lmin/zmo-cli
1
0
Fork 0
Code Issues Pull Requests Actions 6 Packages Projects Releases Wiki Activity

Refactor LLM route-first provider API (#28523)

Browse Source
This commit is contained in:
Kit Langton
2026-05-20 20:15:52 -04:00
committed by GitHub
parent 5381795844
commit 41f6daf96a
87 changed files with 2450 additions and 1520 deletions
Download Patch File Download Diff File Expand all files Collapse all files

17
packages/opencode/test/server/httpapi-event-diagnostics.test.ts
View File

@@ -56,11 +56,11 @@ afterEach(async () => {
})
const inApp = <A, E>(eff: Effect.Effect<A, E, AppServices>) =>
Effect.flatMap(InstanceRef, (ctx) =>
ctx
? Effect.promise(() => AppRuntime.runPromise(eff.pipe(Effect.provideService(InstanceRef, ctx))))
: Effect.die("InstanceRef not provided in test scope"),
)
Effect.gen(function* () {
const ctx = yield* InstanceRef
if (!ctx) return yield* Effect.die("InstanceRef not provided in test scope")
return yield* Effect.promise(() => AppRuntime.runPromise(eff.pipe(Effect.provideService(InstanceRef, ctx))))
})
const publishConnected = inApp(Bus.Service.use((svc) => svc.publish(ServerEvent.Connected, {})))
@@ -112,7 +112,7 @@ const readNextEvent = (reader: ReadableStreamDefaultReader<Uint8Array>) =>
if (result.done || !result.value) return Effect.fail(new Error("event stream closed"))
const frames = decodeFrame(result.value)
if (frames.length === 0) return Effect.fail(new Error("empty SSE frame"))
return Effect.succeed(frames[0]!)
return Effect.succeed(frames[0])
}),
)
@@ -186,8 +186,7 @@ describe("/event SSE delivery diagnostics", () => {
const collected = yield* collectUntilEvent(reader, isPartUpdated)
const updated = collected.find(isPartUpdated)
expect(updated).toBeDefined()
expect((updated as SseEvent).properties.part.id).toBe(partID)
expect(updated?.properties.part.id).toBe(partID)
}),
{ git: true, config: { formatter: false, lsp: false } },
)
@@ -217,7 +216,7 @@ describe("/event SSE delivery diagnostics", () => {
}),
)
expect(event.type).toBe(MessageV2.Event.PartUpdated.type)
expect((event.properties as { part: { id: string } }).part.id).toBe(partID)
expect(event.properties).toMatchObject({ part: { id: partID } })
}),
{ git: true, config: { formatter: false, lsp: false } },
)

4
packages/opencode/test/session/llm-native-recorded.test.ts
View File

@@ -13,7 +13,7 @@ import { Provider } from "@/provider/provider"
import { ModelID, ProviderID } from "@/provider/schema"
import { Filesystem } from "@/util/filesystem"
import { LLMEvent, LLMResponse } from "@opencode-ai/llm"
import { LLMClient, RequestExecutor } from "@opencode-ai/llm/route"
import { LLMClient, RequestExecutor, WebSocketExecutor } from "@opencode-ai/llm/route"
import { RuntimeFlags } from "@/effect/runtime-flags"
import type { Agent } from "../../src/agent/agent"
import { LLM } from "../../src/session/llm"
@@ -137,7 +137,7 @@ async function loadFixture(providerID: string, modelID: string) {
function recordedNativeLLMLayer(spec: ProviderSpec) {
// Only the HTTP client is recorded; RequestExecutor and the opencode LLM stack remain real.
const recordedClient = LLMClient.layer.pipe(
Layer.provide(RequestExecutor.layer),
Layer.provide(Layer.mergeAll(RequestExecutor.layer, WebSocketExecutor.layer)),
Layer.provide(
HttpRecorder.recordingLayer(spec.cassette, {
mode: shouldRecord ? "record" : "replay",

105
packages/opencode/test/session/llm-native.test.ts
View File

@@ -1,8 +1,8 @@
import { describe, expect, test } from "bun:test"
import { ToolFailure } from "@opencode-ai/llm"
import { LLMClient, RequestExecutor } from "@opencode-ai/llm/route"
import { LLMClient, RequestExecutor, WebSocketExecutor } from "@opencode-ai/llm/route"
import { jsonSchema, tool, type ModelMessage } from "ai"
import { Effect } from "effect"
import { Effect, Layer } from "effect"
import { LLMNative } from "@/session/llm/native-request"
import { LLMNativeRuntime } from "@/session/llm/native-runtime"
import type { Provider } from "@/provider/provider"
@@ -138,16 +138,16 @@ describe("session.llm-native.request", () => {
expect(request.model).toMatchObject({
id: "gpt-5-mini",
provider: "openai",
route: "openai-responses",
baseURL: "https://api.openai.com/v1",
headers: {
"x-model": "model-header",
"x-request": "request-header",
},
limits: {
context: 128_000,
output: 32_000,
},
route: { id: "openai-responses" },
})
expect(request.model.route.endpoint.baseURL).toBe("https://api.openai.com/v1")
expect(request.model.route.defaults.headers).toEqual({
"x-model": "model-header",
"x-request": "request-header",
})
expect(request.model.route.defaults.limits).toMatchObject({
context: 128_000,
output: 32_000,
})
expect(request.system).toEqual([
{ type: "text", text: "agent system" },
@@ -211,29 +211,50 @@ describe("session.llm-native.request", () => {
])
})
test("selects native routes from existing provider packages", () => {
expect(
LLMNative.model({ ...baseModel, api: { ...baseModel.api, url: "", npm: "@ai-sdk/anthropic" } }),
).toMatchObject({
route: "anthropic-messages",
baseURL: "https://api.anthropic.com/v1",
test("selects native request routes for provider packages", () => {
const openai = LLMNative.model({
model: { ...baseModel, api: { ...baseModel.api, url: "", npm: "@ai-sdk/openai" } },
apiKey: "test-key",
messages: [],
})
expect(LLMNative.model({ ...baseModel, api: { ...baseModel.api, url: "", npm: "@ai-sdk/google" } })).toMatchObject({
route: "gemini",
baseURL: "https://generativelanguage.googleapis.com/v1beta",
expect(openai.route.id).toBe("openai-responses")
expect(openai.route.endpoint.baseURL).toBe("https://api.openai.com/v1")
const anthropic = LLMNative.model({
model: { ...baseModel, api: { ...baseModel.api, url: "", npm: "@ai-sdk/anthropic" } },
apiKey: "test-key",
messages: [],
})
expect(
LLMNative.model({ ...baseModel, api: { ...baseModel.api, npm: "@ai-sdk/openai-compatible" } }),
).toMatchObject({
route: "openai-compatible-chat",
baseURL: "https://api.openai.com/v1",
expect(anthropic.route.id).toBe("anthropic-messages")
expect(anthropic.route.endpoint.baseURL).toBe("https://api.anthropic.com/v1")
const google = LLMNative.model({
model: { ...baseModel, api: { ...baseModel.api, url: "", npm: "@ai-sdk/google" } },
apiKey: "test-key",
messages: [],
})
expect(
LLMNative.model({ ...baseModel, api: { ...baseModel.api, url: "", npm: "@openrouter/ai-sdk-provider" } }),
).toMatchObject({
route: "openrouter",
baseURL: "https://openrouter.ai/api/v1",
expect(google.route.id).toBe("gemini")
expect(google.route.endpoint.baseURL).toBe("https://generativelanguage.googleapis.com/v1beta")
const compatible = LLMNative.model({
model: {
...baseModel,
providerID: ProviderID.make("opencode"),
api: { ...baseModel.api, url: "https://ai.example.test/v1", npm: "@ai-sdk/openai-compatible" },
},
apiKey: "test-key",
messages: [],
})
expect(compatible.route.id).toBe("openai-compatible-chat")
expect(compatible.route.endpoint.baseURL).toBe("https://ai.example.test/v1")
const openrouter = LLMNative.model({
model: { ...baseModel, api: { ...baseModel.api, url: "", npm: "@openrouter/ai-sdk-provider" } },
apiKey: "test-key",
messages: [],
})
expect(openrouter.route.id).toBe("openrouter")
expect(openrouter.route.endpoint.baseURL).toBe("https://openrouter.ai/api/v1")
})
test("fails fast for unsupported provider packages", () => {
@@ -260,6 +281,20 @@ describe("session.llm-native.request", () => {
type: "supported",
apiKey: "test-openai-key",
})
expect(
LLMNativeRuntime.status({
model: {
...baseModel,
providerID: ProviderID.make("opencode"),
api: { ...baseModel.api, npm: "@ai-sdk/openai-compatible" },
},
provider: { ...providerInfo, id: ProviderID.make("opencode") },
auth: undefined,
}),
).toMatchObject({
type: "supported",
apiKey: "test-openai-key",
})
expect(
LLMNativeRuntime.status({
model: { ...baseModel, providerID: ProviderID.make("google") },
@@ -281,7 +316,7 @@ describe("session.llm-native.request", () => {
provider: providerInfo,
auth: undefined,
}),
).toEqual({ type: "unsupported", reason: "provider package is not OpenAI or Anthropic" })
).toEqual({ type: "unsupported", reason: "provider package is not OpenAI, OpenAI-compatible, or Anthropic" })
expect(
LLMNativeRuntime.status({
@@ -382,12 +417,16 @@ describe("session.llm-native.request", () => {
LLMClient.prepare(
LLMNative.request({
model: baseModel,
apiKey: "test-openai-key",
messages: [{ role: "user", content: "hello" }],
providerOptions: { openai: { store: false } },
maxOutputTokens: 512,
headers: { "x-request": "request-header" },
}),
).pipe(Effect.provide(LLMClient.layer), Effect.provide(RequestExecutor.defaultLayer)),
).pipe(
Effect.provide(LLMClient.layer),
Effect.provide(Layer.mergeAll(RequestExecutor.defaultLayer, WebSocketExecutor.layer)),
),
)
expect(prepared).toMatchObject({

81
packages/opencode/test/session/llm.test.ts
View File

@@ -8,7 +8,7 @@ import { makeRuntime } from "../../src/effect/run-service"
import { InstanceRef } from "../../src/effect/instance-ref"
import { LLM } from "../../src/session/llm"
import type { InstanceContext } from "../../src/project/instance-context"
import { LLMClient, RequestExecutor } from "@opencode-ai/llm/route"
import { LLMClient, RequestExecutor, WebSocketExecutor } from "@opencode-ai/llm/route"
import { Auth } from "@/auth"
import { Config } from "@/config/config"
import { Provider } from "@/provider/provider"
@@ -82,7 +82,7 @@ function llmLayerWithExecutor(executor: Layer.Layer<RequestExecutor.Service>, fl
Layer.provide(Config.defaultLayer),
Layer.provide(Provider.defaultLayer),
Layer.provide(Plugin.defaultLayer),
Layer.provide(LLMClient.layer.pipe(Layer.provide(executor))),
Layer.provide(LLMClient.layer.pipe(Layer.provide(Layer.mergeAll(executor, WebSocketExecutor.layer)))),
Layer.provide(RuntimeFlags.layer(flags)),
)
}
@@ -1975,54 +1975,45 @@ describe("session.llm.stream", () => {
const body = capture.body
expect(capture.url.pathname.endsWith("/messages")).toBe(true)
expect(body.messages).toStrictEqual([
const messages = body.messages as Array<{ role: string; content: Array<Record<string, unknown>> }>
expect(messages[0]?.role).toBe("user")
expect(messages[0]?.content[0]).toMatchObject({
type: "text",
text: "Can you check whether there are any PDF files in my home directory?",
})
expect(messages.some((message) => message.content.some((part) => "cache_control" in part))).toBe(true)
const toolUseIndex = messages.findIndex((message) => message.content.some((part) => part.type === "tool_use"))
expect(toolUseIndex).toBeGreaterThan(0)
expect(messages[toolUseIndex].role).toBe("assistant")
expect(messages[toolUseIndex].content.filter((part) => part.type === "tool_use")).toMatchObject([
{
role: "user",
content: [{ type: "text", text: "Can you check whether there are any PDF files in my home directory?" }],
type: "tool_use",
id: "toolu_01N8mDEzG8DSTs7UPHFtmgCT",
name: "read",
input: { filePath: "/root" },
},
{
role: "assistant",
content: [
{
type: "text",
text: "I checked your home directory and looked for PDF files.",
},
{
type: "tool_use",
id: "toolu_01N8mDEzG8DSTs7UPHFtmgCT",
name: "read",
input: { filePath: "/root" },
},
{
type: "tool_use",
id: "toolu_01APxrADs7VozN8uWzw9WwHr",
name: "glob",
input: { pattern: "**/*.pdf", path: "/root" },
cache_control: {
type: "ephemeral",
},
},
],
},
{
role: "user",
content: [
{
type: "tool_result",
tool_use_id: "toolu_01N8mDEzG8DSTs7UPHFtmgCT",
content: "<path>/root</path>",
},
{
type: "tool_result",
tool_use_id: "toolu_01APxrADs7VozN8uWzw9WwHr",
content: "No files found",
cache_control: {
type: "ephemeral",
},
},
],
type: "tool_use",
id: "toolu_01APxrADs7VozN8uWzw9WwHr",
name: "glob",
input: { pattern: "**/*.pdf", path: "/root" },
},
])
expect(messages[toolUseIndex + 1]).toMatchObject({
role: "user",
content: [
{
type: "tool_result",
tool_use_id: "toolu_01N8mDEzG8DSTs7UPHFtmgCT",
content: "<path>/root</path>",
},
{
type: "tool_result",
tool_use_id: "toolu_01APxrADs7VozN8uWzw9WwHr",
content: "No files found",
},
],
})
},
})
})