NewYour coding agent can read the release notes before it upgrades.Set up the MCP server →
npm · #4234 most downloaded on npm
Last release 3 days ago
30 Sep 2026
Ships on a steady schedule
a new release about every 8 days
Most releases are documented
notes for 38 of the last 60 stable releases
Nothing withdrawn
no release was ever pulled
2 years old
689 releases · first in 2024
69bb613: feat(harness): support request transformations in network sandbox abstraction and use it to apply credential brokering when available
package.json and pnpm-lock.yaml files at build time instead of reading them at runtime to fix runtime errors in certain environmentsOne column per month.
### Patch Changes - Updated dependencies [8d717b3] - @ai-sdk/harness@1.0.71 - @ai-sdk/harness-acp@1.0.8
Updated dependencies [ 14810cb ]
Updated dependencies [ bf29419 ]
Updated dependencies [ 594787e ]
### Patch Changes - Updated dependencies [d300737] - @ai-sdk/harness@1.0.75
### Patch Changes - @ai-sdk/harness@1.0.74
Updated dependencies [ 77d33c0 ]
HarnessAgent via output property80b40aa: feat(harness-cline): implement Cline harness adapter
c38a0db: fix (provider/google): allow empty candidates array when streaming
0509c34: fix (provider/google): add name/content details to tool responses
36b03b0: chore (provider/google): remove topK model setting
@ai-sdk/provider-utils@2.0.0-canary.3
@ai-sdk/provider-utils@2.0.0-canary.2
0d3d3f5: chore (providers): remove baseUrl option
66060f7: chore (release): bump major version to 4.0
11beb9d: fix (provider/google): support tool calls without parameters
1486128: feat: add supportsUrl to language model specification
7944e61: feat (provider/google): enable frequencyPenalty and presencePenalty settings
d595d0d: feat (ai/core): file content parts
5d113fb: feat (provider/google): support fine-tuned models
2e32bd0: fix (provider/google): deal with empty candidatesTokenCount
### Patch Changes - Updated dependencies [273f696] - @ai-sdk/provider-utils@1.0.19
### Patch Changes - Updated dependencies - @ai-sdk/provider-utils@1.0.18 - @ai-sdk/provider@0.0.23
26515cb: feat (ai/provider): introduce ProviderV1 specification
### Patch Changes - Updated dependencies [09f895f] - @ai-sdk/provider-utils@1.0.16
cb94042: fix (provider/google): allow disabling structured generation
### Patch Changes - Updated dependencies [d67fa9c] - @ai-sdk/provider-utils@1.0.15
9c1f4f2: chore (provider/google): remove "models/" from model name
5da9c7f: feat (provider/google): embedding support
### Patch Changes - Updated dependencies [f2c025e] - @ai-sdk/provider@0.0.21 - @ai-sdk/provider-utils@1.0.14
### Patch Changes - Updated dependencies [6ac355e] - @ai-sdk/provider@0.0.20 - @ai-sdk/provider-utils@1.0.13
123134b: fix (provider/google): change default object generation mode to json
89b18ca: fix (ai/provider): send finish reason 'unknown' by default
### Patch Changes - Updated dependencies - @ai-sdk/provider-utils@1.0.10 - @ai-sdk/provider@0.0.18
1e94ed8: feat (provider/google): send json schema in object-json mode
### Patch Changes - Updated dependencies [029af4c] - @ai-sdk/provider@0.0.17 - @ai-sdk/provider-utils@1.0.9
### Patch Changes - Updated dependencies [d58517b] - @ai-sdk/provider@0.0.16 - @ai-sdk/provider-utils@1.0.8
### Patch Changes - Updated dependencies [96aed25] - @ai-sdk/provider@0.0.15 - @ai-sdk/provider-utils@1.0.7
### Patch Changes - Updated dependencies - @ai-sdk/provider-utils@1.0.6
a8d1c9e9: feat (ai/core): parallel image download
Updated dependencies [4f88248f]
b6cae9b : feat(alibaba): preserve reasoning in multi-turn requests by default on supported models
2e59d266: feat (provider/google): add cachedContent optional setting
Updated dependencies [b7290943]
Updated dependencies [d481729f]
5edc6110: feat (ai/core): add custom request header support
Updated dependencies [02f6a088]
### Patch Changes - Updated dependencies - @ai-sdk/provider-utils@0.0.15
4728c37f: feat (core): add text embedding model support to provider registry
Updated dependencies [102ca22f]
09295e2e: feat (@ai-sdk/google): automatically download image URLs
8022d73e: feat (provider/google): add safety setting option on models
f39c0dd2: feat (provider): implement toolChoice support
24683b72: fix (providers): Zod is required dependency
### Patch Changes - Updated dependencies - @ai-sdk/provider@0.0.6 - @ai-sdk/provider-utils@0.0.9
df47a34: feat (provider/google): add models/gemini-1.5-flash-latest to list of supported models
9a7ec8b: feat (provider/google): add token usage information
### Patch Changes - Updated dependencies [0f6bc4e] - @ai-sdk/provider@0.0.5 - @ai-sdk/provider-utils@0.0.8
### Patch Changes - Updated dependencies [325ca55] - @ai-sdk/provider@0.0.4 - @ai-sdk/provider-utils@0.0.7
Your coding agent can read these notes before it upgrades. Set up the MCP server →