[BUG] Muse Spark 1.2 model not working

[harshmann10]

🐛 Muse Spark 1.2 — API 400 Error on Every Request

🐛 What happened?

Muse Spark 1.2 (muse-spark-1.2-contributor) fails on every request with HTTP 400. The API returns an empty assistant response with finish_reason: null, causing the extension to throw OpenCodeRequestError. The model appears to reject the request payload (990KB) without providing a meaningful error message.

🔄 Steps to reproduce

1. Set provider to OpenCode Go
2. Select model muse-spark-1.2-contributor
3. Send any prompt (even a simple one like "hello")
4. Request fails immediately with: Sorry, your request failed. Please try again.

✅ What did you expect?

The model should respond with a completion, or return a descriptive error message explaining why the request was rejected (e.g., payload too large, invalid parameters, model unavailable).

🖥️ Environment

• Extension version: 0.6.0
• VS Code version: (please fill in — Help → About)
• Provider: OpenCode Go
• Model: muse-spark-1.2-contributor

📸 Logs or screenshots

Error 1 (user-facing):

Sorry, your request failed. Please try again.

Client Request Id: f84dae2f-0b63-49cd-a545-60aac4b4b2a0

Reason: OpenCode Go API request failed (400) model=muse-spark-1.2-contributor payloadBytes=990778: {"id":"chatcmpl_oikufvx92vl","object":"chat.completion","created":1787146321,"model":"muse-spark-1.2-contributor","choices":[{"index":0,"message":{"role":"assistant"},"finish_reason":null}]}

Error 2 (console):

OpenCode Go API request failed (HTTP 400) for muse-spark-1.2-contributor: {"id":"chatcmpl_oikufvx92vl","object":"chat.completion","created":1787146321,"model":"muse-spark-1.2-contributor","choices":[{"index":0,"message":{"role":"assistant"},"finish_reason":null}]}

Stack trace:

at buildOpenCodeRequestError (c:\Users\Harsh\.vscode\extensions\ltmoerdani.opencode-copilot-chat-0.6.0\out\errors.js:39:12)
at streamOpenCodeResponse (c:\Users\Harsh\.vscode\extensions\ltmoerdani.opencode-copilot-chat-0.6.0\out\streaming.js:360:73)
at process.processTicksAndRejections (node:internal/process/task_queues:104:5)
at async streamChatCompletions (c:\Users\Harsh\.vscode\extensions\ltmoerdani.opencode-copilot-chat-0.6.0\out\streaming.js:70:5)
at async OpenCodeProvider.provideLanguageModelChatResponse (c:\Users\Harsh\.vscode\extensions\ltmoerdani.opencode-copilot-chat-0.6.0\out\extension.js:2702:13)

[Fahad090NP]

Research note (geographic restriction) — sharing for visibility, not confident this is the cause:

Meta restricts Muse Spark 1.2 / the discounted muse-spark-1.2-contributor tier geographically:

• Fully blocked (all tiers): Afghanistan, Belarus, Cambodia, China (incl. HK & Macau), Cuba, Eritrea, Ethiopia, Haiti, Iran, Iraq, Libya, Myanmar, Nicaragua, North Korea, Pakistan, Russia, Somalia, Syria, Ukraine (Crimea/Donetsk/Luhansk), Venezuela, Western Sahara.
• The discounted contributor tier is additionally blocked in: Australia, Brazil, Canada, EU/EEA, South Korea, United Kingdom (data-privacy / GDPR-style reasons).

Caveat: the reporter here appears to be in India, which is not on either blocked list, so a hard geo-block does not obviously explain the 400. I'm not confident the geo restriction is what's happening here.

More likely hypothesis (worth testing first): the payloadBytes=990778 in the error is the uncompacted conversation history, not a model quirk. The reporter says it fails even on a simple "hello", but if an existing long conversation is resumed, the full history is sent every turn. Notably, running Compact Conversation then sending one message works again in most cases — that strongly points to an oversized-payload rejection rather than a geo block. PR #167 adds conversation-history trimming to fit the model context window, which should mitigate this 990 KB payload case. If the error persists after that lands (or on a brand-new empty conversation), then a geo/account block becomes the leading theory and we can dig into the gateway response headers.

Would be great if the reporter can confirm: does it fail on a fresh, empty chat with no prior history, or only when resuming an existing long conversation?

[Fahad090NP]

Oh Broter @harshmann10 go visit Meta Model API according to OpenCode's Docs, this page will tell you if the model is available for your country or not. In my case, in Pakistan, it is not available. May be because of the text written on our passport cover 😜.

[harshmann10]

@Fahad090NP ohh srry I did not closed on time as this issue previously was due to opencode own endpoint issue but from morning they have fixed it and it's working without any issue in copilot

[Barragek0]

I'm not sure how its working for you, I still get the 400 error on main, but it works with the pull request, with some spacing issues and no in-line reasoning.

Edit: The issue is that it was using the wrong transport, it uses the Responses API and not chat-completions. The registry had no muse-* entry so it fell through to default which routes the request to chat-completions. After fixing that it still failed with 400 indicating a 64 character limit for tool names.

[harshmann10]

@Barragek0 well its weird things as its working on my copilot pretty good without any issue from the morning without any chnage with only one issue as in very msg response it add this line "As GitHub Copilot using OpenCode Go / Muse Spark 1.2 Contributor," non stop on all the reponoses i do not know why as its the only model that do it

[harshmann10]

i am facing the same issue again but on regular vs code as previously i have tested it on vs code insider and its working there correctly even now (do not how?) but in regular vs code i am facing the same issue its not working on both go mode and zen free model

Reason: OpenCode Go API request failed (400) model=muse-spark-1.2-contributor payloadBytes=766810: {"id":"chatcmpl_x0i3apms1pk","object":"chat.completion","created":1787257535,"model":"muse-spark-1.2-contributor","choices":[{"index":0,"message":{"role":"assistant"},"finish_reason":null}]}: OpenCodeRequestError: OpenCode Go API request failed (400) model=muse-spark-1.2-contributor payloadBytes=766810: {"id":"chatcmpl_x0i3apms1pk","object":"chat.completion","created":1787257535,"model":"muse-spark-1.2-contributor","choices":[{"index":0,"message":{"role":"assistant"},"finish_reason":null}]} at buildOpenCodeRequestError (c:\Users\Harsh.vscode\extensions\ltmoerdani.opencode-copilot-chat-0.6.0\out\errors.js:39:12) at streamOpenCodeResponse (c:\Users\Harsh.vscode\extensions\ltmoerdani.opencode-copilot-chat-0.6.0\out\streaming.js:360:73) at process.processTicksAndRejections (node:internal/process/task_queues:104:5) at async streamChatCompletions (c:\Users\Harsh.vscode\extensions\ltmoerdani.opencode-copilot-chat-0.6.0\out\streaming.js:70:5) at async OpenCodeProvider.provideLanguageModelChatResponse (c:\Users\Harsh.vscode\extensions\ltmoerdani.opencode-copilot-chat-0.6.0\out\extension.js:2702:13)

[harshmann10]

@ltmoerdani @Barragek0 Latest 0.7.0 update broke the VS Code and VS Code insider working with muse model

OpenCode Zen stopped sending data before the response was complete (the connection closed unexpectedly). Your message may be cut off — try sending it again; a single resend usually succeeds. If this keeps happening, check your connection, VPN, or firewall.

Sorry, your request failed. Please try again. Client Request Id: 81a8c4bf-2d26-4300-95ee-0fdc531382d8 Reason: OpenCode Zen response stream ended before completion (no [DONE] or finish_reason after 212970 bytes / 12 events, request req-772fbadf).: OpenCodeRequestError: OpenCode Zen response stream ended before completion (no [DONE] or finish_reason after 212970 bytes / 12 events, request req-772fbadf). at streamOpenCodeResponse (c:\Users\Harsh.vscode\extensions\ltmoerdani.opencode-copilot-chat-0.7.0\out\transports\engine.js:395:34) at process.processTicksAndRejections (node:internal/process/task_queues:104:5) at async streamResponsesApi (c:\Users\Harsh.vscode\extensions\ltmoerdani.opencode-copilot-chat-0.7.0\out\transports\responses.js:14:9) at async OpenCodeProvider.provideLanguageModelChatResponse (c:\Users\Harsh.vscode\extensions\ltmoerdani.opencode-copilot-chat-0.7.0\out\provider\OpenCodeProvider.js:328:17)

if I downgrade to previous extension version its working without any issue but after installing latest version its shows this error after model have answered the query every time

[Barragek0]

Closed by #187