Repository navigation
Muse Spark 1.3: resuming a session after idle fails with reasoning encrypted_content not issued to this caller #48915
Description
Description
Resuming an old session on opencode/muse-spark-1.3-contributor-free fails immediately with:
Error from provider (Console): Upstream request failed: [invalid_request_error] reasoning `encrypted_content` was not issued to this caller
A new session with the same model works. Continuing a session after idle, or after using another model then switching back to Muse 1.3, always fails until /new.
OpenCode stores metadata.openai.reasoningEncryptedContent + itemId on reasoning parts and replays them. Those blobs are bound to the Zen/Console caller and expire. HTTP 400 is treated as fatal, so the session stays bricked.
Related: #48741 (image/tool call), #48805 (mid-session model switch). This report is specifically resume-after-idle on 1.18.30.
Plugins
opencode-pty, opencode-websearch-cited@1.2.0
OpenCode version
1.18.30
Steps to reproduce
- Start a session with
opencode/muse-spark-1.3-contributor-freeand send a few messages (or mix in another model, then switch back). - Leave the session idle for a while, or fully restart OpenCode.
- Reopen that same session and send another message with Muse 1.3.
Expected: request succeeds, history stays.
Actual: every retry returns the encrypted_content caller error.
Log:
timestamp=2026-09-14T03:07:17.902Z level=ERROR message="stream error" providerID=opencode modelID=muse-spark-1.3-contributor-free session.id=ses_f6b39b16bffeVTjvIwrR2SLez7 error.error="AI_APICallError: Error from provider (Console): Upstream request failed: [invalid_request_error] reasoning `encrypted_content` was not issued to this caller"
Stored part shape (reasoning, part table):
{
"type": "reasoning",
"text": "...",
"metadata": {
"openai": {
"itemId": "rs_...",
"reasoningEncryptedContent": "..."
}
}
}Please drop reasoningEncryptedContent / reasoning itemId when the provider rejects them, persist the cleaned parts, and retry. Same class of recovery as the Claude stale-thinking handling.
Screenshot and/or share link
N/A. Error is from the Console provider before any stream output.
Operating System
macOS (darwin arm64)
Terminal
OpenCode TUI
Activity
github-actions commentedon Sep 14, 2026
This issue might be a duplicate of existing issues. Please check:
reasoningencrypted_contentwas not issued to this callererror onmuse-spark-1.3-contributor-free, triggered by model switching in the same session — very similar root cause (stale encrypted reasoning blobs replayed across caller boundaries).invalid_request_error— same class of replay-after-restart failure.If your issue is specifically about the idle/restart resume path (rather than mid-session model switching), it may still warrant a separate fix, but please check those issues first and add details there if they cover your case.
1837620622 commentedon Sep 14, 2026
Fix in #48918
Joelincn commentedon Sep 14, 2026
Same error reproduced via a third trigger — egress rotation, no model switch, no idle needed. Details posted in #48805 (comment above): load-balanced egress across 3 Singapore ASNs,
socket closedthen ~10s later theencrypted_contentcaller 400, session permanently bricked. Endorsing the drop-and-retry proposal here.Vargas876 commentedon Sep 15, 2026
Same error on OpenCode Desktop (Windows 11), confirming a fourth trigger: it happens mid-session without idle and without model switching, and after several turns escalates until the session is bricked (HTTP 400, \isRetryable: false).
Additional data points from the desktop app event store (\opencode.global.dat, 55 occurrences across 00:23-23:47 local):
s_...) +
easoningEncryptedContent, \content: null, \summary: []\ (opaque blob only, matching Zen gateway does not forward reasoning/thinking stream for Muse Spark 1.2 #43584).
Also tested and NOT effective: upgrading desktop 1.18.21 -> 1.18.31, model options \store:false\ + \include:[reasoning.encrypted_content],
easoningEffort:none\ +
easoningSummary:auto, and deleting all 160+
easoningEncryptedContent\ blobs from the DB (they regenerate each turn, so the error returns). Fresh /new\ session is the only working workaround, confirming history replay is the trigger.
Endorsing the fix in #48918 (drop
easoningEncryptedContent/itemId and retry). Our case is fully covered by the same recovery path.
thdxr commentedon Oct 11, 2026
Thanks for the report! OpenCode 2.0 is now out and development is focused on V2, so we're closing issues filed against OpenCode 1. Please upgrade (
opencode upgrade --major, details: https://anoma.ly/notes/opencode-2-0), and if this is still relevant on 2.0, reopen this issue or open a new one with your 2.0 version.— from 𝕺𝖕𝖊𝖓𝕮𝖔𝖉𝖊