@effect/ai-openrouter `streamText` fails at end of stream

[nickbreaton]

What version of Effect is running?

3.19.19

What steps can reproduce the bug?

Any call to model.streamText when using OpenRouterLanguageModel layer appears to result in an error.

Reproducible example

https://effect.website/play#76acbd7ee284

import { LanguageModel, Prompt } from "@effect/ai"
import { OpenRouterClient, OpenRouterLanguageModel } from "@effect/ai-openrouter"
import { FetchHttpClient } from "@effect/platform"
import { NodeRuntime } from "@effect/platform-node"
import { Config, ConfigProvider, Effect, Layer, Stream } from "effect"

const prompt = Prompt.make([
  { role: "user", content: [{ type: "text", text: "Tell me a one line joke" }] }
])

const program = Effect.gen(function*() {
  const model = yield* LanguageModel.LanguageModel

  yield* model.streamText({ prompt }).pipe(
    Stream.filter((part) => part.type === "text-delta"),
    Stream.runForEach((part) => Effect.log("!!! PARSED !!!", part.delta))
  )
})

const layer = OpenRouterLanguageModel.layer({ model: "anthropic/claude-haiku-4.5" }).pipe(
  Layer.provide(OpenRouterClient.layerConfig({ apiKey: Config.redacted("OPEN_ROUTER_API_KEY") })),
  Layer.provide(Layer.setConfigProvider(ConfigProvider.fromEnv())),
  Layer.provide(FetchHttpClient.layer)
)

program.pipe(
  Effect.provide(layer),
  NodeRuntime.runMain
)

What is the expected behavior?

The stream completes without error.

What do you see instead?

The stream is successfully parsed until it terminates, the following error occurs:

 MalformedOutput: { "module": "OpenRouterLanguageModel", "method": "makeResponse", "description": "Received response with no valid choices", "cause": undefined }

Using the Effect playground, I'm able to able to inspect the SSE response via Chrome devtools network tab:

: OPENROUTER PROCESSING

: OPENROUTER PROCESSING

data: {"id":"gen-1772933731-Xt9lXqMjuIyE3SnQavae","object":"chat.completion.chunk","created":1772933731,"model":"anthropic/claude-4.5-haiku-20251001","provider":"Amazon Bedrock","choices":[{"index":0,"delta":{"content":"Why don't scientists trust atoms?","role":"assistant"},"finish_reason":null,"native_finish_reason":null}]}

data: {"id":"gen-1772933731-Xt9lXqMjuIyE3SnQavae","object":"chat.completion.chunk","created":1772933731,"model":"anthropic/claude-4.5-haiku-20251001","provider":"Amazon Bedrock","choices":[{"index":0,"delta":{"content":" Because they make up everything!","role":"assistant"},"finish_reason":null,"native_finish_reason":null}]}

data: {"id":"gen-1772933731-Xt9lXqMjuIyE3SnQavae","object":"chat.completion.chunk","created":1772933731,"model":"anthropic/claude-4.5-haiku-20251001","provider":"Amazon Bedrock","choices":[{"index":0,"delta":{"content":"","role":"assistant"},"finish_reason":"stop","native_finish_reason":"stop"}]}

data: {"id":"gen-1772933731-Xt9lXqMjuIyE3SnQavae","object":"chat.completion.chunk","created":1772933731,"model":"anthropic/claude-4.5-haiku-20251001","provider":"Amazon Bedrock","choices":[],"usage":{"prompt_tokens":13,"completion_tokens":16,"total_tokens":29,"cost":0.000093,"is_byok":false,"prompt_tokens_details":{"cached_tokens":0,"cache_write_tokens":0,"audio_tokens":0,"video_tokens":0},"cost_details":{"upstream_inference_cost":0.000093,"upstream_inference_prompt_cost":0.000013,"upstream_inference_completions_cost":0.00008},"completion_tokens_details":{"reasoning_tokens":0,"image_tokens":0,"audio_tokens":0}}}

data: [DONE]

Additional information

Also reported in Discord: https://discord.com/channels/795981131316985866/1338871274398679130/threads/1479636410310066377

[nickbreaton]

The second to last SSE message with "choices":[] is interesting, and potentially points to the root cause.

[nickbreaton]

I have attempted a fix for this: #6117