Ai sdk 5 migration #10903

timtmok · 2025-12-03T16:17:28Z

Address #10818

This used Vercel's MCP server to assist in the migration to v5.

Removed the workaround for the temperature
Updated the field names that changed
Fixed the responses to match the v5 format

Release Notes

New Features

Update ai-sdk to v5

Bug Fixes

N/A

QA Notes

I did some light testing with tool calling with various providers. The chat response handling is where things changed the most.

github-actions · 2025-12-03T16:17:47Z

E2E Tests 🚀
This PR will run tests tagged with: @:critical

^readme ^{valid tags}

sharon-wang

Looking good so far with Amazon Bedrock, OpenAI and OpenRouter via Custom Provider 👍 getPlot tool is working with Amazon Bedrock, so I wonder if our plot image handling is compatible with Anthropic models but not the OpenAI models?

Providers that don't support the responses endpoint aren't working though -- I think we'll need some way to override the chat endpoint for those providers to /chat/completions

extensions/positron-assistant/src/openai-fetch-utils.ts

extensions/positron-assistant/src/models.ts

extensions/positron-assistant/src/utils.ts

wch · 2025-12-05T05:14:15Z

extensions/positron-assistant/src/utils.ts

+ * AI SDK 5 supports images in tool results via the 'content' output type with 'media' parts.
 */
-function getPlotToolResultToAiMessage(part: vscode.LanguageModelToolResultPart2): ai.CoreUserMessage {
+function getPlotToolResultToAiMessage(part: vscode.LanguageModelToolResultPart2): ai.ToolModelMessage {


Just wondering: is this function still needed? I think that convertToolResultToAiMessageExperimentalContent should be able to handle tool results with images.

I removed it and tested the change. Various models and providers were able to handle the tool output without this function. It also fixed the problem with GPT models on Snowflake where it couldn't use the get plot tool output.

wch · 2025-12-05T05:18:51Z

extensions/positron-assistant/src/utils.ts

 						if (cacheBreakpoint && bedrockCacheBreakpoint) {
 							cacheBreakpoint = false;
 							markBedrockCacheBreakpoint(toolMessage);
 						}


This probably doesn't have to be addressed just yet, but if/when Anthropic goes through this code path (instead of the separate Anthropic provider code), then we'll want to support their cache breakpoint markers as well.

extensions/positron-assistant/src/utils.ts

extensions/positron-assistant/src/openai-fetch-utils.ts

wch · 2025-12-05T06:16:11Z

extensions/positron-assistant/src/openai-fetch-utils.ts

+							// Fix empty role field
+							if (choice.delta.role === '') {
+								choice.delta.role = 'assistant';
+							}


Is this check here because of Snowflake, or other providers that don't quite adhere to the OpenAI completions format?

Yeah this was originally added for Snowflake, but it's possible other providers may omit this as well

wch · 2025-12-05T06:40:49Z

extensions/positron-assistant/src/openai-fetch-utils.ts

 					}
 				}

 				transformedLines.push(`data: ${JSON.stringify(data)}`);


Using the types from OpenAI, I suggest actually extracting a lot of the code above into a function which takes as input something that's close to a ChatCompletionChunk, but where the type takes into account the possibly-missing fields. The function would return something that actually is a ChatCompletionChunk.

Then you can really lean into type checking to make sure all the cases are taken care of.

For example, if you define this type, then the role could be "":

type PossiblyBrokenChatCompletionChunk = Omit<OpenAI.ChatCompletionChunk, 'choices'> & { choices: Array<Omit<OpenAI.ChatCompletionChunk['choices'][0], 'delta'> & { delta: Omit<OpenAI.ChatCompletionChunk['choices'][0]['delta'], 'role'> & { role?: OpenAI.ChatCompletionChunk['choices'][0]['delta']['role'] | '' } }> };

I know that's kind of awkward but it does work. It could probably be made a bit simpler and easier to read by referencing the types by name instead of by path. (I would just ask an LLM to add in the other possibly-empty fields as well.)

Next, you'd modify the type guard function above to have the signature:

export function isPossiblyBrokenChatCompletionChunk(obj: unknown): obj is PossiblyBrokenChatCompletionChunk { // body of this function can remain the same }

Then you could add a function with this signature:

function fixPossiblyBrokenChatCompletionChunk(chunk: PossiblyBrokenChatCompletionChunk): OpenAI.ChatCompletionChunk { // Implementation goes here }

Then up above you'd use:

if (isPossiblyBrokenChatCompletionChunk(data)) { const fixedChunk = fixPossiblyBrokenChatCompletionChunk(data) transformedLines.push(`data: ${JSON.stringify(fixedChunk)}`);

I've refactored it so there's a function to check if it's a possible chat completion chunk and another function to fill in any missing fields.

sharon-wang

I can log in with Snowflake now ❄️ ✅

I'm not sure if this is my local state, but I lost the Snowflake icon in the models popup:

The icon is there in a release build of 2026.01.0+24, so not sure if it's my dev build state or possibly something has changed?

extensions/positron-assistant/src/constants.ts

sharon-wang · 2025-12-05T15:32:22Z

extensions/positron-assistant/src/openai-fetch-utils.ts

+							// Fix empty role field
+							if (choice.delta.role === '') {
+								choice.delta.role = 'assistant';
+							}


Yeah this was originally added for Snowflake, but it's possible other providers may omit this as well

sharon-wang · 2025-12-05T15:37:53Z

extensions/positron-assistant/src/openai-fetch-utils.ts

+		// Transform tools to be compatible with OpenAI-compatible providers
+		// Some providers don't support the 'strict' field in tool function definitions
+		if (requestBody.tools && Array.isArray(requestBody.tools)) {
+			log.debug(`[${providerName}] Request contains ${requestBody.tools.length} tools: ${requestBody.tools.map((t: any) => t.function?.name || t.name).join(', ')}`);
+			for (const tool of requestBody.tools) {
+				if (tool.function && tool.function.strict !== undefined) {
+					delete tool.function.strict;
+					log.debug(`[${providerName}] Removed 'strict' field from tool: ${tool.function.name}`);
+				}
+			}
+			log.trace(`[${providerName}] Tools payload: ${JSON.stringify(requestBody.tools)}`);


Thoughts on consolidating the empty input schema handling that is currently in models.ts in AILanguageModel.provideLanguageModelChatResponse() with the handling here, so that all the tool modifications are in one place?

Maybe moving it to the utils.ts for generic handling? Although I removed it, there was some model specific handling. I'm good either way as long as models.ts doesn't grow too large as it is already a all-in-one kind of file.

I really hope that providers adhere to a standard API. This special handling feels like a side effect of providers not implementing the API.

timtmok · 2025-12-10T18:39:53Z

I tried these latest changes with different models and providers and things feel good. The bonus is Snowflake works with GPT models calling the GetPlot tool. Its response is brief but it is able to understand the image.

The response is from GPT-5 but was also similar with GPT-4.1.

The Snowflake icon seems okay for me as well but these changes shouldn't have changed that.

wch · 2025-12-11T05:44:20Z

extensions/positron-assistant/src/openai-fetch-utils.ts

+ * @param obj The object to check
+ * @returns True if the object appears to be a ChatCompletionChunk (possibly malformed)
+ */
+export function isPossiblyBrokenChatCompletionChunk(obj: unknown): obj is PossiblyBrokenChatCompletionChunk {


I think that this type guard isn't correct for the definition of PossiblyBrokenChatCompletionChunk above. With that interface definition, even an empty object, {}, could count as a PossiblyBrokenChatCompletionChunk.

But this function requires it to be an object and have:

id: string

choices: Array<unknown>

object: "chat.completion.chunk"

With this mismatch between the type and the type guard, the type guard could tell you something is not a PossiblyBrokenChatCompletionChunk when in fact it is, and that could be a source of bugs in the future.

The type and the type guard should match (or at least be much closer). I suggest updating the type to include essential things like the id, choices, and object -- whatever things we can reasonably assume that an OpenAI-compatible implementation would include.

wch · 2025-12-11T05:47:37Z

extensions/positron-assistant/src/openai-fetch-utils.ts

-					}
+				// Check if it's a possibly broken chunk and fix it
+				// Otherwise, keep the original line
+				if (isPossiblyBrokenChatCompletionChunk(data)) {


Just a follow on to the note above about the type and type guard mismatch: if isPossiblyBrokenChatCompletionChunk() actually matched the PossiblyBrokenChatCompletionChunk type, then it would always enter this if block and never go into the else. (Assuming that data is a JS object, as opposed to something like a string or null.)

timtmok added 2 commits December 2, 2025 16:04

Update packages to AI SDK 5

16f543a

Fix up tool calls

b95d0b6

timtmok marked this pull request as ready for review December 3, 2025 17:03

timtmok requested a review from sharon-wang December 3, 2025 17:03

sharon-wang requested changes Dec 3, 2025

View reviewed changes

extensions/positron-assistant/src/openai-fetch-utils.ts Show resolved Hide resolved

extensions/positron-assistant/src/models.ts Outdated Show resolved Hide resolved

extensions/positron-assistant/src/models.ts Show resolved Hide resolved

sharon-wang requested a review from wch December 3, 2025 22:42

Fix Snowflake provider

085ec40

timtmok requested a review from sharon-wang December 4, 2025 19:54

wch reviewed Dec 5, 2025

View reviewed changes

sharon-wang reviewed Dec 5, 2025

View reviewed changes

timtmok added 2 commits December 10, 2025 12:45

Clean up types and experimental content

b16ddc4

Revert test code

e42e52b

timtmok requested review from sharon-wang and wch December 10, 2025 18:40

wch reviewed Dec 11, 2025

View reviewed changes

Ai sdk 5 migration #10903

Are you sure you want to change the base?

Ai sdk 5 migration #10903

Conversation

timtmok commented Dec 3, 2025

Release Notes

New Features

Bug Fixes

QA Notes

Uh oh!

github-actions bot commented Dec 3, 2025 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

sharon-wang left a comment

Choose a reason for hiding this comment

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

sharon-wang left a comment

Choose a reason for hiding this comment

Uh oh!

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

timtmok commented Dec 10, 2025

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

4 participants

github-actions bot commented Dec 3, 2025 •

edited

Loading