opencode bug: compaction summary requests hard-capped at 1024 output tokens, ignoring model limit
d1g1t · ghdev · 2026-10-12
- A reproduced issue in opencode (v2.0.27) shows auto/manual compaction failing on long sessions (80K+ input tokens) because the summary request is effectively capped at 1024 output tokens, ignoring the model's configured limit.output (32768 in the report), yielding Compaction summary reached the output token limit — most real session summaries need far more than 1K tokens.
- Reproduction: register a custom openai-compatible local model (llama.cpp-style server running a Qwen3 27B GGUF) with limit.output set; GET /api/model confirms the 32768 limit is loaded, yet the actual POST /v1/chat/completions stops at exactly 1024 completion tokens with stop reason length.
- Impact: users on custom/local models essentially cannot compact long sessions.
More from coding & agent
- Decisions API hits GA as agents automate cloud video editing via feedback loops — dkundel · 2026-10-23
- Swarms API changelog: 8 months of streaming, multi-agent orchestration, and dynamic model pricing — KyeGomezB · 2026-10-23
- A curated library of 50+ free AI agent resources: videos, repos, one Google Doc — ifioknkem · 2026-10-23
- Vercel adds no-account deploys: drag and drop a file to get a live URL — steren · 2026-10-23
- Dev runs 5 Claude Code sessions that coordinate via MCP Agent Mail to root his network cameras — doodlestein · 2026-10-23
- Vercel ships 'vercel deploy --temporary' for ephemeral agent-built deployments — cramforce · 2026-10-23