You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
{{ message }}
Repository navigation
[regression 2026-09-15] muse-spark-1.3-contributor-free: encrypted_content was not issued to this caller en chats nuevos de 2-3 mensajes #49188
Error from provider (Console): Upstream request failed: [invalid_request_error] reasoning encrypted_content was not issued to this caller
Repro
Abrir conversación 100% nueva, sin --continue / --resume / compact
Hacer 1 pregunta simple (ej: por qué sale este error)
Hacer 2-3 preguntas más normales, sin pedir tools
Falla igual. Cada reintento en la misma sesión vuelve a fallar. Solo funciona abrir otra sesión nueva, que también falla a los 2-3 mensajes.
Antes vs hoy
Antes: solo pasaba en chats muy extensos, lo entendía e iniciaba uno nuevo.
Hoy 2026-09-15: pasa en chats nuevos con 2-3 preguntas. No uso compact salvo para pruebas. No cambié modelo, key, cuenta ni proyecto.
Working dir vacío: C:\Users\desarrollo4\Documents\Default Project, no es git repo.
opencode --version local no disponible en este entorno, el error sale del LLM antes de ejecutar shell.
Esperado
Sesión corta sin tools no debería devolver invalid_request_error de reasoning encriptado.
Sospecha
Validación upstream más estricta de reasoning.encrypted_content en el gateway Console para este modelo free, similar a lo visto en @ai-sdk/azure Responses API con store:false (issue #29879, PRs #28678 / #29000). El bloque encriptado devuelto en turno N es rechazado al reenviarlo en N+1.
Plugins
No response
OpenCode version
No response
Steps to reproduce
No response
Screenshot and/or share link
No response
Operating System
No response
Terminal
No response
Reacted by Yan Henridev, Kike Garcia and npetroff24
Your report specifically highlights that this is now occurring in very short new sessions (2-3 messages) as of 2026-09-15, which could indicate a regression worth tracking separately. The root cause analysis you provided (encrypted blob from turn N being rejected at turn N+1) matches the discussion in #48915.
My earlier comment recommended github:arfoux/rs-guard, but that plugin is OpenCode V2 only (@opencode/plugin v2 API: context/compaction/generate + retry hooks for stateless replay and retry-without-reasoning). It will not load on OpenCode V1.
OpenCode V2 — full self-heal (github:arfoux/rs-guard):
opencode plugin add github:arfoux/rs-guard
Forces stateless full replay on every request (no previous_response_id, store: false, no include: reasoning.encrypted_content), strips server-issued ids (rs_*, msg_*) from metadata, keeps at most 1 newest encrypted blob. On a matching error it retries once with no reasoning at all.
V1 (@opencode-ai/plugin) has no request/retry hooks, so it works with what V1 offers: chat.params strips include: reasoning.encrypted_content / continuation ids and forces store: false; experimental.chat.messages.transform strips stale ids/blobs; experimental.session.compacting keeps reasoning state out of summaries; session.error shows a guidance toast (retry, or fresh session if it persists).
Tested on:muse-spark-1.3-contributor-free via Gateway Console — fails on message 2–3 without the plugin, works consistently with the V2 plugin. (For the report by @egkike on opencode 1.18.31 / V1: use rs-guard-v1; it reduces the failure surface but cannot auto-retry — if the error still appears, just retry, or start a fresh session.)
Caveat: Both are client-side mitigations, not a server fix. The root cause is upstream replaying stale encrypted/reasoning state that the Console free tier rejects. Same family as #28678 / #29000 but for Console/Gateway models.
Same symptom here on a fresh short session: muse-spark-1.3-contributor-free (exact model ID opencode/muse-spark-1.3-contributor-free) via the Console gateway fails at turn 2 with reasoning encrypted_content was not issued to this caller, no model switching, no resume/compact involved. opencode 1.18.31, Ubuntu 24.04. Consolidating my report #49368 here.
Extra data point: the same rejection also reproduces outside opencode. Running the same Console models (muse-spark-1.3-contributor-free via the command-code provider) from the Pi coding agent hits the same reasoning encrypted_content was not issued to this caller failure. No opencode client code involved on that path, which supports the gateway-side root cause — stale encrypted/reasoning state rejected by the Console free tier regardless of client.
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.
Activity
github-actions commentedon Sep 15, 2026
This issue might be a duplicate of existing issues. Please check:
encrypted_contentwas not issued to this caller #48973: Same error with Muse Spark 1.3 Free in short sessionsencrypted_contenterror after 2-3 messages (including with images, similar session length)reasoningEncryptedContentblobs expire and a proposed fixYour report specifically highlights that this is now occurring in very short new sessions (2-3 messages) as of 2026-09-15, which could indicate a regression worth tracking separately. The root cause analysis you provided (encrypted blob from turn N being rejected at turn N+1) matches the discussion in #48915.
arfoux commentedon Sep 16, 2026
Correction — OpenCode version matters:
My earlier comment recommended
github:arfoux/rs-guard, but that plugin is OpenCode V2 only (@opencode/pluginv2 API:context/compaction/generate+retryhooks for stateless replay and retry-without-reasoning). It will not load on OpenCode V1.github:arfoux/rs-guard):previous_response_id,store: false, noinclude: reasoning.encrypted_content), strips server-issued ids (rs_*,msg_*) from metadata, keeps at most 1 newest encrypted blob. On a matching error it retries once with no reasoning at all.github:arfoux/rs-guard-v1):@opencode-ai/plugin) has no request/retry hooks, so it works with what V1 offers:chat.paramsstripsinclude: reasoning.encrypted_content/ continuation ids and forcesstore: false;experimental.chat.messages.transformstrips stale ids/blobs;experimental.session.compactingkeeps reasoning state out of summaries;session.errorshows a guidance toast (retry, or fresh session if it persists).Tested on:
muse-spark-1.3-contributor-freevia Gateway Console — fails on message 2–3 without the plugin, works consistently with the V2 plugin. (For the report by @egkike on opencode 1.18.31 / V1: users-guard-v1; it reduces the failure surface but cannot auto-retry — if the error still appears, just retry, or start a fresh session.)Caveat: Both are client-side mitigations, not a server fix. The root cause is upstream replaying stale encrypted/reasoning state that the Console free tier rejects. Same family as #28678 / #29000 but for Console/Gateway models.
egkike commentedon Sep 16, 2026
Same symptom here on a fresh short session:
muse-spark-1.3-contributor-free(exact model IDopencode/muse-spark-1.3-contributor-free) via the Console gateway fails at turn 2 withreasoning encrypted_content was not issued to this caller, no model switching, no resume/compact involved. opencode 1.18.31, Ubuntu 24.04. Consolidating my report #49368 here.egkike commentedon Sep 16, 2026
Extra data point: the same rejection also reproduces outside opencode. Running the same Console models (
muse-spark-1.3-contributor-freevia the command-code provider) from the Pi coding agent hits the samereasoning encrypted_content was not issued to this callerfailure. No opencode client code involved on that path, which supports the gateway-side root cause — stale encrypted/reasoning state rejected by the Console free tier regardless of client.Tarquinen commentedon Oct 20, 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