _cacheOptions (with underscore prefix) suggests it's now ignored, but the JSDoc comment on lines 208-211 still documents cacheOptions: { type: "ephemeral" } as a valid usage pattern. This could mislead users.
Run the following script to check if the caching functionality and documentation are consistent:
🏁 Script executed:
#!/bin/bash
# Verify caching parameter usage and documentation consistency
echo "=== Checking for cacheOptions usage in codebase ==="
rg -n "cacheOptions" --type=ts -C3
echo "=== Checking JSDoc examples that mention cacheOptions ==="
rg -n "cacheOptions.*ephemeral" --type=ts -A5 -B5
echo "=== Checking if _cacheOptions parameter is actually used ==="
ast-grep --pattern 'function generateOpenRouterText($$$, _cacheOptions$$$) {
$$$
}'Length of output: 4598
🏁 Script executed:
#!/bin/bash
set -euo pipefail
echo "=== grep for _cacheOptions ==="
rg -n "_cacheOptions" -S || true
echo
echo "=== grep for cacheOptions (all hits) ==="
rg -n "cacheOptions" --type=ts -S -n -C2 || true
echo
echo "=== all generateText call sites (with context) ==="
rg -n --type=ts -C3 "generateText\(" || true
echo
echo "=== src/llm.ts lines 180-260 (JSDoc + nearby code) ==="
sed -n '180,260p' src/llm.ts || true
echo
echo "=== src/llm.ts lines 380-440 (function signature + body head) ==="
sed -n '380,440p' src/llm.ts || true
echo
echo "=== src/document-processor.ts lines 660-700 (call site shown earlier) ==="
sed -n '660,700p' src/document-processor.ts || true
echo
echo "=== types.ts cacheOptions type area ==="
sed -n '80,110p' src/types.ts || trueLength of output: 12384
Fix ignored cacheOptions in generateOpenRouterText
The parameter _cacheOptions?: { type: 'ephemeral' } in src/llm.ts (generateOpenRouterText — line ~409) is never referenced; generateText passes overrideConfig?.cacheOptions (src/llm.ts ~lines 239–243) and callers (src/document-processor.ts:676–679) supply cacheOptions, so the option is effectively ignored. Either remove the underscore and actually consume cacheOptions (implement ephemeral behavior) or stop accepting/passing it and update the JSDoc/types to avoid misleading users.
Offending line (still present):
_cacheOptions?: { type: 'ephemeral' },
🤖 Prompt for AI Agents
In src/llm.ts around line 409, the parameter `_cacheOptions?: { type:
'ephemeral' }` is declared but never used; either consume it or stop accepting
it — choose to remove it to avoid misleading API: delete the unused
`_cacheOptions` parameter from generateOpenRouterText, update the function's
JSDoc/type signature to remove cacheOptions, remove any passing of this arg from
callers (e.g., update calls that supply cacheOptions so they no longer pass it),
and run/adjust any affected typings or docs so the option is no longer
advertised; if you prefer the other approach instead, rename `_cacheOptions` to
`cacheOptions`, implement handling for `{ type: 'ephemeral' }` within
generateOpenRouterText (apply ephemeral behavior to caching logic) and ensure
callers pass the option and types/JSDoc reflect the implemented behavior.
3 participants
Summary
Changes
Dependencies Updated
@ai-sdk/anthropic: ^1.2.11 → ^2.0.17@ai-sdk/google: ^1.2.18 → ^2.0.14@ai-sdk/openai: ^1.3.22 → ^2.0.32@openrouter/ai-sdk-provider: ^0.4.5 → ^1.2.0ai: ^4.3.17 → ^5.0.48Testing Framework
Breaking Changes
None - Plugin API remains unchanged
Testing
All existing tests pass with the new Bun test runner
Summary by CodeRabbit
New Features
Improvements
Chores
Tests