chore: update nix node_modules hashes

bun.lock

@@ -350,7 +350,6 @@
         "@ai-sdk/cohere": "3.0.27",
         "@ai-sdk/gateway": "3.0.104",
         "@ai-sdk/google-vertex": "4.0.128",
-        "@ai-sdk/groq": "3.0.31",
         "@ai-sdk/mistral": "3.0.51",
         "@ai-sdk/openai-compatible": "2.0.41",
         "@ai-sdk/perplexity": "3.0.26",

nix/hashes.json

@@ -1,8 +1,8 @@
 {
   "nodeModules": {
-    "x86_64-linux": "sha256-QWLIdvu985FH5I9cZJOAuoeFeXU+4Jx9RzBB9RPoeeQ=",
-    "aarch64-linux": "sha256-SSzGD5hMj2vFvyw+dUPR9g/ZH6qhs0ZyZ/DnltZt3N8=",
-    "aarch64-darwin": "sha256-CeFUxiV+e8pKho+YcSclC3soQBogoxNMxwyIMztAExU=",
-    "x86_64-darwin": "sha256-FYwcACzU72y0+KtOpFfU7ndak8vMasqMgd5NLS6+XtY="
+    "x86_64-linux": "sha256-Q7BQ46mKePJtaKzhHxahIXy/pZczPmm5cQuBDrgd2Bc=",
+    "aarch64-linux": "sha256-pqk4iUhXzEc4ei9zpeGpPjX7Q6pxH1K5rgotD5Wf91s=",
+    "aarch64-darwin": "sha256-1q3mK5zLqQA0vz7KErDOkjeAnmsTReI0lhBJfIobC/E=",
+    "x86_64-darwin": "sha256-dBMQ6tZxt5VjgWTZELHgPk6fVhBfNYfmY+AnQ3iJ88Q="
   }
 }

packages/ai/src/providers/groq.ts

@@ -0,0 +1,116 @@
+import { Effect, Schema } from "effect"
+import type { ProviderPackage } from "../provider-package.js"
+import { OpenAIChat } from "../protocols/openai-chat.js"
+import { ProviderShared } from "../protocols/shared.js"
+import { AuthOptions, type ProviderAuthOption } from "../route/auth-options.js"
+import { Route, type RouteDefaultsInput } from "../route/client.js"
+import { Endpoint } from "../route/endpoint.js"
+import { Framing } from "../route/framing.js"
+import { Protocol } from "../route/protocol.js"
+import { ProviderID, type ModelID, type LLMRequest } from "../schema/index.js"
+import { profiles } from "./openai-compatible-profile.js"
+import type { OpenAIProviderOptionsInput } from "./openai-options.js"
+
+export const id = ProviderID.make("groq")
+
+export type ProviderOptions = Pick<OpenAIProviderOptionsInput, "reasoningEffort"> & {
+  /** Controls visible reasoning on GPT-OSS; other models always use parsed reasoning. */
+  readonly includeReasoning?: boolean
+  readonly parallelToolCalls?: boolean
+  readonly serviceTier?: "on_demand" | "flex" | "auto" | "performance" | (string & {})
+  readonly user?: string
+}
+
+export type LanguageModelOptions = Omit<RouteDefaultsInput, "providerOptions"> &
+  ProviderAuthOption<"optional"> & {
+    readonly baseURL?: string
+    readonly providerOptions?: ProviderOptions
+  }
+
+export interface Settings extends ProviderPackage.Settings {
+  readonly apiKey?: string
+  readonly baseURL?: string
+  readonly providerOptions?: ProviderOptions
+}
+
+const Options = Schema.Struct({
+  includeReasoning: Schema.optional(Schema.Boolean),
+  parallelToolCalls: Schema.optional(Schema.Boolean),
+  serviceTier: Schema.optional(Schema.String),
+  user: Schema.optional(Schema.String),
+})
+
+export const protocol = Protocol.make({
+  id: "groq-chat",
+  body: {
+    schema: Schema.Struct({
+      ...OpenAIChat.bodyFields,
+      reasoning_format: Schema.optional(Schema.Literal("parsed")),
+      include_reasoning: Schema.optional(Schema.Boolean),
+      parallel_tool_calls: Schema.optional(Schema.Boolean),
+      service_tier: Schema.optional(Schema.String),
+      user: Schema.optional(Schema.String),
+    }),
+    from: Effect.fn("Groq.fromRequest")(function* (request: LLMRequest) {
+      const options = yield* ProviderShared.validateWith(Schema.decodeUnknownEffect(Options))(
+        request.providerOptions ?? {},
+      )
+      const gptOSS = request.model.id.startsWith("openai/gpt-oss-")
+      return {
+        ...(yield* OpenAIChat.fromRequest(request)),
+        reasoning_format: gptOSS ? undefined : ("parsed" as const),
+        include_reasoning: gptOSS ? options.includeReasoning : undefined,
+        parallel_tool_calls: options.parallelToolCalls,
+        service_tier: options.serviceTier,
+        user: options.user,
+      }
+    }),
+  },
+  stream: OpenAIChat.protocol.stream,
+})
+
+export const route = Route.make({
+  id: "groq-chat",
+  provider: id,
+  providerMetadataKey: "openai",
+  protocol,
+  endpoint: Endpoint.path("/chat/completions", { baseURL: profiles.groq.baseURL }),
+  framing: Framing.sse,
+})
+
+export const configure = (input: LanguageModelOptions = {}) => {
+  const { apiKey: _apiKey, auth: _auth, baseURL, ...defaults } = input
+  const configured = route.with({
+    ...defaults,
+    endpoint: { baseURL: baseURL ?? profiles.groq.baseURL },
+    auth: AuthOptions.bearer(input, "GROQ_API_KEY"),
+  })
+  return {
+    id,
+    model: (modelID: string | ModelID) =>
+      configured.model<ProviderOptions>({
+        id: modelID,
+        compatibility: {
+          maxTokensField: "max_completion_tokens",
+          reasoningField: "reasoning",
+          requireReasoning: false,
+          supportsStore: false,
+          supportsStrictMode: false,
+        },
+      }),
+    configure,
+  }
+}
+
+export const provider = configure()
+
+export const model: ProviderPackage.Definition<Settings, ProviderOptions>["model"] = (modelID, settings) =>
+  configure({
+    apiKey: settings.apiKey,
+    baseURL: settings.baseURL,
+    headers: settings.headers === undefined ? undefined : { ...settings.headers },
+    http: settings.body === undefined ? undefined : { body: { ...settings.body } },
+    providerOptions: settings.providerOptions,
+  }).model(modelID)
+
+export * as Groq from "./groq.js"

packages/ai/src/providers/index.ts

@@ -12,6 +12,7 @@ export * as GoogleVertex from "./google-vertex.js"
 export * as GoogleVertexChat from "./google-vertex-chat.js"
 export * as GoogleVertexMessages from "./google-vertex-messages.js"
 export * as GoogleVertexResponses from "./google-vertex-responses.js"
+export * as Groq from "./groq.js"
 export * as OpenAI from "./openai.js"
 export * as OpenAICompatible from "./openai-compatible.js"
 export * as OpenAICompatibleResponses from "./openai-compatible-responses.js"

packages/ai/test/fixtures/recordings/groq-chat/streams-text-with-usage.json

@@ -0,0 +1,29 @@
+{
+  "version": 1,
+  "metadata": {
+    "model": "openai/gpt-oss-20b",
+    "tags": ["prefix:groq-chat", "provider:groq", "protocol:groq-chat", "text", "usage"],
+    "name": "groq-chat/streams-text-with-usage",
+    "recordedAt": "2026-08-26T14:40:09.833Z"
+  },
+  "interactions": [
+    {
+      "transport": "http",
+      "request": {
+        "method": "POST",
+        "url": "https://api.groq.com/openai/v1/chat/completions",
+        "headers": {
+          "content-type": "application/json"
+        },
+        "body": "{\"model\":\"openai/gpt-oss-20b\",\"messages\":[{\"role\":\"user\",\"content\":\"Reply with exactly one word: hello\"}],\"stream\":true,\"stream_options\":{\"include_usage\":true},\"reasoning_effort\":\"low\",\"max_completion_tokens\":512,\"include_reasoning\":false,\"service_tier\":\"on_demand\",\"user\":\"recorded-test\"}"
+      },
+      "response": {
+        "status": 200,
+        "headers": {
+          "content-type": "text/event-stream"
+        },
+        "body": "data: {\"id\":\"chatcmpl-660048b0-3c99-4215-a582-e116b77eb881\",\"object\":\"chat.completion.chunk\",\"created\":1787755209,\"model\":\"openai/gpt-oss-20b\",\"system_fingerprint\":\"fp_66891002f6\",\"choices\":[{\"index\":0,\"delta\":{\"role\":\"assistant\",\"content\":\"\"},\"logprobs\":null,\"finish_reason\":null}],\"x_groq\":{\"id\":\"req_01m0z87915eep9bpf10gg7331e\",\"seed\":94036161}}\n\ndata: {\"id\":\"chatcmpl-660048b0-3c99-4215-a582-e116b77eb881\",\"object\":\"chat.completion.chunk\",\"created\":1787755209,\"model\":\"openai/gpt-oss-20b\",\"system_fingerprint\":\"fp_66891002f6\",\"choices\":[{\"index\":0,\"delta\":{\"content\":\"hello\"},\"logprobs\":null,\"finish_reason\":null}]}\n\ndata: {\"id\":\"chatcmpl-660048b0-3c99-4215-a582-e116b77eb881\",\"object\":\"chat.completion.chunk\",\"created\":1787755209,\"model\":\"openai/gpt-oss-20b\",\"system_fingerprint\":\"fp_66891002f6\",\"choices\":[{\"index\":0,\"delta\":{},\"logprobs\":null,\"finish_reason\":\"stop\"}],\"x_groq\":{\"id\":\"req_01m0z87915eep9bpf10gg7331e\",\"usage\":{\"queue_time\":0.10886435,\"prompt_tokens\":78,\"prompt_time\":0.003693734,\"completion_tokens\":20,\"completion_time\":0.020459983,\"total_tokens\":98,\"total_time\":0.024153717,\"completion_tokens_details\":{\"reasoning_tokens\":10}}},\"usage\":{\"queue_time\":0.10886435,\"prompt_tokens\":78,\"prompt_time\":0.003693734,\"completion_tokens\":20,\"completion_time\":0.020459983,\"total_tokens\":98,\"total_time\":0.024153717,\"completion_tokens_details\":{\"reasoning_tokens\":10}}}\n\ndata: {\"id\":\"chatcmpl-660048b0-3c99-4215-a582-e116b77eb881\",\"object\":\"chat.completion.chunk\",\"created\":1787755209,\"model\":\"openai/gpt-oss-20b\",\"system_fingerprint\":\"fp_66891002f6\",\"choices\":[],\"usage\":{\"queue_time\":0.10886435,\"prompt_tokens\":78,\"prompt_time\":0.003693734,\"completion_tokens\":20,\"completion_time\":0.020459983,\"total_tokens\":98,\"total_time\":0.024153717,\"completion_tokens_details\":{\"reasoning_tokens\":10}},\"service_tier\":\"on_demand\"}\n\ndata: [DONE]\n\n"
+      }
+    }
+  ]
+}

packages/ai/test/provider-package.test.ts

@@ -29,6 +29,7 @@ describe("provider package entrypoints", () => {
       import("@opencode-ai/ai/providers/togetherai"),
       import("@opencode-ai/ai/providers/cerebras"),
       import("@opencode-ai/ai/providers/deepinfra"),
+      import("@opencode-ai/ai/providers/groq"),
     ])
 
     for (const module of modules) expect(module.model).toBeFunction()

packages/ai/test/provider/groq.recorded.test.ts

@@ -0,0 +1,185 @@
+import { configure } from "@opencode-ai/ai/providers/groq"
+import { describe, expect } from "bun:test"
+import { Effect } from "effect"
+import { LLM, LLMEvent, LLMRequest, LLMResponse, Message, ToolChoice, ToolDefinition } from "../../src/index.js"
+import { LLMClient } from "../../src/route.js"
+import { compileRequest } from "../../src/route/client.js"
+import { recordedTests } from "../recorded-test.js"
+
+const apiKey = process.env.GROQ_API_KEY ?? "fixture"
+const recorded = recordedTests({
+  prefix: "groq-chat",
+  provider: "groq",
+  protocol: "groq-chat",
+  requires: ["GROQ_API_KEY"],
+})
+
+const weather = ToolDefinition.make({
+  name: "lookup_weather",
+  description: "Look up the current weather for a city",
+  inputSchema: {
+    type: "object",
+    properties: { city: { type: "string", enum: ["Paris", "London"] } },
+    required: ["city"],
+    additionalProperties: false,
+  },
+})
+
+describe("Groq recorded", () => {
+  recorded.effect.with(
+    "streams text with usage",
+    { tags: ["text", "usage"], metadata: { model: "openai/gpt-oss-20b" } },
+    () =>
+      Effect.gen(function* () {
+        const request = LLM.request({
+          model: configure({
+            apiKey,
+            providerOptions: {
+              includeReasoning: false,
+              reasoningEffort: "low",
+              serviceTier: "on_demand",
+              user: "recorded-test",
+            },
+          }).model("openai/gpt-oss-20b"),
+          prompt: "Reply with exactly one word: hello",
+          generation: { maxTokens: 512 },
+        })
+        const compiled = yield* compileRequest(request)
+        expect(compiled.body).toMatchObject({
+          max_completion_tokens: 512,
+          stream_options: { include_usage: true },
+          include_reasoning: false,
+          service_tier: "on_demand",
+          user: "recorded-test",
+        })
+        expect(compiled.body.max_tokens).toBeUndefined()
+        expect(compiled.body.store).toBeUndefined()
+        expect(compiled.body.reasoning_format).toBeUndefined()
+
+        const response = yield* LLMClient.generate(request)
+        expect(response.text.toLowerCase().trim()).toBe("hello")
+        expect(response.reasoning).toBe("")
+        expect(response.events.some(LLMEvent.is.textDelta)).toBe(true)
+        expectUsage(response)
+      }),
+    60_000,
+  )
+
+  for (const item of [
+    {
+      name: "continues Qwen parallel tool calls",
+      model: configure({ apiKey, providerOptions: { parallelToolCalls: true, reasoningEffort: "none" } }).model(
+        "qwen/qwen3.6-27b",
+      ),
+      cities: ["Paris", "London"],
+      reasoning: false,
+    },
+    {
+      name: "replays GPT OSS reasoning through a tool loop",
+      model: configure({ apiKey, providerOptions: { includeReasoning: true, reasoningEffort: "low" } }).model(
+        "openai/gpt-oss-20b",
+      ),
+      cities: ["Paris"],
+      reasoning: true,
+    },
+  ]) {
+    recorded.effect.with(
+      item.name,
+      {
+        tags: ["tool", "tool-loop", "usage", item.reasoning ? "reasoning" : "parallel"],
+        metadata: { model: item.model.id },
+      },
+      () =>
+        Effect.gen(function* () {
+          const request = LLM.request({
+            model: item.model,
+            prompt: `Look up the current weather in ${item.cities.join(" and ")}. Call lookup_weather once for each city in the same response before answering. After receiving all results, report each city's weather in one short sentence.`,
+            tools: [weather],
+            toolChoice: "required",
+            generation: { maxTokens: 1536 },
+          })
+          const compiled = yield* compileRequest(request)
+          expect(compiled.body.stream_options).toEqual({ include_usage: true })
+          expect(compiled.body.store).toBeUndefined()
+          expect(compiled.body.reasoning_format).toBe(item.reasoning ? undefined : "parsed")
+          expect(compiled.body.tools[0].function.strict).toBeUndefined()
+          if (!item.reasoning) expect(compiled.body.parallel_tool_calls).toBe(true)
+
+          const first = yield* LLMClient.generate(request)
+          expect(first.finishReason.normalized).toBe("tool-calls")
+          expect(first.toolCalls).toHaveLength(item.cities.length)
+          expect(new Set(first.toolCalls.map((call) => call.id)).size).toBe(item.cities.length)
+          expect(first.toolCalls.map((call) => call.input)).toEqual(
+            expect.arrayContaining(item.cities.map((city) => ({ city }))),
+          )
+          expect(first.toolCalls.every((call) => call.name === "lookup_weather")).toBe(true)
+          expectUsage(first)
+          if (item.reasoning) {
+            expect(first.reasoning.length).toBeGreaterThan(0)
+            expect(first.events.some(LLMEvent.is.reasoningDelta)).toBe(true)
+          }
+
+          const followUp = LLMRequest.update(request, {
+            toolChoice: ToolChoice.make("none"),
+            messages: [
+              ...request.messages,
+              first.message,
+              ...first.toolCalls.map((call) =>
+                Message.tool({ id: call.id, name: call.name, result: { condition: "sunny", temperature: "18C" } }),
+              ),
+            ],
+          })
+          const replay = yield* compileRequest(followUp)
+          if (item.reasoning) {
+            expect(replay.body.messages).toEqual(
+              expect.arrayContaining([expect.objectContaining({ role: "assistant", reasoning: first.reasoning })]),
+            )
+          }
+          expect(replay.body.reasoning_format).toBe(item.reasoning ? undefined : "parsed")
+
+          const second = yield* LLMClient.generate(followUp)
+          expect(second.finishReason.normalized).toBe("stop")
+          expect(second.toolCalls).toHaveLength(0)
+          expect(second.text.toLowerCase()).toContain("sunny")
+          item.cities.forEach((city) => expect(second.text).toContain(city))
+          expectUsage(second)
+        }),
+      60_000,
+    )
+  }
+
+  recorded.effect.with(
+    "streams Qwen parsed reasoning",
+    { tags: ["reasoning", "usage"], metadata: { model: "qwen/qwen3.6-27b" } },
+    () =>
+      Effect.gen(function* () {
+        const request = LLM.request({
+          model: configure({
+            apiKey,
+            providerOptions: { reasoningEffort: "default" },
+          }).model("qwen/qwen3.6-27b"),
+          prompt:
+            "What is 173 multiplied by 219? Think through the arithmetic, then reply with only the final integer.",
+          generation: { maxTokens: 2048 },
+        })
+        const compiled = yield* compileRequest(request)
+        expect(compiled.body).toMatchObject({ reasoning_format: "parsed", reasoning_effort: "default" })
+        expect(compiled.body.include_reasoning).toBeUndefined()
+
+        const response = yield* LLMClient.generate(request)
+        expect(response.text.replaceAll(",", "").trim()).toBe("37887")
+        expect(response.text).not.toContain("<think>")
+        expect(response.reasoning.length).toBeGreaterThan(0)
+        expect(response.events.some(LLMEvent.is.reasoningDelta)).toBe(true)
+        expectUsage(response)
+      }),
+    60_000,
+  )
+})
+
+function expectUsage(response: LLMResponse) {
+  expect(response.usage).toBeDefined()
+  expect(response.usage?.inputTokens).toBeGreaterThan(0)
+  expect(response.usage?.outputTokens).toBeGreaterThan(0)
+  expect(response.events.filter(LLMEvent.is.finish)).toHaveLength(1)
+}

packages/ai/test/provider/groq.test.ts

@@ -0,0 +1,112 @@
+import { expect } from "bun:test"
+import { Effect } from "effect"
+import { LanguageModel, LLM, Message } from "../../src/index.js"
+import { OpenAIChat } from "../../src/protocols/openai-chat.js"
+import { Groq } from "../../src/providers/groq.js"
+import { compileRequest } from "../../src/route/client.js"
+import { it } from "../lib/effect.js"
+import { weatherTool } from "../recorded-scenarios.js"
+
+it.effect("Groq reuses Chat streaming and defaults to parsed reasoning", () =>
+  Effect.gen(function* () {
+    expect(Groq.protocol.stream).toBe(OpenAIChat.protocol.stream)
+    const model = Groq.configure({ apiKey: "fixture" }).model("llama-3.3-70b-versatile")
+    expect(model.route.endpoint.baseURL).toBe("https://api.groq.com/openai/v1")
+    const compiled = yield* compileRequest(
+      LLM.request({ model, prompt: "Hello", tools: [weatherTool], generation: { maxTokens: 64 } }),
+    )
+    expect(compiled.body).toMatchObject({
+      max_completion_tokens: 64,
+      stream_options: { include_usage: true },
+      reasoning_format: "parsed",
+    })
+    for (const key of ["store", "max_tokens", "include_reasoning", "parallel_tool_calls", "service_tier", "user"])
+      expect(compiled.body[key]).toBeUndefined()
+    expect(compiled.body.tools?.[0]?.function).not.toHaveProperty("strict")
+  }),
+)
+
+it.effect("Groq lowers its own options for custom catalog identities and endpoints", () =>
+  Effect.gen(function* () {
+    const model = LanguageModel.update(
+      Groq.model("qwen/qwen3.6-27b", {
+        apiKey: "fixture",
+        baseURL: "https://gateway.example/v1",
+        headers: { "x-client": "test" },
+        body: { custom: "value" },
+        providerOptions: {
+          reasoningEffort: "default",
+          parallelToolCalls: true,
+          serviceTier: "flex",
+          user: "test-user",
+        },
+      }),
+      { provider: "custom-groq" },
+    )
+    const compiled = yield* compileRequest(
+      LLM.request({ model, prompt: "Hello", providerOptions: { parallelToolCalls: false, includeReasoning: false } }),
+    )
+    expect(model.route.endpoint.baseURL).toBe("https://gateway.example/v1")
+    expect(model.route.defaults.headers).toEqual({ "x-client": "test" })
+    expect(model.route.defaults.http?.body).toEqual({ custom: "value" })
+    expect(compiled.body).toMatchObject({
+      reasoning_effort: "default",
+      reasoning_format: "parsed",
+      parallel_tool_calls: false,
+      service_tier: "flex",
+      user: "test-user",
+    })
+    expect(compiled.body.include_reasoning).toBeUndefined()
+    for (const key of ["reasoningFormat", "reasoningEffort", "parallelToolCalls", "serviceTier"])
+      expect(compiled.body).not.toHaveProperty(key)
+  }),
+)
+
+it.effect("Groq replays reasoning only when present and preserves explicit reasoning exclusion", () =>
+  Effect.gen(function* () {
+    const compiled = yield* compileRequest(
+      LLM.request({
+        model: Groq.configure({ apiKey: "fixture" }).model("openai/gpt-oss-20b"),
+        messages: [
+          Message.user("Think"),
+          Message.assistant([
+            { type: "reasoning", text: "Thinking" },
+            { type: "text", text: "Answer" },
+          ]),
+          Message.user("Again"),
+          Message.assistant("Answer only"),
+          Message.user("Continue"),
+        ],
+        providerOptions: { reasoningEffort: "low", includeReasoning: false },
+      }),
+    )
+    expect(compiled.body).toMatchObject({ reasoning_effort: "low", include_reasoning: false })
+    expect(compiled.body.reasoning_format).toBeUndefined()
+    expect(compiled.body.messages[1]).toMatchObject({ reasoning: "Thinking", content: "Answer" })
+    expect(compiled.body.messages[1]).not.toHaveProperty("reasoning_content")
+    expect(compiled.body.messages[3]).not.toHaveProperty("reasoning")
+  }),
+)
+
+it.effect("Groq omits reasoning_format for the GPT-OSS family by default", () =>
+  Effect.gen(function* () {
+    for (const id of ["openai/gpt-oss-20b", "openai/gpt-oss-120b", "openai/gpt-oss-safeguard-20b"]) {
+      const compiled = yield* compileRequest(
+        LLM.request({ model: Groq.configure({ apiKey: "fixture" }).model(id), prompt: "Hello" }),
+      )
+      expect(compiled.body.reasoning_format).toBeUndefined()
+      expect(compiled.body.include_reasoning).toBeUndefined()
+    }
+  }),
+)
+
+it.effect("Groq validates option types", () =>
+  Effect.gen(function* () {
+    for (const providerOptions of [{ includeReasoning: "false" }, { parallelToolCalls: "false" }]) {
+      const error = yield* compileRequest(
+        LLM.request({ model: Groq.configure({ apiKey: "fixture" }).model("qwen"), prompt: "Hello", providerOptions }),
+      ).pipe(Effect.flip)
+      expect(error.reason._tag).toBe("InvalidRequest")
+    }
+  }),
+)

packages/ai/test/provider/native-providers.test.ts

@@ -2,7 +2,7 @@ import { describe, expect } from "bun:test"
 import { ConfigProvider, Effect } from "effect"
 import { HttpClientRequest } from "effect/unstable/http"
 import { LLM, Message, ToolDefinition } from "../../src/index.js"
-import { Cerebras, DeepInfra, TogetherAI } from "../../src/providers/index.js"
+import { Cerebras, DeepInfra, Groq, TogetherAI } from "../../src/providers/index.js"
 import { compileRequest } from "../../src/route/client.js"
 import { it } from "../lib/effect.js"
 import { dynamicResponse } from "../lib/http.js"
@@ -155,6 +155,12 @@ describe("native OpenAI-compatible providers", () => {
           token: "deepinfra-secret",
           url: "https://api.deepinfra.com/v1/openai/chat/completions",
         },
+        {
+          model: Groq.configure().model("llama"),
+          env: { GROQ_API_KEY: "groq-secret" },
+          token: "groq-secret",
+          url: "https://api.groq.com/openai/v1/chat/completions",
+        },
       ]
 
       yield* Effect.forEach(scenarios, (scenario) =>

packages/core/package.json

@@ -105,7 +105,6 @@
     "@ai-sdk/cohere": "3.0.27",
     "@ai-sdk/gateway": "3.0.104",
     "@ai-sdk/google-vertex": "4.0.128",
-    "@ai-sdk/groq": "3.0.31",
     "@ai-sdk/mistral": "3.0.51",
     "@ai-sdk/openai-compatible": "2.0.41",
     "@ai-sdk/perplexity": "3.0.26",

packages/core/src/aisdk-native.ts

@@ -55,6 +55,7 @@ export function map(input: MapInput): Mapping | undefined {
       }
     case "@ai-sdk/cerebras":
     case "@ai-sdk/deepinfra":
+    case "@ai-sdk/groq":
     case "@ai-sdk/togetherai":
       return {
         package: `@opencode-ai/ai/providers/${input.packageName.slice("@ai-sdk/".length)}`,

packages/core/src/model-resolver.ts

@@ -337,6 +337,7 @@ function usesAPIKeyAuth(packageName: string | undefined) {
     name === "@ai-sdk/deepinfra" ||
     name === "@ai-sdk/openai-compatible" ||
     name === "@ai-sdk/google" ||
+    name === "@ai-sdk/groq" ||
     name === "@ai-sdk/togetherai" ||
     name === "@ai-sdk/xai" ||
     name === "@openrouter/ai-sdk-provider" ||
@@ -349,6 +350,7 @@ function usesAPIKeyAuth(packageName: string | undefined) {
     name === "@opencode-ai/ai/providers/deepinfra" ||
     name === "@opencode-ai/ai/providers/openai-compatible" ||
     name === "@opencode-ai/ai/providers/google" ||
+    name === "@opencode-ai/ai/providers/groq" ||
     name === "@opencode-ai/ai/providers/togetherai" ||
     name === "@opencode-ai/ai/providers/xai" ||
     name === "@opencode-ai/ai/providers/openrouter" ||

packages/core/src/plugin/provider.ts

@@ -11,7 +11,6 @@ import { GatewayPlugin } from "./provider/gateway.js"
 import { GithubCopilotPlugin } from "./provider/github-copilot.js"
 import { GitLabPlugin } from "./provider/gitlab.js"
 import { GoogleVertexPlugin } from "./provider/google-vertex.js"
-import { GroqPlugin } from "./provider/groq.js"
 import { KiloPlugin } from "./provider/kilo.js"
 import { LLMGatewayPlugin } from "./provider/llmgateway.js"
 import { LMStudioPlugin } from "./provider/lmstudio.js"
@@ -45,7 +44,6 @@ export const ProviderPlugins: PluginInternal.InternalPlugin[] = [
   GithubCopilotPlugin,
   GitLabPlugin,
   GoogleVertexPlugin,
-  GroqPlugin,
   KiloPlugin,
   LLMGatewayPlugin,
   LMStudioPlugin,

packages/core/src/plugin/provider/groq.ts

@@ -1,10 +0,0 @@
-import { createProviderPlugin } from "./factory.js"
-
-export const GroqPlugin = createProviderPlugin({
-  id: "opencode.provider.groq",
-  package: "@ai-sdk/groq",
-  load: async (options) => {
-    const { createGroq } = await import("@ai-sdk/groq")
-    return createGroq(options)
-  },
-})

packages/core/src/provider.ts

@@ -62,6 +62,7 @@ const builtins = new Map<string, () => Promise<unknown>>([
     "@opencode-ai/ai/providers/google-vertex/messages",
     () => import("@opencode-ai/ai/providers/google-vertex/messages"),
   ],
+  ["@opencode-ai/ai/providers/groq", () => import("@opencode-ai/ai/providers/groq")],
   ["@opencode-ai/ai/providers/openai", () => import("@opencode-ai/ai/providers/openai")],
   ["@opencode-ai/ai/providers/openai/chat", () => import("@opencode-ai/ai/providers/openai/chat")],
   ["@opencode-ai/ai/providers/openai/responses", () => import("@opencode-ai/ai/providers/openai/responses")],

packages/core/test/aisdk-native.test.ts

@@ -61,8 +61,8 @@ describe("AISDKNative", () => {
     })
   })
 
-  test("maps Cerebras, DeepInfra, and Together AI settings, headers, and reasoning options to native providers", () => {
-    for (const name of ["cerebras", "deepinfra", "togetherai"]) {
+  test("maps Cerebras, DeepInfra, Groq, and Together AI settings, headers, and reasoning options to native providers", () => {
+    for (const name of ["cerebras", "deepinfra", "groq", "togetherai"]) {
       expect(
         map(`@ai-sdk/${name}`, {
           apiKey: "secret",

packages/core/test/model-resolver.test.ts

@@ -915,6 +915,12 @@ describe("ModelResolver", () => {
           { reasoning: { effort: "high" } },
           { reasoning: { effort: "high" } },
         ],
+        [
+          "@ai-sdk/groq",
+          "@opencode-ai/ai/providers/groq",
+          { reasoningEffort: "high", parallelToolCalls: false },
+          { reasoningEffort: "high", parallelToolCalls: false },
+        ],
         [
           "@ai-sdk/togetherai",
           "@opencode-ai/ai/providers/togetherai",
@@ -973,6 +979,7 @@ describe("ModelResolver", () => {
         ["@ai-sdk/google", "@opencode-ai/ai/providers/google", "api-model"],
         ["@ai-sdk/google-vertex", "@opencode-ai/ai/providers/google-vertex", "api-model"],
         ["@ai-sdk/google-vertex/anthropic", "@opencode-ai/ai/providers/google-vertex/messages", "claude-sonnet-4-6"],
+        ["@ai-sdk/groq", "@opencode-ai/ai/providers/groq", "api-model"],
         ["@ai-sdk/openai", "@opencode-ai/ai/providers/openai", "api-model"],
         ["@ai-sdk/openai-compatible", "@opencode-ai/ai/providers/openai-compatible", "api-model"],
         ["@openrouter/ai-sdk-provider", "@opencode-ai/ai/providers/openrouter", "api-model"],
@@ -1102,6 +1109,11 @@ describe("ModelResolver", () => {
       const togetherai = yield* ModelResolver.fromCatalogModel(
         model(Provider.aisdk("@ai-sdk/togetherai"), { settings: { reasoningEffort: "high" } }),
       )
+      const groq = yield* ModelResolver.fromCatalogModel(
+        model(Provider.aisdk("@ai-sdk/groq"), {
+          settings: { reasoningEffort: "high", parallelToolCalls: false },
+        }),
+      )
       const xai = yield* ModelResolver.fromCatalogModel(
         model(Provider.aisdk("@ai-sdk/xai"), { settings: { reasoningEffort: "high" } }),
       )
@@ -1132,6 +1144,10 @@ describe("ModelResolver", () => {
       expect(togetherai.route.id).toBe("togetherai-chat")
       expect(togetherai.route.defaults.providerOptions).toEqual({ reasoningEffort: "high" })
       expect(String(togetherai.provider)).toBe("test-provider")
+      expect(groq.route.id).toBe("groq-chat")
+      expect(groq.route.protocol).toBe("groq-chat")
+      expect(groq.route.defaults.providerOptions).toEqual({ reasoningEffort: "high", parallelToolCalls: false })
+      expect(String(groq.provider)).toBe("test-provider")
       expect(xai.route.id).toBe("openai-responses")
       expect(xai.route.defaults.providerOptions).toEqual({
         reasoningEffort: "high",

packages/core/test/plugin/provider-factory.test.ts

@@ -7,7 +7,6 @@ import { PluginHost } from "@opencode-ai/core/plugin/host"
 import { AlibabaPlugin } from "@opencode-ai/core/plugin/provider/alibaba"
 import { CoherePlugin } from "@opencode-ai/core/plugin/provider/cohere"
 import { GatewayPlugin } from "@opencode-ai/core/plugin/provider/gateway"
-import { GroqPlugin } from "@opencode-ai/core/plugin/provider/groq"
 import { MistralPlugin } from "@opencode-ai/core/plugin/provider/mistral"
 import { PerplexityPlugin } from "@opencode-ai/core/plugin/provider/perplexity"
 import { VenicePlugin } from "@opencode-ai/core/plugin/provider/venice"
@@ -21,7 +20,6 @@ const providers = [
   { id: "alibaba", plugin: AlibabaPlugin, package: "@ai-sdk/alibaba", provider: "alibaba.chat" },
   { id: "cohere", plugin: CoherePlugin, package: "@ai-sdk/cohere", provider: "cohere.chat" },
   { id: "gateway", plugin: GatewayPlugin, package: "@ai-sdk/gateway", provider: "gateway" },
-  { id: "groq", plugin: GroqPlugin, package: "@ai-sdk/groq", provider: "groq.chat" },
   { id: "mistral", plugin: MistralPlugin, package: "@ai-sdk/mistral", provider: "mistral.chat" },
   { id: "perplexity", plugin: PerplexityPlugin, package: "@ai-sdk/perplexity", provider: "perplexity" },
   { id: "venice", plugin: VenicePlugin, package: "venice-ai-sdk-provider", provider: "custom-provider.chat" },

packages/core/test/provider.test.ts

@@ -12,6 +12,7 @@ describe("Provider", () => {
       "@opencode-ai/ai/providers/google-vertex/chat",
       "@opencode-ai/ai/providers/google-vertex/responses",
       "@opencode-ai/ai/providers/google-vertex/messages",
+      "@opencode-ai/ai/providers/groq",
       "@opencode-ai/ai/providers/togetherai",
     ]
 

packages/tui/src/routes/session/index.tsx

@@ -3360,7 +3360,7 @@ function executeCalls(value: unknown): ExecuteCall[] {
 
 export function executeCallSummary(call: ExecuteCall) {
   const args = primitiveInputSummary(call.input ?? {}).replace(/\s+/g, " ")
-  return `↳ ${call.tool}${call.status === "error" ? " (failed)" : ""}${args ? ` ${args}` : ""}`
+  return `${call.tool}${args ? ` ${args}` : ""}`
 }
 
 function ExecuteCallView(props: { call: Accessor<ExecuteCall> }) {
@@ -3370,11 +3370,16 @@ function ExecuteCallView(props: { call: Accessor<ExecuteCall> }) {
   const [hover, setHover] = createSignal(false)
   const input = createMemo(() => Object.entries(props.call().input ?? {}))
   const expandable = createMemo(() => input().length > 0)
-  const title = createMemo(() => `↳ ${props.call().tool}${props.call().status === "error" ? " (failed)" : ""}`)
+  const expandedColor = createMemo(() => theme.raise(theme.text.subdued))
+  const color = createMemo(() => {
+    if (props.call().status === "error") return theme.text.feedback.error.default
+    if (hover()) return theme.text.default
+    return expanded() ? expandedColor() : theme.text.subdued
+  })
 
   return (
     <box
-      paddingLeft={3 + INLINE_TOOL_ICON_WIDTH}
+      paddingLeft={3}
       onMouseOver={() => expandable() && setHover(true)}
       onMouseOut={() => setHover(false)}
       onMouseUp={() => {
@@ -3382,21 +3387,16 @@ function ExecuteCallView(props: { call: Accessor<ExecuteCall> }) {
         setExpanded((value) => !value)
       }}
     >
-      <text
-        wrapMode="none"
-        truncate
-        fg={
-          props.call().status === "error"
-            ? theme.text.feedback.error.default
-            : hover()
-              ? theme.text.default
-              : theme.text.subdued
-        }
-      >
-        {expanded() ? title() : executeCallSummary(props.call())}
-      </text>
+      <box flexDirection="row">
+        <box width={INLINE_TOOL_ICON_WIDTH} flexShrink={0}>
+          <text fg={color()}>{props.call().status === "error" ? "✗" : "›"}</text>
+        </box>
+        <text flexGrow={1} wrapMode="none" truncate fg={color()}>
+          {expanded() ? props.call().tool : executeCallSummary(props.call())}
+        </text>
+      </box>
       <Show when={expanded()}>
-        <box paddingLeft={2}>
+        <box paddingLeft={1} border={["left"]} borderColor={expandedColor()}>
           <For each={input()}>
             {([key, value]) => (
               <box flexDirection="row">

packages/tui/test/cli/tui/inline-tool-wrap-snapshot.test.tsx

@@ -191,13 +191,13 @@ describe("TUI inline tool wrapping", () => {
         status: "completed",
         input: { sessionID: "ses_example", notify: true },
       }),
-    ).toBe("↳ session.prompt [sessionID=ses_example, notify=true]")
+    ).toBe("session.prompt [sessionID=ses_example, notify=true]")
     expect(executeCallSummary({ tool: "session.get", status: "error", input: { nested: { hidden: true } } })).toBe(
-      "↳ session.get (failed)",
+      "session.get",
     )
     expect(
       executeCallSummary({ tool: "session.prompt", status: "completed", input: { text: "first line\nsecond line" } }),
-    ).toBe("↳ session.prompt [text=first line second line]")
+    ).toBe("session.prompt [text=first line second line]")
   })
 
   test("summarizes generic tool arguments on one line", () => {