fix(opencode): route Muse Spark through the Responses API

muse-spark-1.2-contributor-free returned HTTP 500 on /zen/v1/chat/completions.
The model is only served by /zen/v1/responses, so route it there via a per-model
targetFormat and normalize the Chat fields the Responses API rejects
(max_tokens -> max_output_tokens, reasoning_effort -> reasoning{effort,summary}),
clamping max/ultra down to the highest effort the model accepts (xhigh).

Routing stays per-model: the other free models (big-pickle, hy3-free, mimo,
nemotron, laguna) are not served by /responses and keep /chat/completions.
This commit is contained in:
anojndr
2026-08-28 11:32:56 +07:00
parent 14401c433c
commit ab044e6d6d
6 changed files with 203 additions and 13 deletions

View File

@@ -9,6 +9,7 @@ import { DEFAULT_MAX_TOKENS, DEFAULT_MIN_TOKENS } from "../../open-sse/config/ru
import mimoFree from "../../open-sse/providers/registry/mimo-free.js";
import opencode from "../../open-sse/providers/registry/opencode.js";
import antigravity from "../../open-sse/providers/registry/antigravity.js";
import { OpenCodeExecutor } from "../../open-sse/executors/opencode.js";
describe("compat base URLs / version", () => {
it("OPENAI_COMPAT_BASE", () => {
@@ -46,3 +47,36 @@ describe("antigravity retry (intentional change: 429=6, 503=3)", () => {
expect(antigravity.transport.retry["503"].attempts).toBe(3);
});
});
describe("OpenCode Free endpoint routing", () => {
const MUSE = "muse-spark-1.2-contributor-free";
it("declares the Responses format only on the Muse Spark model", () => {
expect(opencode.transport.format).toBeUndefined();
const muse = opencode.models.find((m) => m.id === MUSE);
expect(muse?.targetFormat).toBe("openai-responses");
});
it("routes Muse Spark to /responses and every other model to /chat/completions", () => {
const executor = new OpenCodeExecutor();
expect(executor.buildUrl(MUSE)).toBe("https://opencode.ai/zen/v1/responses");
expect(executor.buildUrl(`${MUSE}(xhigh)`)).toBe("https://opencode.ai/zen/v1/responses");
expect(executor.buildUrl("big-pickle")).toBe("https://opencode.ai/zen/v1/chat/completions");
expect(executor.buildUrl("hy3-free")).toBe("https://opencode.ai/zen/v1/chat/completions");
});
it("normalizes Chat token/thinking fields only for the Responses model", () => {
const executor = new OpenCodeExecutor();
const muse = { max_tokens: 4096, reasoning_effort: "high" };
executor.transformRequest(MUSE, muse, true, {});
expect(muse.max_output_tokens).toBe(4096);
expect(muse.max_tokens).toBeUndefined();
expect(muse.reasoning).toEqual({ effort: "high", summary: "auto" });
const chat = { max_tokens: 4096, reasoning_effort: "high" };
executor.transformRequest("big-pickle", chat, true, {});
expect(chat.max_tokens).toBe(4096);
expect(chat.max_output_tokens).toBeUndefined();
expect(chat.reasoning_effort).toBe("high");
});
});

View File

@@ -0,0 +1,94 @@
import { describe, expect, it } from "vitest";
import { getCapabilitiesForModel } from "../../open-sse/providers/capabilities.js";
import { PROVIDER_MODELS } from "../../open-sse/config/providerModels.js";
import { getThinkingLevels } from "../../open-sse/providers/thinkingLevels.js";
import { FORMATS } from "../../open-sse/translator/formats.js";
import { OpenCodeExecutor } from "../../open-sse/executors/opencode.js";
import "../translator/registerAll.js";
import { translateRequest } from "../../open-sse/translator/index.js";
const MODEL = "muse-spark-1.2-contributor-free";
const PROVIDER = "opencode";
const input = [{
type: "message",
role: "user",
content: [{ type: "input_text", text: "Think, then answer: 2 + 2?" }],
}];
describe("OpenCode Free Muse Spark thinking", () => {
it("advertises reasoning and the requested model limits", () => {
expect(PROVIDER_MODELS.oc?.some((model) => model.id === MODEL)).toBe(true);
expect(getCapabilitiesForModel(PROVIDER, MODEL)).toMatchObject({
reasoning: true,
thinkingFormat: "openai",
contextWindow: 1048576,
maxOutput: 131072,
});
expect(getCapabilitiesForModel(PROVIDER, `oc/${MODEL}`)).toMatchObject({
reasoning: true,
contextWindow: 1048576,
maxOutput: 131072,
});
expect(getThinkingLevels(PROVIDER, MODEL)).toEqual([
"none",
"minimal",
"low",
"medium",
"high",
"xhigh",
]);
});
it("clamps max to xhigh and emits the Responses reasoning shape", () => {
const body = {
input,
reasoning: { effort: "max" },
max_tokens: 131072,
};
const out = new OpenCodeExecutor().transformRequest(MODEL, body, true, {
connectionId: "opencode-muse-spark-test",
});
expect(out.reasoning).toEqual({ effort: "xhigh", summary: "auto" });
expect(out.reasoning_effort).toBeUndefined();
expect(out.max_output_tokens).toBe(131072);
expect(out.max_tokens).toBeUndefined();
});
it("leaves the other free models on Chat Completions", () => {
const executor = new OpenCodeExecutor();
const body = { messages: [{ role: "user", content: "hi" }], max_tokens: 1024 };
executor.transformRequest("big-pickle", body, true, {});
expect(executor.buildUrl("big-pickle")).toBe("https://opencode.ai/zen/v1/chat/completions");
expect(body.max_tokens).toBe(1024);
expect(body.max_output_tokens).toBeUndefined();
});
it("translates Chat Completions max thinking into a Responses request", () => {
const body = {
model: `oc/${MODEL}`,
messages: [{ role: "user", content: "Think, then answer: 2 + 2?" }],
reasoning_effort: "max",
max_tokens: 131072,
};
const translated = translateRequest(
FORMATS.OPENAI,
FORMATS.OPENAI_RESPONSES,
MODEL,
body,
true,
{},
PROVIDER,
);
const out = new OpenCodeExecutor().transformRequest(MODEL, translated, true, {
connectionId: "opencode-muse-spark-translation-test",
});
expect(out.reasoning).toEqual({ effort: "xhigh", summary: "auto" });
expect(out.max_output_tokens).toBe(131072);
expect(out.max_tokens).toBeUndefined();
});
});