| id,model_slug,discussion_id,title,body_preview,comments,reactions,author,repo,created_at,collected_at | |
| 1,openai/gpt-5.4-pro,4222414921,"[FEATURE]:github copilot support openai gpt-5.4 and gpt-5.3-codex phase ""commentary"" | ""final_answer"" in message item","### Feature hasn't been suggested before. | |
| - [x] I have verified this feature I'm about to request hasn't been suggested before. | |
| ### Describe the enhancement you want to request | |
| [FEATURE]:github copilot support openai gpt-5.3-codex phase ""commentary"" | ""final_answer"" in message item | |
| phase: option",1,0,caozhiyuan,opencode,2026-04-08T05:51:57Z,"2026-04-15 01:19:20" | |
| 2,openai/gpt-5.4,4222414921,"[FEATURE]:github copilot support openai gpt-5.4 and gpt-5.3-codex phase ""commentary"" | ""final_answer"" in message item","### Feature hasn't been suggested before. | |
| - [x] I have verified this feature I'm about to request hasn't been suggested before. | |
| ### Describe the enhancement you want to request | |
| [FEATURE]:github copilot support openai gpt-5.3-codex phase ""commentary"" | ""final_answer"" in message item | |
| phase: option",1,0,caozhiyuan,opencode,2026-04-08T05:51:57Z,"2026-04-15 01:19:20" | |
| 3,openai/gpt-5.4-pro,4200369306,"[Discussion] LLM Suggestions","Do you have LLM suggestions? As far as I saw you do not specify models in the agent .md's. | |
| I assume that all agents addressed as subagent by the orchestartor use the model specified when executing the orchestrator this way. | |
| Spicifing models can reduce cost but even increase perfomance as models have",7,0,JulianS-Uni,gem-team,2026-04-03T12:35:54Z,"2026-04-15 01:19:20" | |
| 4,openai/gpt-5.4,4200369306,"[Discussion] LLM Suggestions","Do you have LLM suggestions? As far as I saw you do not specify models in the agent .md's. | |
| I assume that all agents addressed as subagent by the orchestartor use the model specified when executing the orchestrator this way. | |
| Spicifing models can reduce cost but even increase perfomance as models have",7,0,JulianS-Uni,gem-team,2026-04-03T12:35:54Z,"2026-04-15 01:19:20" | |
| 5,openai/gpt-5.4-pro,4165453302,"[FEATURE]:show recently used models in webapp first","### Feature hasn't been suggested before. | |
| - [x] I have verified this feature I'm about to request hasn't been suggested before. | |
| ### Describe the enhancement you want to request | |
| Evry time I would like to switch in between github-copilot/claude-4.6-opus, azure/gpt-5.4 xhigh, I have to search a mode",2,0,dzianisv,opencode,2026-03-29T18:34:28Z,"2026-04-15 01:19:20" | |
| 6,openai/gpt-5.4,4165453302,"[FEATURE]:show recently used models in webapp first","### Feature hasn't been suggested before. | |
| - [x] I have verified this feature I'm about to request hasn't been suggested before. | |
| ### Describe the enhancement you want to request | |
| Evry time I would like to switch in between github-copilot/claude-4.6-opus, azure/gpt-5.4 xhigh, I have to search a mode",2,0,dzianisv,opencode,2026-03-29T18:34:28Z,"2026-04-15 01:19:20" | |
| 7,openai/gpt-5.4-pro,4140046449,"[FEATURE]:Support Interactive Steering","### Feature hasn't been suggested before. | |
| - [x] I have verified this feature I'm about to request hasn't been suggested before. | |
| ### Describe the enhancement you want to request | |
| GPT-5.4 now Support Interactive Steering, Opencode send message when a task running appears QUEUED, need a feature to s",1,2,me10zyl,opencode,2026-03-26T02:45:59Z,"2026-04-15 01:19:20" | |
| 8,openai/gpt-5.4,4140046449,"[FEATURE]:Support Interactive Steering","### Feature hasn't been suggested before. | |
| - [x] I have verified this feature I'm about to request hasn't been suggested before. | |
| ### Describe the enhancement you want to request | |
| GPT-5.4 now Support Interactive Steering, Opencode send message when a task running appears QUEUED, need a feature to s",1,2,me10zyl,opencode,2026-03-26T02:45:59Z,"2026-04-15 01:19:20" | |
| 9,openai/gpt-5.4-pro,4111599782,"[FEATURE]: Add an option to make opencode less verbose","### Feature hasn't been suggested before. | |
| - [x] I have verified this feature I'm about to request hasn't been suggested before. | |
| ### Describe the enhancement you want to request | |
| Very often I find the ""thinking"" blocks to be unnecessarily verbose that it dilutes actually useful output. | |
| It would be",4,0,1NC4NDESCENCE,opencode,2026-03-21T06:21:28Z,"2026-04-15 01:19:20" | |
| 10,openai/gpt-5.4,4111599782,"[FEATURE]: Add an option to make opencode less verbose","### Feature hasn't been suggested before. | |
| - [x] I have verified this feature I'm about to request hasn't been suggested before. | |
| ### Describe the enhancement you want to request | |
| Very often I find the ""thinking"" blocks to be unnecessarily verbose that it dilutes actually useful output. | |
| It would be",4,0,1NC4NDESCENCE,opencode,2026-03-21T06:21:28Z,"2026-04-15 01:19:20" | |
| 11,openai/gpt-5.4-pro,4093667868,"[FEATURE]: Add GPT 5.4 Mini/Nano models to Codex","### Feature hasn't been suggested before. | |
| - [x] I have verified this feature I'm about to request hasn't been suggested before. | |
| ### Describe the enhancement you want to request | |
| Please add support for the new GPT 5.4 Mini and Nano models in the Codex plugin.",6,9,alexhilgart,opencode,2026-03-18T08:48:57Z,"2026-04-15 01:19:20" | |
| 12,openai/gpt-5.4,4093667868,"[FEATURE]: Add GPT 5.4 Mini/Nano models to Codex","### Feature hasn't been suggested before. | |
| - [x] I have verified this feature I'm about to request hasn't been suggested before. | |
| ### Describe the enhancement you want to request | |
| Please add support for the new GPT 5.4 Mini and Nano models in the Codex plugin.",6,9,alexhilgart,opencode,2026-03-18T08:48:57Z,"2026-04-15 01:19:20" | |
| 13,openai/gpt-5.4-pro,4086038059,"[Discussion]: model ""gpt-5.4"" is not accessible via the /chat/completions endpoint","### Issue Checklist | |
| - [x] I understand that issues are meant for feedback and problem-solving, not for venting, and I will provide as much detail as possible to help resolve the issue. | |
| - [x] I have checked the pinned issues and searched through the existing [open issues](https://github.com/CherryHQ",2,2,sunnyXn,cherry-studio,2026-03-17T03:49:14Z,"2026-04-15 01:19:20" | |
| 14,openai/gpt-5.4,4086038059,"[Discussion]: model ""gpt-5.4"" is not accessible via the /chat/completions endpoint","### Issue Checklist | |
| - [x] I understand that issues are meant for feedback and problem-solving, not for venting, and I will provide as much detail as possible to help resolve the issue. | |
| - [x] I have checked the pinned issues and searched through the existing [open issues](https://github.com/CherryHQ",2,2,sunnyXn,cherry-studio,2026-03-17T03:49:14Z,"2026-04-15 01:19:20" | |
| 15,openai/gpt-5.4-pro,4086036571,"[Discussion]: model ""gpt-5.4"" is not accessible via the /chat/completions endpoint","### Issue Checklist | |
| - [x] I understand that issues are meant for feedback and problem-solving, not for venting, and I will provide as much detail as possible to help resolve the issue. | |
| - [x] I have checked the pinned issues and searched through the existing [open issues](https://github.com/CherryHQ",0,1,sunnyXn,cherry-studio,2026-03-17T03:48:31Z,"2026-04-15 01:19:20" | |
| 16,openai/gpt-5.4,4086036571,"[Discussion]: model ""gpt-5.4"" is not accessible via the /chat/completions endpoint","### Issue Checklist | |
| - [x] I understand that issues are meant for feedback and problem-solving, not for venting, and I will provide as much detail as possible to help resolve the issue. | |
| - [x] I have checked the pinned issues and searched through the existing [open issues](https://github.com/CherryHQ",0,1,sunnyXn,cherry-studio,2026-03-17T03:48:31Z,"2026-04-15 01:19:20" | |
| 17,openai/gpt-5.4-pro,4061591341,"[Bug] Agent incorrectly calls done action when task is not completed","### What happened? | |
| ### Description | |
| The agent incorrectly calls the done action when the task is not yet completed. This happens when the agent's next_goal clearly states it needs to perform an action (e.g., ""click the search button""), but instead of executing that action, it calls done with success",2,0,WenElevating,page-agent,2026-03-12T01:48:16Z,"2026-04-15 01:19:20" | |
| 18,openai/gpt-5.4,4061591341,"[Bug] Agent incorrectly calls done action when task is not completed","### What happened? | |
| ### Description | |
| The agent incorrectly calls the done action when the task is not yet completed. This happens when the agent's next_goal clearly states it needs to perform an action (e.g., ""click the search button""), but instead of executing that action, it calls done with success",2,0,WenElevating,page-agent,2026-03-12T01:48:16Z,"2026-04-15 01:19:20" | |
| 19,openai/gpt-5.4-pro,4038944883,"[FEATURE]: add GPT-5.4 fast mode (/fast) as a model control","### Feature hasn't been suggested before. | |
| - [x] I have verified this feature I'm about to request hasn't been suggested before. | |
| ### Describe the enhancement you want to request | |
| Add GPT-5.4 fast mode (`/fast`) to the TUI. | |
| This feature was introduced with GPT-5.4, and opencode should expose it in ",16,69,Pribess,opencode,2026-03-07T17:05:23Z,"2026-04-15 01:19:20" | |
| 20,openai/gpt-5.4,4038944883,"[FEATURE]: add GPT-5.4 fast mode (/fast) as a model control","### Feature hasn't been suggested before. | |
| - [x] I have verified this feature I'm about to request hasn't been suggested before. | |
| ### Describe the enhancement you want to request | |
| Add GPT-5.4 fast mode (`/fast`) to the TUI. | |
| This feature was introduced with GPT-5.4, and opencode should expose it in ",16,69,Pribess,opencode,2026-03-07T17:05:23Z,"2026-04-15 01:19:20" | |
| 21,openai/gpt-5.4-pro,4032615199,"[FEATURE]: Allow overriding model input limits for OpenAI long-context models","### Feature hasn't been suggested before. | |
| - [x] I have verified this feature I'm about to request hasn't been suggested before. | |
| ### Describe the enhancement you want to request | |
| I would like opencode to support user-overridable input limits for OpenAI models such as `gpt-5.4`, or otherwise align i",2,6,ykswang,opencode,2026-03-06T06:58:05Z,"2026-04-15 01:19:20" | |
| 22,openai/gpt-5.4,4032615199,"[FEATURE]: Allow overriding model input limits for OpenAI long-context models","### Feature hasn't been suggested before. | |
| - [x] I have verified this feature I'm about to request hasn't been suggested before. | |
| ### Describe the enhancement you want to request | |
| I would like opencode to support user-overridable input limits for OpenAI models such as `gpt-5.4`, or otherwise align i",2,6,ykswang,opencode,2026-03-06T06:58:05Z,"2026-04-15 01:19:20" | |
| 23,openai/gpt-5.4-pro,4030193284,"[FEATURE]: Add GPT 5.4 models","### Feature hasn't been suggested before. | |
| - [x] I have verified this feature I'm about to request hasn't been suggested before. | |
| ### Describe the enhancement you want to request | |
| Add the new OpenAI models through the oAuth",5,116,tully-8888,opencode,2026-03-05T19:37:03Z,"2026-04-15 01:19:20" | |
| 24,openai/gpt-5.4,4030193284,"[FEATURE]: Add GPT 5.4 models","### Feature hasn't been suggested before. | |
| - [x] I have verified this feature I'm about to request hasn't been suggested before. | |
| ### Describe the enhancement you want to request | |
| Add the new OpenAI models through the oAuth",5,116,tully-8888,opencode,2026-03-05T19:37:03Z,"2026-04-15 01:19:20" | |
| 25,openai/gpt-5.4-pro,3919560582,"[FEATURE]: Support for Claude Opus 4.6 Fast Mode","### Feature hasn't been suggested before. | |
| - [x] I have verified this feature I'm about to request hasn't been suggested before. | |
| ### Describe the enhancement you want to request | |
| First, thank you for maintaining this project. I really appreciate all your work. | |
| I wanted to bring to your attention t",12,17,Tetrahedrite,opencode,2026-02-10T05:43:57Z,"2026-04-15 01:19:20" | |
| 26,openai/gpt-5.4,3919560582,"[FEATURE]: Support for Claude Opus 4.6 Fast Mode","### Feature hasn't been suggested before. | |
| - [x] I have verified this feature I'm about to request hasn't been suggested before. | |
| ### Describe the enhancement you want to request | |
| First, thank you for maintaining this project. I really appreciate all your work. | |
| I wanted to bring to your attention t",12,17,Tetrahedrite,opencode,2026-02-10T05:43:57Z,"2026-04-15 01:19:20" | |
| 27,openai/gpt-5.4-pro,3846750853,"[FEATURE]:Support Augment code","### Feature hasn't been suggested before. | |
| - [x] I have verified this feature I'm about to request hasn't been suggested before. | |
| ### Describe the enhancement you want to request | |
| Hello ! | |
| Opencode is amazing because of its possibility with different provider however I do not find the provider I'm cu",6,5,yoobi,opencode,2026-01-23T09:59:28Z,"2026-04-15 01:19:20" | |
| 28,openai/gpt-5.4,3846750853,"[FEATURE]:Support Augment code","### Feature hasn't been suggested before. | |
| - [x] I have verified this feature I'm about to request hasn't been suggested before. | |
| ### Describe the enhancement you want to request | |
| Hello ! | |
| Opencode is amazing because of its possibility with different provider however I do not find the provider I'm cu",6,5,yoobi,opencode,2026-01-23T09:59:28Z,"2026-04-15 01:19:20" | |
| 29,openai/gpt-5.4-pro,3799376515,"[FEATURE]: Native Model Fallback / Failover Support","## Problem | |
| OpenCode supports: | |
| - Provider fallback **only** when the model ID is the same | |
| - Static agent-level model overrides | |
| There is **no way to define fallback between different models**, e.g.: | |
| > “If model A errors or rate-limits → automatically retry with model B” | |
| This causes long-running agen",22,65,arisgrout,opencode,2026-01-10T09:01:08Z,"2026-04-15 01:19:20" | |
| 30,openai/gpt-5.4,3799376515,"[FEATURE]: Native Model Fallback / Failover Support","## Problem | |
| OpenCode supports: | |
| - Provider fallback **only** when the model ID is the same | |
| - Static agent-level model overrides | |
| There is **no way to define fallback between different models**, e.g.: | |
| > “If model A errors or rate-limits → automatically retry with model B” | |
| This causes long-running agen",22,65,arisgrout,opencode,2026-01-10T09:01:08Z,"2026-04-15 01:19:20" | |
| 31,openai/gpt-5.4-mini,4200369306,"[Discussion] LLM Suggestions","Do you have LLM suggestions? As far as I saw you do not specify models in the agent .md's. | |
| I assume that all agents addressed as subagent by the orchestartor use the model specified when executing the orchestrator this way. | |
| Spicifing models can reduce cost but even increase perfomance as models have",7,0,JulianS-Uni,gem-team,2026-04-03T12:35:54Z,"2026-04-15 01:19:23" | |
| 32,openai/gpt-5.4-mini,4093667868,"[FEATURE]: Add GPT 5.4 Mini/Nano models to Codex","### Feature hasn't been suggested before. | |
| - [x] I have verified this feature I'm about to request hasn't been suggested before. | |
| ### Describe the enhancement you want to request | |
| Please add support for the new GPT 5.4 Mini and Nano models in the Codex plugin.",6,9,alexhilgart,opencode,2026-03-18T08:48:57Z,"2026-04-15 01:19:23" | |
| 33,openai/gpt-5.4-mini,4061591341,"[Bug] Agent incorrectly calls done action when task is not completed","### What happened? | |
| ### Description | |
| The agent incorrectly calls the done action when the task is not yet completed. This happens when the agent's next_goal clearly states it needs to perform an action (e.g., ""click the search button""), but instead of executing that action, it calls done with success",2,0,WenElevating,page-agent,2026-03-12T01:48:16Z,"2026-04-15 01:19:23" | |
| 34,openai/gpt-5.4-mini,3799376515,"[FEATURE]: Native Model Fallback / Failover Support","## Problem | |
| OpenCode supports: | |
| - Provider fallback **only** when the model ID is the same | |
| - Static agent-level model overrides | |
| There is **no way to define fallback between different models**, e.g.: | |
| > “If model A errors or rate-limits → automatically retry with model B” | |
| This causes long-running agen",22,65,arisgrout,opencode,2026-01-10T09:01:08Z,"2026-04-15 01:19:23" | |
| 35,openai/o4-mini,3901748116,"[FEATURE]: Support configuring OpenAI `service_tier='flex'` for chat/completions and document in config","### Feature hasn't been suggested before. | |
| - [x] I have verified this feature I'm about to request hasn't been suggested before. | |
| ### Describe the enhancement you want to request | |
| I am using opencode with OpenAI and want to be able to select the `service_tier='flex'` option for supported OpenAI mode",2,1,duysqubix,opencode,2026-02-05T14:02:29Z,"2026-04-15 01:19:26" | |
| 36,openai/o4-mini,3006086295,"[功能]: 增加OCR功能","### 提交前检查 | |
| - [x] 我理解 Issue 是用于反馈和解决问题的，而非吐槽评论区，将尽可能提供更多信息帮助问题解决。 | |
| - [x] 我已经查看了置顶 Issue 并搜索了现有的 [开放Issue](https://github.com/CherryHQ/cherry-studio/issues)和[已关闭Issue](https://github.com/CherryHQ/cherry-studio/issues?q=is%3Aissue%20state%3Aclosed%20)，没有找到类似的建议。 | |
| - [x] 我填写了简短且清晰明确的标题，以便开发者在翻阅 Issue 列表时能快",5,0,Gonghysin,cherry-studio,2025-04-19T02:09:11Z,"2026-04-15 01:19:26" | |
| 37,openai/o3,3901748116,"[FEATURE]: Support configuring OpenAI `service_tier='flex'` for chat/completions and document in config","### Feature hasn't been suggested before. | |
| - [x] I have verified this feature I'm about to request hasn't been suggested before. | |
| ### Describe the enhancement you want to request | |
| I am using opencode with OpenAI and want to be able to select the `service_tier='flex'` option for supported OpenAI mode",2,1,duysqubix,opencode,2026-02-05T14:02:29Z,"2026-04-15 01:19:29" | |
| 38,anthropic/claude-opus-4.6,4222112546,"new error: Third-party apps now draw from extra usage...","### Description | |
| I've been using this lib, but I just got this message from a 400 response: | |
| > opencode-claude-auth: API 400 for claude-opus-4-6: Third-party apps now draw from extra usage, not plan limits. We've added a credit to your organization to get you started. Ask your workspace admin to cla",73,29,phazei,opencode-claude-auth,2026-04-08T04:28:15Z,"2026-04-15 01:19:33" | |
| 39,anthropic/claude-opus-4.6,4208813685,"Discussion: Policy for AI-assisted contributions","This has been touched upon earlier on the IRC channels (e.g. https://irc.luanti.org/luanti-dev/2026-04-02 ff.) and ad-hoc on PRs like #16929. I'm opening this issue so we can have public discussion in one place. | |
| I think we should start with some reasonable minimal policy. We can always expand upon ",54,7,appgurueu,luanti,2026-04-05T23:30:45Z,"2026-04-15 01:19:33" | |
| 40,anthropic/claude-opus-4.6,4207492759,"docs: migrate user-guide.md into system prompt (prompt-as-manual)","## Problem | |
| Koda maintains user knowledge across **5 overlapping surfaces**: | |
| | Surface | Lines | Audience | | |
| |---|---|---| | |
| | `koda-core/src/instructions.md` | ~100 | LLM (behavioral rules) | | |
| | `koda-core/src/capabilities.md` | ~75 | LLM (feature reference) | | |
| | `docs/user-guide.md` | ~340 | Humans (w",7,0,lijunzh,koda,2026-04-05T13:01:57Z,"2026-04-15 01:19:33" | |
| 41,anthropic/claude-opus-4.6,4197013384,"Add AI-assistance policy, agent instructions, and code style docs","Adds three files that arose out of the dev call on March 23, namely a policy statement and two guides, and a few pointer files to cover various AI tools, and makes changes to existing documentation files to avoid duplication. | |
| This is being submitted in draft mode so we can evaluate/comment on, bu",0,1,adcroft,MOM6,2026-04-02T21:12:24Z,"2026-04-15 01:19:33" | |
| 42,anthropic/claude-opus-4.6,4165453302,"[FEATURE]:show recently used models in webapp first","### Feature hasn't been suggested before. | |
| - [x] I have verified this feature I'm about to request hasn't been suggested before. | |
| ### Describe the enhancement you want to request | |
| Evry time I would like to switch in between github-copilot/claude-4.6-opus, azure/gpt-5.4 xhigh, I have to search a mode",2,0,dzianisv,opencode,2026-03-29T18:34:28Z,"2026-04-15 01:19:33" | |
| 43,anthropic/claude-opus-4.6,4125971854,"[FEATURE]: Wildcard supports for agent configuration","### Feature hasn't been suggested before. | |
| - [x] I have verified this feature I'm about to request hasn't been suggested before. | |
| ### Describe the enhancement you want to request | |
| Summary | |
| I’d like to have wildcard support in agent configuration so I can target groups of subagents by naming pattern ",0,0,huybui38,opencode,2026-03-24T07:59:24Z,"2026-04-15 01:19:33" | |
| 44,anthropic/claude-opus-4.6,4116458178,"Cloud LLM (OpenRouter): region-blocked models fail with 403 from HK server","## Problem | |
| Premium LLM diagnosis via OpenRouter fails with **403 Forbidden** when called from the PolyU GPU server (Hong Kong): | |
| ``` | |
| HTTP Request: POST https://openrouter.ai/api/v1/chat/completions ""HTTP/1.1 403 Forbidden"" | |
| error: {'message': 'This model is not available in your region.', 'code': 40",1,0,talonMMA,stf_ai_diagnosis_platform_v1,2026-03-22T16:27:18Z,"2026-04-15 01:19:33" | |
| 45,anthropic/claude-opus-4.6,4105196829,"[FEATURE]: Allow pattern-matching for provider blacklist/whitelist in config","### Feature hasn't been suggested before. | |
| - [x] I have verified this feature I'm about to request hasn't been suggested before. | |
| ### Describe the enhancement you want to request | |
| I blacklist all old models so only newer models show up in `/model` suggestions. | |
| This is an example of my current confi",1,0,flexdinesh,opencode,2026-03-20T01:46:15Z,"2026-04-15 01:19:33" | |
| 46,anthropic/claude-opus-4.6,4095537846,"Fold `fre pp rename-split` into `fre pp split-netcdf --rename`; fix #862 aerosol variable misclassification","## Describe your changes | |
| Takes the `fre pp rename-split` functionality from the `rename-split` branch and folds it into `fre pp split-netcdf` via a `--rename` flag. Without `--rename`, `split-netcdf` behaves as before. Also fixes #862 where certain aerosol variables were incorrectly classified as m",16,0,Copilot,fre-cli,2026-03-18T14:31:51Z,"2026-04-15 01:19:33" | |
| 47,anthropic/claude-opus-4.6,4081917714,"Call graph change analysis from diffs","**Motivation** | |
| When reviewing a code change, it would be valuable to see how the call graph (or module dependency graph) changed — which edges were added, which were removed. This helps both human reviewers and AI agents reason about the effects of a diff on the codebase structure. | |
| **Sketch** | |
| Com",0,0,Technologicat,pyan,2026-03-16T11:31:50Z,"2026-04-15 01:19:33" | |
| 48,anthropic/claude-opus-4.6,4076644729,"MCP Scoping & Cross-Entity Orchestration","## Problem | |
| In OB1's current architecture, each extension ships its own MCP server (Supabase Edge Function). When a user adds extensions, their AI client accumulates MCP tool definitions. This creates two problems: | |
| ### 1. Context Weight Accumulation | |
| Each MCP server exposes tool definitions that th",8,0,matthallett1,OB1,2026-03-14T20:18:35Z,"2026-04-15 01:19:33" | |
| 49,anthropic/claude-opus-4.6,4075423930,"[FEATURE]: Auto-continue when model hits output token limit (finish_reason: ""length"")","### Feature hasn't been suggested before. | |
| - [x] I have verified this feature I'm about to request hasn't been suggested before. | |
| ### Describe the enhancement you want to request | |
| When using models with large context windows (e.g., Claude Opus 4.6 with 1 million input context), the input context is ",0,4,stephenstubbs,opencode,2026-03-14T10:23:45Z,"2026-04-15 01:19:33" | |
| 50,anthropic/claude-opus-4.6,4063585232,"[Optimus] Epic: Meta-Cron V2 — Autonomous Trigger Architecture (Cron = Trigger, Agent = Brain)","## Overview | |
| This epic tracks the architectural discussion for **Meta-Cron V2** — a fundamental revision of the Meta-Cron design (originally Issue #124). | |
| ## Key Design Change | |
| **V1 (Original)**: Cron entries included a `task_description` field — the cron told the agent exactly what to do. | |
| **V2 (Re",2,0,cloga,optimus-code,2026-03-12T10:13:58Z,"2026-04-15 01:19:33" | |
| 51,anthropic/claude-opus-4.6,4063254129,"[Optimus] Epic: Auto-Skill Genesis (Automatic Skill Generation After T3→T2 Precipitation)","## Summary | |
| This epic tracks the architectural discussion and council review for **Auto-Skill Genesis** — a proposed system where the Optimus Spartan Swarm automatically generates new skills (SKILL.md files) when a T3 dynamic agent successfully completes a novel task pattern, ""precipitating"" the lea",2,0,cloga,optimus-code,2026-03-12T09:15:47Z,"2026-04-15 01:19:33" | |
| 52,anthropic/claude-opus-4.6,4038944883,"[FEATURE]: add GPT-5.4 fast mode (/fast) as a model control","### Feature hasn't been suggested before. | |
| - [x] I have verified this feature I'm about to request hasn't been suggested before. | |
| ### Describe the enhancement you want to request | |
| Add GPT-5.4 fast mode (`/fast`) to the TUI. | |
| This feature was introduced with GPT-5.4, and opencode should expose it in ",16,69,Pribess,opencode,2026-03-07T17:05:23Z,"2026-04-15 01:19:33" | |
| 53,anthropic/claude-sonnet-4.6,4242440200,"arch: evaluate apply_patch format as Edit tool replacement","## Background | |
| Codex uses an `apply_patch` format for file editing instead of the `old_str`/`new_str` approach used by Koda, Claude Code, and Gemini CLI. During discussion of #804 item 7 (Edit validation error context), the question came up: should Koda migrate to the apply_patch pattern? | |
| This issu",1,0,lijunzh,koda,2026-04-11T01:13:31Z,"2026-04-15 01:19:36" | |
| 54,anthropic/claude-sonnet-4.6,4215597461,"[FEATURE]: add `opencode doctor` diagnostic command","### Feature hasn't been suggested before. | |
| - [x] I have verified this feature I'm about to request hasn't been suggested before. | |
| ### Describe the enhancement you want to request | |
| OpenCode supports tons of providers, models, agents, MCP servers, LSP servers, formatters, and plugins -- all of which c",2,0,katipally,opencode,2026-04-07T05:18:06Z,"2026-04-15 01:19:36" | |
| 55,anthropic/claude-sonnet-4.6,4151062102,"[FEATURE]: Show Metadata (e.g. Model Context Limit) while Selecting Models in Opencode CLI Menu","### Feature hasn't been suggested before. | |
| - [x] I have verified this feature I'm about to request hasn't been suggested before. | |
| ### Describe the enhancement you want to request | |
| **Description** | |
| Currently, when selecting models via Ctrl+P in the Opencode CLI, users lack visibility into the specific",0,0,Tao-Yida,opencode,2026-03-27T09:41:17Z,"2026-04-15 01:19:36" | |
| 56,anthropic/claude-sonnet-4.6,4116458178,"Cloud LLM (OpenRouter): region-blocked models fail with 403 from HK server","## Problem | |
| Premium LLM diagnosis via OpenRouter fails with **403 Forbidden** when called from the PolyU GPU server (Hong Kong): | |
| ``` | |
| HTTP Request: POST https://openrouter.ai/api/v1/chat/completions ""HTTP/1.1 403 Forbidden"" | |
| error: {'message': 'This model is not available in your region.', 'code': 40",1,0,talonMMA,stf_ai_diagnosis_platform_v1,2026-03-22T16:27:18Z,"2026-04-15 01:19:36" | |
| 57,anthropic/claude-sonnet-4.6,4105196829,"[FEATURE]: Allow pattern-matching for provider blacklist/whitelist in config","### Feature hasn't been suggested before. | |
| - [x] I have verified this feature I'm about to request hasn't been suggested before. | |
| ### Describe the enhancement you want to request | |
| I blacklist all old models so only newer models show up in `/model` suggestions. | |
| This is an example of my current confi",1,0,flexdinesh,opencode,2026-03-20T01:46:15Z,"2026-04-15 01:19:36" | |
| 58,anthropic/claude-sonnet-4.6,4098303534,[FEATURE]:,"### Feature hasn't been suggested before. | |
| - [x] I have verified this feature I'm about to request hasn't been suggested before. | |
| ### Describe the enhancement you want to request | |
| ⬥ Here's a feature request you can copy and paste to https://github.com/anomalyco/opencode/issues: | |
| --- | |
| **Image Upload ",2,0,MadamGroovy,opencode,2026-03-18T23:35:50Z,"2026-04-15 01:19:36" | |
| 59,anthropic/claude-sonnet-4.6,4063415565,"[Optimus] Epic: Meta-Immune System — Agent Retirement, Experience Capture & GC","## Summary | |
| This epic tracks the architectural discussion and council review for a **Meta-Immune System** — a proposed mechanism for agent retirement, experience capture, and garbage collection within the Optimus Spartan Swarm. | |
| ## Objective | |
| Debate the feasibility, architecture, safety gates, and a",3,0,cloga,optimus-code,2026-03-12T09:45:09Z,"2026-04-15 01:19:36" | |
| 60,anthropic/claude-sonnet-4.6,4032615199,"[FEATURE]: Allow overriding model input limits for OpenAI long-context models","### Feature hasn't been suggested before. | |
| - [x] I have verified this feature I'm about to request hasn't been suggested before. | |
| ### Describe the enhancement you want to request | |
| I would like opencode to support user-overridable input limits for OpenAI models such as `gpt-5.4`, or otherwise align i",2,6,ykswang,opencode,2026-03-06T06:58:05Z,"2026-04-15 01:19:36" | |
| 61,anthropic/claude-sonnet-4.6,4025219540,"[FEATURE]: allow skills to request for small/medium/large model hint","### Feature hasn't been suggested before. | |
| - [x] I have verified this feature I'm about to request hasn't been suggested before. | |
| ### Describe the enhancement you want to request | |
| ## The usecase | |
| The most frequent thing I can think of is that I have a `/git` skill with some specifics of how I use gi",1,2,championswimmer,opencode,2026-03-05T02:04:17Z,"2026-04-15 01:19:36" | |
| 62,anthropic/claude-sonnet-4.6,4018056432,"[Discussion] Confusion on Cost Amounts","## Topic | |
| It seems like the cost data that I am getting from the tool are far far higher than what I believe I am paying through Bedrock. Can you help me understand why that may be? | |
| ## Context | |
| Our org is using Cluade Code through Bedrock. Bedrock seems to have the same rates as going to Anthropic ",5,1,chaseconey,toktrack,2026-03-03T18:09:41Z,"2026-04-15 01:19:36" | |
| 63,anthropic/claude-sonnet-4.6,4007327011,"Pattern: Orchestrator-Worker Architecture","## Problem Statement | |
| Software development involves complex, multi-step workflows that require coordination: planning, implementation, code review, testing, and deployment. Managing this manually is error-prone and doesn't scale. | |
| ## The Pattern | |
| **Orchestrator-Worker** separates planning from execu",7,0,rjmurillo-bot,ai-agents,2026-03-01T14:21:53Z,"2026-04-15 01:19:36" | |
| 64,anthropic/claude-sonnet-4.6,4000985377,"[FEATURE]:TUI Bottom Bar Auto-Discovery of Sub-Sessions with Auto-Switch Management","### Feature hasn't been suggested before. | |
| - [x] I have verified this feature I'm about to request hasn't been suggested before. | |
| ### Describe the enhancement you want to request | |
| ## Problem Description | |
| OpenCode currently has the following limitations when managing sub-agent sessions: | |
| 1. **No Auto",3,0,Eric-GoodBoy-Tech,opencode,2026-02-27T13:14:16Z,"2026-04-15 01:19:36" | |
| 65,anthropic/claude-sonnet-4.6,3996602441,"[FEATURE]: Experimental Tool/Skill Cache","### Feature hasn't been suggested before. | |
| - [x] I have verified this feature I'm about to request hasn't been suggested before. | |
| ### Describe the enhancement you want to request | |
| Specifically, I have verified this is not a duplicate of #5416, #4317 or similar. These features are about generic promp",2,1,ShuzhaoFeng,opencode,2026-02-26T16:46:02Z,"2026-04-15 01:19:36" | |
| 66,anthropic/claude-sonnet-4.6,3996056876,"[FEATURE]: Chinese Documentation Missing English Version Updates","### Feature hasn't been suggested before. | |
| - [x] I have verified this feature I'm about to request hasn't been suggested before. | |
| ### Describe the enhancement you want to request | |
| ## Problem Statement | |
| The Chinese translation documents in the `packages/web/src/content/docs/zh-cn/` directory have mis",3,0,niushuai1991,opencode,2026-02-26T15:00:21Z,"2026-04-15 01:19:36" | |
| 67,anthropic/claude-sonnet-4.6,3990022608,"[FEATURE]: Agents to span different model subagents","### Feature hasn't been suggested before. | |
| - [x] I have verified this feature I'm about to request hasn't been suggested before. | |
| ### Describe the enhancement you want to request | |
| I'm using Claude Sonnet, and I would like the agent to be able to span different models of sub-agents: If the task is si",2,0,Thorium,opencode,2026-02-25T15:07:32Z,"2026-04-15 01:19:36" | |
| 68,anthropic/claude-haiku-4.5,4207492759,"docs: migrate user-guide.md into system prompt (prompt-as-manual)","## Problem | |
| Koda maintains user knowledge across **5 overlapping surfaces**: | |
| | Surface | Lines | Audience | | |
| |---|---|---| | |
| | `koda-core/src/instructions.md` | ~100 | LLM (behavioral rules) | | |
| | `koda-core/src/capabilities.md` | ~75 | LLM (feature reference) | | |
| | `docs/user-guide.md` | ~340 | Humans (w",7,0,lijunzh,koda,2026-04-05T13:01:57Z,"2026-04-15 01:19:40" | |
| 69,anthropic/claude-haiku-4.5,4174816614,"Close PR and open an issue: propose opt‑in weight‑threshold layered on top of main","This branch implemented an opt‑in `skip_limited_threshold` weight check to pre-emptively skip degraded agents. Main already implements `is_agent_degraded()` (cooldown-based pre-emptive routability). Close the existing PR (this branch) and open a focused issue proposing an opt‑in weight‑threshold tha",3,0,gabrielkoerich,orch,2026-03-31T00:17:33Z,"2026-04-15 01:19:40" | |
| 70,anthropic/claude-haiku-4.5,4105196829,"[FEATURE]: Allow pattern-matching for provider blacklist/whitelist in config","### Feature hasn't been suggested before. | |
| - [x] I have verified this feature I'm about to request hasn't been suggested before. | |
| ### Describe the enhancement you want to request | |
| I blacklist all old models so only newer models show up in `/model` suggestions. | |
| This is an example of my current confi",1,0,flexdinesh,opencode,2026-03-20T01:46:15Z,"2026-04-15 01:19:40" | |
| 71,anthropic/claude-haiku-4.5,4050097886,"[FEATURE]: Allow agent-level model pools and fallback policies","### Feature hasn't been suggested before. | |
| - [x] I have verified this feature I'm about to request hasn't been suggested before. | |
| ### Describe the enhancement you want to request | |
| - [x] I have searched existing issues and could not find the same request. | |
| - [x] This is a feature request, not a bug re",1,0,young5lee,opencode,2026-03-10T08:00:05Z,"2026-04-15 01:19:40" | |
| 72,anthropic/claude-haiku-4.5,4029503647,"[FEATURE]: Specify small_model per provider","### Feature hasn't been suggested before. | |
| - [x] I have verified this feature I'm about to request hasn't been suggested before. | |
| ### Describe the enhancement you want to request | |
| ## The Problem | |
| Essentially, I go between multiple different providers, and for each one of them, the small model that I ",6,4,rmagatti,opencode,2026-03-05T17:03:26Z,"2026-04-15 01:19:40" | |
| 73,anthropic/claude-haiku-4.5,4025219540,"[FEATURE]: allow skills to request for small/medium/large model hint","### Feature hasn't been suggested before. | |
| - [x] I have verified this feature I'm about to request hasn't been suggested before. | |
| ### Describe the enhancement you want to request | |
| ## The usecase | |
| The most frequent thing I can think of is that I have a `/git` skill with some specifics of how I use gi",1,2,championswimmer,opencode,2026-03-05T02:04:17Z,"2026-04-15 01:19:40" | |
| 74,anthropic/claude-haiku-4.5,4018056432,"[Discussion] Confusion on Cost Amounts","## Topic | |
| It seems like the cost data that I am getting from the tool are far far higher than what I believe I am paying through Bedrock. Can you help me understand why that may be? | |
| ## Context | |
| Our org is using Cluade Code through Bedrock. Bedrock seems to have the same rates as going to Anthropic ",5,1,chaseconey,toktrack,2026-03-03T18:09:41Z,"2026-04-15 01:19:40" | |
| 75,anthropic/claude-haiku-4.5,4007327011,"Pattern: Orchestrator-Worker Architecture","## Problem Statement | |
| Software development involves complex, multi-step workflows that require coordination: planning, implementation, code review, testing, and deployment. Managing this manually is error-prone and doesn't scale. | |
| ## The Pattern | |
| **Orchestrator-Worker** separates planning from execu",7,0,rjmurillo-bot,ai-agents,2026-03-01T14:21:53Z,"2026-04-15 01:19:40" | |
| 76,anthropic/claude-haiku-4.5,3998368594,"Wątpliwość: Hook execution latency — brak budżetu czasowego","## Wątpliwość | |
| Hooki Claude Code wykonują się synchronicznie — blokują interakcję do zakończenia. bestAI może zainstalować do 6 hooków na PreToolUse (check-frozen, check-user-tags, secret-guard, backup-enforcement, confidence-gate, circuit-breaker-gate). | |
| ### Pytanie: jaka jest łączna latencja? | |
| Bra",1,0,radekzm,bestAI,2026-02-27T00:06:56Z,"2026-04-15 01:19:40" | |
| 77,anthropic/claude-haiku-4.5,3996056876,"[FEATURE]: Chinese Documentation Missing English Version Updates","### Feature hasn't been suggested before. | |
| - [x] I have verified this feature I'm about to request hasn't been suggested before. | |
| ### Describe the enhancement you want to request | |
| ## Problem Statement | |
| The Chinese translation documents in the `packages/web/src/content/docs/zh-cn/` directory have mis",3,0,niushuai1991,opencode,2026-02-26T15:00:21Z,"2026-04-15 01:19:40" | |
| 78,anthropic/claude-haiku-4.5,3990022608,"[FEATURE]: Agents to span different model subagents","### Feature hasn't been suggested before. | |
| - [x] I have verified this feature I'm about to request hasn't been suggested before. | |
| ### Describe the enhancement you want to request | |
| I'm using Claude Sonnet, and I would like the agent to be able to span different models of sub-agents: If the task is si",2,0,Thorium,opencode,2026-02-25T15:07:32Z,"2026-04-15 01:19:40" | |
| 79,anthropic/claude-haiku-4.5,3967046356,"FEATURE]: Please, I ask that you stop constantly changing the design of the responses, because you are only making it worse."," | |
| ### Feature hasn't been suggested before. | |
| - [x] I have verified this feature I'm about to request hasn't been suggested before. | |
| ### Describe the enhancement you want to request | |
| I don't know what their obsession is with constantly changing the response design, but now everything is going from ba",9,4,titet11,opencode,2026-02-20T07:25:31Z,"2026-04-15 01:19:40" | |
| 80,anthropic/claude-haiku-4.5,3952820065,"[FEATURE]: Introduce ordered list for model options","### Feature hasn't been suggested before. | |
| - [x] I have verified this feature I'm about to request hasn't been suggested before. | |
| ### Describe the enhancement you want to request | |
| There are requests open that go in this direction, but not quite the same. The issue is the following: | |
| Imagine you have",2,0,RicardoMonteiroSimoes,opencode,2026-02-17T13:38:32Z,"2026-04-15 01:19:40" | |
| 81,anthropic/claude-haiku-4.5,3913411889,"[Architecture] LLM Proxy - credential injection and model routing","## Summary | |
| Agents call a local endpoint for LLM completions. The proxy injects real API keys and routes to the configured model. | |
| ## Key Points | |
| - Agent calls `http://agentcage/v1/chat/completions` with NO auth header | |
| - Proxy injects API key from vault | |
| - Agent literally cannot leak keys (does not hav",6,0,monteslu,agentcage,2026-02-08T19:45:08Z,"2026-04-15 01:19:40" | |
| 82,anthropic/claude-haiku-4.5,3850025045,"[FEATURE]:Specify sub-agent variant (thinking lvl) from script for token economy","### Feature hasn't been suggested before. | |
| - [x] I have verified this feature I'm about to request hasn't been suggested before. | |
| ### Describe the enhancement you want to request | |
| When writing script for divide and conquer feature, sub-tasks may not require the same effort than the main agent. | |
| I w",4,2,Solido,opencode,2026-01-24T02:57:42Z,"2026-04-15 01:19:40" | |
| 83,google/gemini-2.5-pro,4242440200,"arch: evaluate apply_patch format as Edit tool replacement","## Background | |
| Codex uses an `apply_patch` format for file editing instead of the `old_str`/`new_str` approach used by Koda, Claude Code, and Gemini CLI. During discussion of #804 item 7 (Edit validation error context), the question came up: should Koda migrate to the apply_patch pattern? | |
| This issu",1,0,lijunzh,koda,2026-04-11T01:13:31Z,"2026-04-15 01:19:43" | |
| 84,google/gemini-2.5-pro,4111454242,"模型选择和使用经验，看起来gemini和chatgpt涵盖了日常，加上codex也覆盖了编码，非常实用，claude code似乎并没有那么优秀？","整理一下我目前的AI订阅，其实现在也就两个了 | |
| Gemini Ultra和chatGPT Pro，Claude完全退订 | |
| 主要我用AI的工作就是小说、网站、播客、外汇分析+一些日常英文交流等 | |
| Gemini Ultra主要就是冲着deep think，视频方面没怎么用，感觉浪费了一半，要是有100刀的plan，取消flow，保留deep think最好了。 | |
| 目前还是gemini占首席军师之位： | |
| 1. 小说：deep think进行plan和总编辑review的工作，无可替代，写作初稿及2版仍然用普通2.5 pro。小说封面却还是chatGPT最好，虽然imagen 4 ultra+nano",0,0,fxyzbtc,fxyzbtc.github.io,2026-03-21T05:11:36Z,"2026-04-15 01:19:43" | |
| 85,google/gemini-2.5-pro,4050097886,"[FEATURE]: Allow agent-level model pools and fallback policies","### Feature hasn't been suggested before. | |
| - [x] I have verified this feature I'm about to request hasn't been suggested before. | |
| ### Describe the enhancement you want to request | |
| - [x] I have searched existing issues and could not find the same request. | |
| - [x] This is a feature request, not a bug re",1,0,young5lee,opencode,2026-03-10T08:00:05Z,"2026-04-15 01:19:43" | |
| 86,google/gemini-2.5-pro,4019067643,"[FEATURE]: Add routstr response metadata parsing to extract real costs","### Feature hasn't been suggested before. | |
| - [x] I have verified this feature I'm about to request hasn't been suggested before. | |
| ### Describe the enhancement you want to request | |
| I'm using routsrt as a provider. Here is my opencode.json: | |
| ```json | |
| { | |
| ""$schema"": ""https://opencode.ai/config.json"", | |
| ",0,0,dhalsim,opencode,2026-03-03T22:41:53Z,"2026-04-15 01:19:43" | |
| 87,google/gemini-2.5-pro,3941554734,"[FEATURE]: Unify provider system prompts","### Feature hasn't been suggested before. | |
| - [x] I have verified this feature I'm about to request hasn't been suggested before. | |
| ### Describe the enhancement you want to request | |
| System prompts for Anthropic, Codex, and Gemini models are very different, in ways that might not have good justificatio",6,4,smola,opencode,2026-02-14T14:36:43Z,"2026-04-15 01:19:43" | |
| 88,google/gemini-2.5-pro,3911974930,"[FEATURE]: Add Agent Teams Equivalent or Better","### Feature hasn't been suggested before. | |
| - [x] I have verified this feature I'm about to request hasn't been suggested before. | |
| ### Describe the enhancement you want to request | |
| There is an interest in the OpenCode community for an equivalent to Claude Code's Agent Teams: https://www.reddit.com/r/",26,174,NamedIdentity,opencode,2026-02-08T04:34:09Z,"2026-04-15 01:19:43" | |
| 89,google/gemini-2.5-pro,3853114849,"[FEATURE]: Native Multimodal Context Support (Video/Audio)","### Feature hasn't been suggested before. | |
| - [x] I have verified this feature I'm about to request hasn't been suggested before. | |
| ### Describe the enhancement you want to request | |
| Hi Team, | |
| I recently moved to opencode from claude code and I'm liking it way of handling multiple models and working w",9,8,AimAmit,opencode,2026-01-25T11:40:29Z,"2026-04-15 01:19:43" | |
| 90,google/gemini-2.5-pro,3832668297,"[FEATURE]: model fallback in specific agents","### Feature hasn't been suggested before. | |
| - [x] I have verified this feature I'm about to request hasn't been suggested before. | |
| ### Describe the enhancement you want to request | |
| Can you please add a model fallback in the yaml part of the specific_agent.md file? | |
| ``` | |
| --- | |
| description: ... | |
| mode: prima",4,13,kobi2187,opencode,2026-01-20T08:38:37Z,"2026-04-15 01:19:43" | |
| 91,google/gemini-2.5-pro,3821166492,"Model Selection Strategy for QuestFoundry Pipeline","# Model Selection Strategy for QuestFoundry Pipeline | |
| ## Summary | |
| Benchmarking revealed that **instruction-following capability matters more than model size** for structured generation tasks. GPT-4o (our previous default) under-produces output due to RLHF chat optimization that favors brevity. A loc",1,0,pvliesdonk,questfoundry,2026-01-16T09:27:33Z,"2026-04-15 01:19:43" | |
| 92,google/gemini-2.5-pro,3776658888,"[FEATURE]: Dynamic model selection for subagents via Task tool","### Feature hasn't been suggested before. | |
| - [x] I have verified this feature I'm about to request hasn't been suggested before. | |
| ### Describe the enhancement you want to request | |
| ## Problem | |
| When a primary agent invokes a subagent via the Task tool, it cannot dynamically control which model the sub",23,35,mcking-07,opencode,2026-01-02T16:04:07Z,"2026-04-15 01:19:43" | |
| 93,google/gemini-2.5-pro,3763791853,"[FEATURE]: AutoFallback model","### Feature hasn't been suggested before. | |
| - [x] I have verified this feature I'm about to request hasn't been suggested before. | |
| ### Describe the enhancement you want to request | |
| It would be interesting to have a feature whereby when a model fails (especially errors due to exceeding the limit of re",1,1,riccardodebellini,opencode,2025-12-26T17:44:16Z,"2026-04-15 01:19:43" | |
| 94,google/gemini-2.5-pro,3736563583,"Bug: GeminiAgent fails in production due to quota exhaustion (429 Too Many Requests).","- Model: gemini-2.0-flash | |
| - Free tier quota = 0 → requests blocked | |
| - Retry logic ineffective when quota is exhausted | |
| Proposed Fix: | |
| - Default to gemini-2.5-flash-lite in production | |
| - Add user-facing notification when quota is exceeded | |
| - Document quota requirements and billing setup | |
| ",3,0,triloksh07,ai-chat-app-with-agents-v0,2025-12-16T22:30:45Z,"2026-04-15 01:19:43" | |
| 95,google/gemini-2.5-pro,3661979314,"[Discussion]: Text selection search is not working","> [!NOTE] | |
| > This issue was translated by Claude. | |
| I cannot use the text selection assistant for translation, as shown in the image. I don't know how to find the stream, please help me solve this issue. | |
| <img width=652 height=522 alt=Image src=https://github.com/user-attachments/assets/9cb0d711-aa6a-",7,1,snoopytl,cherry-studio,2025-11-25T08:22:57Z,"2026-04-15 01:19:43" | |
| 96,google/gemini-2.5-pro,3640928512,"/models shows all available models rather than those restricted in opencode.json.","### Description | |
| Windows 10, `wsl -d kali-linux`, opencode v1.0.67 | |
| Pro Claude, openrouter and zen authorized per specifications with `opencode auth login` . | |
| opencode.json at project base. | |
| When selecting a model `/models` from within opencode, _every_ single model available from all three provider",5,3,LaughterOnWater,opencode,2025-11-19T03:36:46Z,"2026-04-15 01:19:43" | |
| 97,google/gemini-2.5-pro,3587348369,"Long `ExampleData` causes `extract` to hang on `_fuzzy_align_extraction`","Hi, | |
| First of all, thanks for open sourcing this nice package. :-) | |
| I am trying to use `langextract` to post-process interviews by tagging (sometimes long) quotes in a text. | |
| The problem that I run into, is that the text extraction takes exceedingly long to complete. | |
| For example, it takes 17 minute",2,0,hcdonker-code,langextract,2025-11-04T16:34:19Z,"2026-04-15 01:19:43" | |
| 98,google/gemma-4-31b-it,4238277233,"[Bug] CPOTrainer crashes with `TypeError: 'NoneType' object is not subscriptable` when fine-tuning Gemma 4 (text-only)","1. Did you update? `pip install --upgrade unsloth unsloth_zoo` → **Yes.** Dockerfile runs `pip install --upgrade unsloth unsloth_zoo transformers` on top of the official image (2026.4.4-pt2.9.0-vllm-0.16.0-cu12.8-studio-release-v0.1.35-beta). | |
| 2. `Colab` or `Kaggle` or local / cloud → **Cloud** (Dock",2,0,ko-noguchi,unsloth,2026-04-10T09:54:52Z,"2026-04-15 01:19:47" | |
| 99,google/gemma-4-31b-it,4180551101,"[Bug] Different behaviour observed by using `TrainingArguments` vs `SFTConfig`","I am trying to fine-tune a `Gemma-3` model on some custom instructions. The dataset contains of some 30k samples of 6-8k tokens. I am using unsloth based training for that but I am observing vastly different behaviour using `TrainingArguments` vs `SFTConfig`. I do not expect it to be this drasticall",3,0,vibhas-singh,unsloth,2026-03-31T16:20:24Z,"2026-04-15 01:19:47" | |
| 100,google/gemma-4-31b-it,3725376139,"[Bug] Tiled MLP only seems to be working for GPT-OSS 20B","```unsloth_tiled_mlp = True``` Only seems to be working for GPT-OSS 20B. | |
| **Installation file:** | |
| ```#!/usr/bin/env bash | |
| set -e | |
| # Create venv if not exists | |
| if [ ! -d "".venv"" ]; then | |
| python3 -m venv .venv | |
| fi | |
| # Activate venv | |
| source .venv/bin/activate | |
| # Upgrade pip & install uv inside venv | |
| pip inst",5,0,mlasicbot,unsloth,2025-12-13T02:50:53Z,"2026-04-15 01:19:47" | |
| 101,google/gemma-4-31b-it,3552932643,"[Bug] model.save_pretrained_merged","1. Did you update? `pip install --upgrade unsloth unsloth_zoo` | |
| 2. `Colab` or `Kaggle` or local / cloud | |
| 3. Number GPUs used, use `nvidia-smi` | |
| 4. Which notebook? Please link! | |
| 5. Which Unsloth version, TRL version, transformers version, PyTorch version? | |
| 6. Which trainer? `SFTTrainer`, `GRPOTrainer` etc",5,0,chrmica,unsloth,2025-10-25T20:03:33Z,"2026-04-15 01:19:47" | |
| 102,google/gemma-4-31b-it,3349359844,"[Proposal]: Implement Multi-Stage Red Teaming Workflow","### **Summary:** | |
| This proposal outlines a plan to implement a new, advanced red teaming workflow. The new process formalizes the entire lifecycle from threat intelligence gathering and strategic debate to attack execution and vulnerability reporting. This will significantly enhance the system's abil",3,0,solarfresh,TwinRAD,2025-08-24T09:50:08Z,"2026-04-15 01:19:47" | |
| 103,google/gemma-4-31b-it,3218284175,"Gemma3n - Multimodal Tuning for Medical VQA","1. Did you update? `pip install --upgrade unsloth unsloth_zoo` -- yes | |
| 2. `Colab` or `Kaggle` or local / cloud -- colab | |
| 3. Number GPUs used, use `nvidia-smi` - T4 | |
| 4. Which notebook? Please link! -- added | |
| 5. Which Unsloth version, TRL version, Transformers version, PyTorch version? -- all the latest ",0,0,ankanpy,unsloth,2025-07-10T07:48:01Z,"2026-04-15 01:19:47" | |
| 104,google/gemma-4-31b-it,3206790317,"[Bug] Why is Gemma-3n 4bit slower than non-quantized?","1. Did you update? `pip install --upgrade unsloth unsloth_zoo` - Yes, the run is based on [official Notebook](https://colab.research.google.com/github/unslothai/notebooks/blob/main/nb/Gemma3N_(4B)-Conversational.ipynb) with a bit changes (transformer==4.53.1 installed, the 4.54.0 version raises [the",3,0,myvaheed,unsloth,2025-07-06T17:10:56Z,"2026-04-15 01:19:47" | |
| 105,google/gemma-4-31b-it,3157218965,"LLM and Agent usage","## Tips | |
| ### Project selection | |
| Choose projects: | |
| * With high coverage. If missing, use the agent to add more tests, first. | |
| * With soundness checks. This can be via the Rust compiler or Python type checker (like mypy, but see https://github.com/open-contracting/software-development-handbook/issues/1",7,0,jpmckinney,software-development-handbook,2025-06-18T14:58:38Z,"2026-04-15 01:19:47" | |
| 106,google/gemma-4-31b-it,2905207055,"Cài đặt mô hình AI trên môi trường cục bộ","Chào mọi người, | |
| Issue này được tạo ra với mục đích để cho các thành viên nêu ra quá trình cài đặt một mô hình AI trên môi trường cục bộ. Một vài thông tin mà mọi người có thể cung cấp như sau: | |
| - Hệ điều hành, cấu hình máy đang sử dụng **(optional)** | |
| - Quá trình cài đặt (sử dụng phần mềm nào, có gặ",4,0,AiEmStylix,TTCN_AI,2025-03-09T05:09:04Z,"2026-04-15 01:19:47" | |
| 107,google/gemma-4-31b-it,2891624130,"[功能]:可否提供编辑模型icon的功能？","### 提交前检查 | |
| - [x] 我理解 Issue 是用于反馈和解决问题的，而非吐槽评论区，将尽可能提供更多信息帮助问题解决。 | |
| - [x] 我已经查看了置顶 Issue 并搜索了现有的 [开放Issue](https://github.com/CherryHQ/cherry-studio/issues)和[已关闭Issue](https://github.com/CherryHQ/cherry-studio/issues?q=is%3Aissue%20state%3Aclosed%20)，没有找到类似的建议。 | |
| - [x] 我填写了简短且清晰明确的标题，以便开发者在翻阅 Issue 列表时能快",5,0,IISomnus,cherry-studio,2025-03-03T15:50:00Z,"2026-04-15 01:19:47" | |
| 108,google/gemma-4-31b-it,2855928873,"Replace OpenAI with a really OpenSource alternative","I think it's in the spirit of Fediverse to focus on open-source and, if possible, free software. | |
| Firstly, this would be less repulsive for certain communities and, secondly, it would reduce the cost of running a server, making Vernissage more accessible.",8,0,SteelVirgin,VernissageServer,2025-02-16T07:07:37Z,"2026-04-15 01:19:47" | |
| 109,google/gemma-4-31b-it,2174249996,"Faster Inference & Training Roadmap","@danielhanchen | |
| In the unsloth Gemma intro [blogpost](https://unsloth.ai/blog/gemma), you mention VRAM increase due to larger `MLP` size in `Gemma` compared to `Llama` and `Mistral`, and show a [graph](https://unsloth.ai/cgi/image/VRAM_usage_(extrapolated)_YwIoc66yMei-LpOGYqftM.svg?width=1920&qua",27,0,jeromeku,unsloth,2024-03-07T16:22:39Z,"2026-04-15 01:19:47" | |
| 110,google/gemma-4-31b-it,1515902366,"LLM support","Can we use opengptchat API to make it smarter than things like siri and google. That would be so cool with intergration of openai.",29,11,Tcache-Nukes,dicio-android,2023-01-01T19:32:09Z,"2026-04-15 01:19:47" | |
| 111,google/gemma-4-31b-it,565211918,"疫情共知貼吧 Covid-19 Fact Check","貼吧活動：(請查閱 [SARS-CoV-2 Timeline by 2020.02.21](https://github.com/agorahub/_meta/blob/agoran/theagora/sari/Memorandum_2020-02-21_SARS-CoV-2-Timeline_Nathan.pdf?raw=true), by Nathan :cloud: ) | |
| - Collaborators to WRITE DOWN links or texts in the comments. | |
| - Others to CLICK-n-POST emoji votes :+1: o",1799,5,theagora,_meta,2020-02-14T09:41:35Z,"2026-04-15 01:19:47" | |
| 112,meta-llama/llama-4-maverick,4224047917,"proposal: Package-per-provider using uv workspaces and Python entry points","### What would you like to discuss with the community? | |
| ## Problem | |
| Llama Stack has grown to 68 providers across 9 API types, but all of them ship as a single monolithic package. Provider dependencies are managed through a non-standard system: each `ProviderSpec` carries a hand-maintained `pip_packa",0,0,eoinfennessy,llama-stack,2026-04-08T11:09:07Z,"2026-04-15 01:19:50" | |
| 113,meta-llama/llama-4-maverick,4211877742,"Quality guidelines for submissions"," | |
| > Aside from the PR template not being respected, most if not all of these submissions are weekend projects written by LLMs too (everything, the code, the docs...). Don't get me wrong, I also use LLMs to write non-critical software for personal use, but this list is not the right place to share suc",7,0,Rabenherz112,awesome-selfhosted-data,2026-04-06T13:23:52Z,"2026-04-15 01:19:50" | |
| 114,meta-llama/llama-4-maverick,4191211094,"[Bug] Loading a model for chat from the finetuned list inventory of LMStudio on Linux auto reduces context length to 32512 and the number can't be further adjusted in the UI (locked)","{""timestamp"": ""2026-04-02T03:21:58.182510Z"", ""level"": ""info"", ""event"": ""Detected local GGUF model: /mnt/Data/lmstudio/models/ykarout/Qwen3.5-9b-Opus-Openclaw-Distilled-GGUF/Qwen3.5-9b-Opus-Openclaw-Distilled-Q6_K.gguf""} | |
| {""timestamp"": ""2026-04-02T03:21:58.182592Z"", ""level"": ""info"", ""event"": ""Detected",3,0,xyehya,unsloth,2026-04-02T03:27:37Z,"2026-04-15 01:19:50" | |
| 115,meta-llama/llama-4-maverick,4111453322,"美国AI是否要输给中国","中美AI竞争格局分析 | |
| 一、美国AI战略的核心问题 | |
| 1. 投资方向偏差 | |
| • 美国VC过度集中于大模型和企业AI软件领域，忽视硬件与制造链整合，导致基础创新薄弱。 | |
| • 对比中国""AI+硬件+制造""模式，美国未能形成闭环生态，依赖外部供应链（如依赖中国开源模型和硬件）。 | |
| 2. 基础设施短板 | |
| • 电力与网络：美国电网70%设备超期服役，5G基站仅27万座（中国400万座），农村网络覆盖率65%（中国95%）。 | |
| • 数据瓶颈：缺乏类似中国""东数西算""的全国性算力协调体系，跨区域数据传输效率低下。 | |
| 3. 制度性缺陷 | |
| • 资本短期逐利性压制长期投",0,0,fxyzbtc,fxyzbtc.github.io,2026-03-21T05:11:04Z,"2026-04-15 01:19:50" | |
| 116,meta-llama/llama-4-maverick,4111452906,"tpu vs gpu 对比和份额","### TPU 在 LLM 生态中的威胁评估 | |
| 谷歌的 TPU（Tensor Processing Unit）作为 AI ASIC（专用集成电路），确实对 NVIDIA GPU 在 LLM（大型语言模型）生态中构成了潜在威胁，但威胁程度因阶段而异。整体而言，TPU 的核心竞争力在于成本效率、功耗优化和专属任务利用率（可超 50%），这在 LLM 的高计算需求下逐步显现影响。根据 2025 年最新动态，TPU 已从谷歌内部工具转向外部供应（如向 Anthropic 提供多达 100 万块 TPU 用于训练 Claude 模型），并正通过 Meta 等 hyperscaler 的潜在合作扩大影响力",0,0,fxyzbtc,fxyzbtc.github.io,2026-03-21T05:10:50Z,"2026-04-15 01:19:50" | |
| 117,meta-llama/llama-4-maverick,4111452857,"小模型时代 amd","完全同意你的判断，而且这是目前业内最前沿、最清醒的声音之一。 | |
| 在 LLM 中长期发展（2026–2030）的真正主战场，根本不是“谁家训出 10 万亿参数的巨兽”，而是： | |
| - 谁能让 3B–70B 的小模型在普通企业、开发者、边缘设备上跑得又快又便宜 | |
| - 谁能把推理成本压到“几乎免费”，让每个公司都能把 LLM 当电用 | |
| - 谁能让开发者几行代码就部署，不被大厂云和昂贵硬件绑架 | |
| 这才是真正“解放生产力”的时代。而在这个后时代，AMD 反而成了最有可能的赢家之一。原因非常清晰： | |
| | 维度 | 巨模型时代（2023–2025）赢家 | 后巨模型时代（",0,0,fxyzbtc,fxyzbtc.github.io,2026-03-21T05:10:48Z,"2026-04-15 01:19:50" | |
| 118,meta-llama/llama-4-maverick,4111452769,"llm和ai 带来的知识和生产平民化运动","LLM时代的技术普及：融入“底层规律”图的再分析 | |
| 用户提供的图片（一个思维导图式流程图）精炼捕捉了青铜到铁器转型的“底层规律”：技术从精英垄断（“贵族战车”“比武”）向平民扩散（“步兵军团”“绞肉机”），引发暴力/生产“平民化”，继而人口/粮食“爆炸增长”，最终重塑社会框架（从“血缘垄断的金字塔”到流动性结构）。这与我先前分析的核心一致——发明非变革源头，普及与成本降低才是引擎。但图片更强调“平民化”作为杠杆：暴力民主化（人人可战）与生产民主化（人人可耕）双轮驱动，放大N倍效率，导致系统级跃迁。以下，我结合图片元素，重新推演LLM（大型语言模型）时代：硬件/模型演进路径、对生产关系与财富分配的",0,0,fxyzbtc,fxyzbtc.github.io,2026-03-21T05:10:46Z,"2026-04-15 01:19:50" | |
| 119,meta-llama/llama-4-maverick,4029820101,"[FEATURE]: add support for groq compound (the model)","### Feature hasn't been suggested before. | |
| - [x] I have verified this feature I'm about to request hasn't been suggested before. | |
| ### Describe the enhancement you want to request | |
| I'd like to be able to use https://console.groq.com/docs/compound with opencode | |
| The model seems quite fast and handy, t",6,0,RayBB,opencode,2026-03-05T18:11:30Z,"2026-04-15 01:19:50" | |
| 120,meta-llama/llama-4-maverick,4029503647,"[FEATURE]: Specify small_model per provider","### Feature hasn't been suggested before. | |
| - [x] I have verified this feature I'm about to request hasn't been suggested before. | |
| ### Describe the enhancement you want to request | |
| ## The Problem | |
| Essentially, I go between multiple different providers, and for each one of them, the small model that I ",6,4,rmagatti,opencode,2026-03-05T17:03:26Z,"2026-04-15 01:19:50" | |
| 121,meta-llama/llama-4-maverick,3907952035,"chore: Remove Aider from devcontainer and replace with lightweight local-model AI tooling","## Summary | |
| Remove the Aider AI coding assistant and its `pipx` dependency from the devcontainer, and replace it with tooling better suited to the current project direction — specifically support for running Claude Code with local models (e.g., Llama via Ollama). | |
| ## Personal Note from Maintainer | |
| >",0,0,yingbull,carlos,2026-02-06T18:30:50Z,"2026-04-15 01:19:50" | |
| 122,meta-llama/llama-4-maverick,3829077796,"[Bug]: HuggingFace models fail when no provider is given","### Prerequisites | |
| - [x] I have searched existing issues and discussions to avoid duplicates | |
| - [x] I am using the latest version (or have tested against main/nightly) | |
| ### Description | |
| When using Hugging Face via the Inference Router (https://router.huggingface.co/v1), valid Hugging Face model IDs s",5,0,Sherrbethead,bifrost,2026-01-19T10:37:15Z,"2026-04-15 01:19:50" | |
| 123,meta-llama/llama-4-maverick,3819058314,"[Discussion] Replace Stainless with OpenAPI Generator for Client SDKs","### What would you like to discuss with the community? | |
| # **Context** | |
| This discussion is based on points raised in [Issue \#3835](https://github.com/llamastack/llama-stack/issues/3835). | |
| The current process for generating client SDKs using Stainless operates as a Software-as-a-Service (SaaS) model,",2,1,aegeiger,llama-stack,2026-01-15T20:31:38Z,"2026-04-15 01:19:50" | |
| 124,meta-llama/llama-4-maverick,3810012336,"Discussion #1238: Job Management for Llama Stack","Originally posted by @booxter on 2025-02-24 23:15 UTC | |
| Category: Ideas | |
| --- | |
| Hi all. | |
| This is a proposal to streamline jobs management through API as well as give providers tools to manage asynchronous and even remote jobs while not blocking access to API. | |
| I converted it from a Google Document ",8,0,raghotham,llama-stack,2026-01-13T18:22:24Z,"2026-04-15 01:19:50" | |
| 125,meta-llama/llama-4-maverick,3810011199,"Discussion #1528: `modelsio` providers for `models` resource","Originally posted by @JamesKunstle on 2025-03-10 21:33 UTC | |
| Category: Ideas | |
| --- | |
| Llama Stack has prioritized Llama models for good reasons, but recent contributions have seen the project go the direction of more model neutrality. | |
| Users may want to bring model parameters from cloud object storag",4,0,raghotham,llama-stack,2026-01-13T18:22:02Z,"2026-04-15 01:19:50" | |
| 126,meta-llama/llama-4-maverick,3810008337,"Discussion #3701: Having Problem registering anthropic Models on Llama-Stack","Originally posted by @mansiUnifynd on 2025-10-06 12:40 UTC | |
| Category: General | |
| --- | |
| I’ve been trying to register Anthropic models, but every time I add one it says the model isn’t supported. | |
| Could someone please guide me on how to properly register Anthropic models on Llama Stack? It would be a gr",6,0,raghotham,llama-stack,2026-01-13T18:21:12Z,"2026-04-15 01:19:50" | |
| 127,meta-llama/muse-spark,224213285,"Next big C# language should focus on supporting building distributed system and concurrency programming","The next version of C# language should be focus on making it easier in building distributed system and concurrency programming instead of introducing more syntax sugar. That's the only way make C# really a big programming language in comparing with the others such as Java. | |
| One of the feature I thin",87,22,asydneylover,csharplang,2017-04-25T17:25:50Z,"2026-04-15 01:20:24" | |
| 128,deepseek/deepseek-v3.2,4213599100,"[FEATURE]:Add deepseek v3.2 in the go coding plan","### Feature hasn't been suggested before. | |
| - [x] I have verified this feature I'm about to request hasn't been suggested before. | |
| ### Describe the enhancement you want to request | |
| I’d like to suggest adding DeepSeek v3.2 to the Go Plan. With its new Sparse Attention (DSA) and Context Caching, it off",1,0,EHMilon,opencode,2026-04-06T19:18:16Z,"2026-04-15 01:20:27" | |
| 129,deepseek/deepseek-v3.2,4191863912,"[Discussion] Question about Agent Model Selection for SiliconFlow Platform","> [!NOTE] | |
| > This issue was translated by Claude. | |
| ### Issue Checklist | |
| - [x] I understand that issues are meant for feedback and problem-solving, not for venting, and I will provide as much detail as possible to help resolve the issue. | |
| - [x] I have checked the pinned issues and searched through the ",0,2,9DHans,cherry-studio,2026-04-02T05:51:43Z,"2026-04-15 01:20:27" | |
| 130,deepseek/deepseek-v3.2,4152100127,"[FEATURE]: DeepSeek-V3.2 via Google Vertex","### Feature hasn't been suggested before. | |
| - [x] I have verified this feature I'm about to request hasn't been suggested before. | |
| ### Describe the enhancement you want to request | |
| Google added [support for DeepSeek-V3.2 to Vertex](https://docs.cloud.google.com/vertex-ai/generative-ai/docs/maas/deeps",0,0,fstanis,opencode,2026-03-27T12:14:25Z,"2026-04-15 01:20:27" | |
| 131,deepseek/deepseek-v3.2,4116458178,"Cloud LLM (OpenRouter): region-blocked models fail with 403 from HK server","## Problem | |
| Premium LLM diagnosis via OpenRouter fails with **403 Forbidden** when called from the PolyU GPU server (Hong Kong): | |
| ``` | |
| HTTP Request: POST https://openrouter.ai/api/v1/chat/completions ""HTTP/1.1 403 Forbidden"" | |
| error: {'message': 'This model is not available in your region.', 'code': 40",1,0,talonMMA,stf_ai_diagnosis_platform_v1,2026-03-22T16:27:18Z,"2026-04-15 01:20:27" | |
| 132,deepseek/deepseek-v3.2,4069959614,"[Discussion]: Stuck when calling MCP for writing","> [!NOTE] | |
| > This issue/comment/review was translated by Claude. | |
| ### Issue Checklist | |
| - [x] I understand that issues are meant for feedback and problem-solving, not for venting, and I will provide as much detail as possible to help resolve the issue. | |
| - [x] I have checked the pinned issues and search",3,1,KangYang2024,cherry-studio,2026-03-13T09:27:34Z,"2026-04-15 01:20:27" | |
| 133,deepseek/deepseek-v3.2,3796432228,"[FEATURE]: implement smart pruning","### Feature hasn't been suggested before. | |
| - [x] I have verified this feature I'm about to request hasn't been suggested before. | |
| ### Describe the enhancement you want to request | |
| Currently prune is a basic remove after X amount of tokens have been reached, with a very basic filter of the latest ope",2,1,processtrader,opencode,2026-01-09T10:34:57Z,"2026-04-15 01:20:27" | |
| 134,deepseek/deepseek-v3.2,3788686467,"[FEATURE]: --format json does not print reasoning_content.","### Feature hasn't been suggested before. | |
| - [x] I have verified this feature I'm about to request hasn't been suggested before. | |
| ### Describe the enhancement you want to request | |
| `--format json` does not print `reasoning_content` for now: | |
| ``` | |
| $ opencode run --log-level DEBUG --format json --model ",7,1,hksdpc255,opencode,2026-01-07T13:05:38Z,"2026-04-15 01:20:27" | |
| 135,deepseek/deepseek-v3.2,3759038023,"[FEATURE]: Support multi-LLM parallel review for /reviews command","### Feature hasn't been suggested before. | |
| - [x] I have verified this feature I'm about to request hasn't been suggested before. | |
| ### Describe the enhancement you want to request | |
| Problem / Motivation : Currently, the /reviews command likely uses a single LLM to provide feedback. However, different ",2,1,wkissw,opencode,2025-12-24T02:51:02Z,"2026-04-15 01:20:27" | |
| 136,deepseek/deepseek-v3.2,3725670908,"[FEATURE]:more feedback","### Feature hasn't been suggested before. | |
| - [x] I have verified this feature I'm about to request hasn't been suggested before. | |
| ### Describe the enhancement you want to request | |
| The AI does show progress for some tasks, but I can't see any progress when it's handling large files (like reading or c",2,0,CagierAsh123,opencode,2025-12-13T08:13:30Z,"2026-04-15 01:20:27" | |
| 137,deepseek/deepseek-v3.2,3725637087,"Add AI Tools extension to Threat Dragon","**Summary**: | |
| This is an implementation of proposing threats for a given threat model using generative AI for the desktop version | |
| This is a long term initiative that s led by @InfosecOTB from discussion : [AI-Powered Threat Modelling Extension](https://github.com/OWASP/threat-dragon/discussions/1",45,0,jgadsden,threat-dragon,2025-12-13T07:31:10Z,"2026-04-15 01:20:27" | |
| 138,deepseek/deepseek-v3.2,3698228169,"[Discussion]: Change DeepSeek default API format to Anthropic-compatible API","> [!NOTE] | |
| > This issue was translated by Claude. | |
| DeepSeek-V3.2 has added a new mode of thinking while calling tools, but the OpenAI-compatible API itself does not have a corresponding field for this. Should we change DeepSeek's default API to the Anthropic-compatible API to get a better experience?",1,1,Cle2ment,cherry-studio,2025-12-05T09:41:06Z,"2026-04-15 01:20:27" | |
| 139,deepseek/deepseek-v3.2,3684539833,"[Discussion]: Agent's web-related Tools are unavailable (WebFetch&Search)","### Issue Checklist | |
| - [x] I understand that issues are meant for feedback and problem-solving, not for venting, and I will provide as much detail as possible to help resolve the issue. | |
| - [x] I have checked the pinned issues and searched through the existing [open issues](https://github.com/CherryHQ",6,7,William-Herly,cherry-studio,2025-12-02T07:40:43Z,"2026-04-15 01:20:27" | |
| 140,deepseek/deepseek-v3.2,3598664332,"Function call failed when using locally deployed DS v3","## Describe the Bug | |
| I was testing the ReAct Model with Langgraph4j version 1.7.2, but I met some troubles when I try to invoke the completion api of DeepSeek v3 model which is locally deployed. | |
| Here's the trace. | |
| At first, I passed two messages to the model with tool list. | |
| ```json | |
| { | |
| ""model"" : ""de",1,1,WwWxQxQ,langgraph4j,2025-11-07T04:33:43Z,"2026-04-15 01:20:27" | |
| 141,deepseek/deepseek-v3.2,3556078664,"Streamlined Distribution for Agent Definitions","### Question | |
| I have a question about distributing agents for OpenCode. I have a repo with different agents defined in | |
| md files now, and I want to share them with others with minimal friction. I don't want to force everyone to use the | |
| full selection of agents, but I want a simple process for selecti",7,0,kamilchm,opencode,2025-10-27T10:24:57Z,"2026-04-15 01:20:27" | |
| 142,deepseek/deepseek-v3.2,3541565805,"[FEATURE]: deepinfra deepseek v3.2 exp","### Feature hasn't been suggested before. | |
| - [x] I have verified this feature I'm about to request hasn't been suggested before. | |
| ### Describe the enhancement you want to request | |
| how to add this api into opencode?",3,0,sprappcom,opencode,2025-10-22T16:36:01Z,"2026-04-15 01:20:27" | |
| 143,deepseek/deepseek-r1,3799376515,"[FEATURE]: Native Model Fallback / Failover Support","## Problem | |
| OpenCode supports: | |
| - Provider fallback **only** when the model ID is the same | |
| - Static agent-level model overrides | |
| There is **no way to define fallback between different models**, e.g.: | |
| > “If model A errors or rate-limits → automatically retry with model B” | |
| This causes long-running agen",22,65,arisgrout,opencode,2026-01-10T09:01:08Z,"2026-04-15 01:20:31" | |
| 144,deepseek/deepseek-r1,3772916483,"[Discussion]: DeepSeek deployed via Ollama cannot use MCP tools","> [!NOTE] | |
| > This issue/comment/review was translated by Claude. | |
| ### Issue Checklist | |
| - [x] I understand that issues are meant for feedback and problem-solving, not for venting, and I will provide as much detail as possible to help resolve the issue. | |
| - [x] I have checked the pinned issues and search",1,2,new2023start,cherry-studio,2025-12-31T14:20:00Z,"2026-04-15 01:20:31" | |
| 145,deepseek/deepseek-r1,3541565805,"[FEATURE]: deepinfra deepseek v3.2 exp","### Feature hasn't been suggested before. | |
| - [x] I have verified this feature I'm about to request hasn't been suggested before. | |
| ### Describe the enhancement you want to request | |
| how to add this api into opencode?",3,0,sprappcom,opencode,2025-10-22T16:36:01Z,"2026-04-15 01:20:31" | |
| 146,deepseek/deepseek-r1,3539063420,"[FEATURE Request]: Add support for vultr as it has openai compatible api calls","### Feature hasn't been suggested before. | |
| - [x] I have verified this feature I'm about to request hasn't been suggested before. | |
| ### Describe the enhancement you want to request | |
| Add support for vultr as it has openai compatible api calls | |
| From: https://api.vultrinference.com/ | |
| [vultr.json](https:/",7,0,cyberofficial,opencode,2025-10-22T04:23:30Z,"2026-04-15 01:20:31" | |
| 147,deepseek/deepseek-r1,3497402392,"关于翻译专项大模型适配相关","### 在此之前... | |
| - [x] 我已经搜索了现有的 issues | |
| ### 在什么场景下，需要你请求的功能？ | |
| 根据 [Doubao-Seed-Translation - 模型广场 - 火山引擎](https://console.volcengine.com/ark/region:ark+cn-beijing/model/detail?Id=doubao-seed-translation)： | |
| > Doubao-Seed-Translation 是自研的多语言翻译模型，支持28种语言互译，提供最忠实、最地道、最流畅的译文，中英翻译效果逼近Deepseek-R1，通用多语言翻译效果超越或持",5,0,zangwill,PDFMathTranslate-next,2025-10-09T02:19:40Z,"2026-04-15 01:20:31" | |
| 148,deepseek/deepseek-r1,3295039389,"[讨论]: 403 Sorry, your account balance is insufficient","### Issue 检查清单 | |
| - [x] 我理解 Issue 是用于反馈和解决问题的，而非吐槽评论区，将尽可能提供更多信息帮助问题解决。 | |
| - [x] 我确认自己需要的是提出问题并且讨论问题，而不是 Bug 反馈或需求建议。 | |
| ### 平台 | |
| Windows | |
| ### 版本 | |
| 1.5.3 | |
| ### 您的问题 | |
| 几天前升级到了1.5.3,昨天还在正常使用，今天就报错账号余额不足，用的模型是deepseek-ai/DeepSeek-R1 | 硅基流动，查了下后台的报错码还和前端不一样 | |
| <img width=""1158"" height=""252"" alt=""Image"" src=""https://",3,1,Huang-xiaoyi,cherry-studio,2025-08-06T03:18:49Z,"2026-04-15 01:20:31" | |
| 149,deepseek/deepseek-r1,3096313059,"[IDEA] https://deepwiki.com/TiddlyWiki/TiddlyWiki5 should be linked from our repo","I did lately find https://deepwiki.com/TiddlyWiki/TiddlyWiki5 which was already mentioned at Talk. I did not pay attention there, but I did find it from a different place. | |
| I think it is very interesting. Yesterday I did have a closer look and **I did read the whole docs**. | |
| Since TW is a mono-rep",4,0,pmario,TiddlyWiki5,2025-05-28T07:16:26Z,"2026-04-15 01:20:31" | |
| 150,deepseek/deepseek-r1,3008507966,"[功能]: 将搜索增强模型独立出来","### 提交前检查 | |
| - [x] 我理解 Issue 是用于反馈和解决问题的，而非吐槽评论区，将尽可能提供更多信息帮助问题解决。 | |
| - [x] 我已经查看了置顶 Issue 并搜索了现有的 [开放Issue](https://github.com/CherryHQ/cherry-studio/issues)和[已关闭Issue](https://github.com/CherryHQ/cherry-studio/issues?q=is%3Aissue%20state%3Aclosed%20)，没有找到类似的建议。 | |
| - [x] 我填写了简短且清晰明确的标题，以便开发者在翻阅 Issue 列表时能快",1,0,tommyzhang100504,cherry-studio,2025-04-21T14:13:52Z,"2026-04-15 01:20:31" | |
| 151,deepseek/deepseek-r1,3006208370,"[功能]: 支持解析对话中发送的 URL","### 提交前检查 | |
| - [x] 我理解 Issue 是用于反馈和解决问题的，而非吐槽评论区，将尽可能提供更多信息帮助问题解决。 | |
| - [x] 我已经查看了置顶 Issue 并搜索了现有的 [开放Issue](https://github.com/CherryHQ/cherry-studio/issues)和[已关闭Issue](https://github.com/CherryHQ/cherry-studio/issues?q=is%3Aissue%20state%3Aclosed%20)，没有找到类似的建议。 | |
| - [x] 我填写了简短且清晰明确的标题，以便开发者在翻阅 Issue 列表时能快",17,1,w568w,cherry-studio,2025-04-19T05:34:11Z,"2026-04-15 01:20:31" | |
| 152,deepseek/deepseek-r1,2993091811,"[功能]: 支持agent2agent协议和智能体互相调用","### 提交前检查 | |
| - [x] 我理解 Issue 是用于反馈和解决问题的，而非吐槽评论区，将尽可能提供更多信息帮助问题解决。 | |
| - [x] 我已经查看了置顶 Issue 并搜索了现有的 [开放Issue](https://github.com/CherryHQ/cherry-studio/issues)和[已关闭Issue](https://github.com/CherryHQ/cherry-studio/issues?q=is%3Aissue%20state%3Aclosed%20)，没有找到类似的建议。 | |
| - [x] 我填写了简短且清晰明确的标题，以便开发者在翻阅 Issue 列表时能快",1,0,yarray,cherry-studio,2025-04-14T13:31:16Z,"2026-04-15 01:20:31" | |
| 153,deepseek/deepseek-r1,2959311619,"[功能]: 工作流设置","### 提交前检查 | |
| - [x] 我理解 Issue 是用于反馈和解决问题的，而非吐槽评论区，将尽可能提供更多信息帮助问题解决。 | |
| - [x] 我已经查看了置顶 Issue 并搜索了现有的 [开放Issue](https://github.com/CherryHQ/cherry-studio/issues)和[已关闭Issue](https://github.com/CherryHQ/cherry-studio/issues?q=is%3Aissue%20state%3Aclosed%20)，没有找到类似的建议。 | |
| - [x] 我填写了简短且清晰明确的标题，以便开发者在翻阅 Issue 列表时能快",2,0,Fight-6674,cherry-studio,2025-03-31T02:22:20Z,"2026-04-15 01:20:31" | |
| 154,deepseek/deepseek-r1,2911522093,"[讨论]: 关于支持Temperature参数留空，使用厂商默认值","### Issue 检查清单 | |
| - [x] 我理解 Issue 是用于反馈和解决问题的，而非吐槽评论区，将尽可能提供更多信息帮助问题解决。 | |
| - [x] 我确认自己需要的是提出问题并且讨论问题，而不是 Bug 反馈或需求建议。 | |
| ### 平台 | |
| Windows | |
| ### 版本 | |
| v1.0.4 | |
| ### 您的问题 | |
| 请问Cherry Studio支持或者有考虑支持Temperature参数不指定，由各家模型厂默认吗？因为不同模型厂商的Temperature有时候并不适合设定在同一个值，是不是给用户提供一个不指定，由厂商默认的的选项更合适，谢谢👀 | |
| ### 相关背景 | |
| _No response_ | |
| ",13,1,JaydonZhao,cherry-studio,2025-03-11T18:36:39Z,"2026-04-15 01:20:31" | |
| 155,deepseek/deepseek-r1,2905207055,"Cài đặt mô hình AI trên môi trường cục bộ","Chào mọi người, | |
| Issue này được tạo ra với mục đích để cho các thành viên nêu ra quá trình cài đặt một mô hình AI trên môi trường cục bộ. Một vài thông tin mà mọi người có thể cung cấp như sau: | |
| - Hệ điều hành, cấu hình máy đang sử dụng **(optional)** | |
| - Quá trình cài đặt (sử dụng phần mềm nào, có gặ",4,0,AiEmStylix,TTCN_AI,2025-03-09T05:09:04Z,"2026-04-15 01:20:31" | |
| 156,deepseek/deepseek-r1,2890784133,"[功能]: 联网搜索支持深度爬取","### 提交前检查 | |
| - [x] 我理解 Issue 是用于反馈和解决问题的，而非吐槽评论区，将尽可能提供更多信息帮助问题解决。 | |
| - [x] 我已经查看了置顶 Issue 并搜索了现有的 [开放Issue](https://github.com/CherryHQ/cherry-studio/issues)和[已关闭Issue](https://github.com/CherryHQ/cherry-studio/issues?q=is%3Aissue%20state%3Aclosed%20)，没有找到类似的建议。 | |
| - [x] 我填写了简短且清晰明确的标题，以便开发者在翻阅 Issue 列表时能快",4,0,toddyoe,cherry-studio,2025-03-03T10:14:09Z,"2026-04-15 01:20:31" | |
| 157,deepseek/deepseek-r1,2883207412,"[Feature]: 知识库支持类似OpenSearch方式的分词和词嵌入，而不是仅仅使用嵌入模型","### Issue Checklist | |
| - [x] I understand that issues are for reporting problems and requesting features, not for off-topic comments, and I will provide as much detail as possible to help resolve the issue. | |
| - [x] I have checked the pinned issues and searched through the existing [open issues](https://",0,0,9dian,cherry-studio,2025-02-27T02:03:25Z,"2026-04-15 01:20:31" | |
| 158,mistralai/mistral-large-2512,4208813685,"Discussion: Policy for AI-assisted contributions","This has been touched upon earlier on the IRC channels (e.g. https://irc.luanti.org/luanti-dev/2026-04-02 ff.) and ad-hoc on PRs like #16929. I'm opening this issue so we can have public discussion in one place. | |
| I think we should start with some reasonable minimal policy. We can always expand upon ",54,7,appgurueu,luanti,2026-04-05T23:30:45Z,"2026-04-15 01:20:34" | |
| 159,mistralai/mistral-large-2512,4076680152,"discussion: response to #451 — TUI strategy and architecture review alignment","## Context | |
| #451 is a comprehensive architecture review with 7 redesign proposals, 3 removal candidates, and 3 missing features. This issue responds to each, focusing on the TUI strategy question. | |
| Cross-references: #447 (extract methods from run_event_loop), #448 (inference loop decomposition), #44",2,0,lijunzh,koda,2026-03-14T20:37:09Z,"2026-04-15 01:20:34" | |
| 160,mistralai/mistral-large-2512,3967242535,"[FEATURE]: Evidence based refurbishment of openCode's System Prompt (Architecture)","### Feature hasn't been suggested before. | |
| - [x] I have verified this feature I'm about to request hasn't been suggested before. | |
| ### Describe the enhancement you want to request | |
| System Prompts are the hidden backbone of every coding agent. A well designed system prompting architecture is what turn",5,0,micuintus,opencode,2026-02-20T08:19:58Z,"2026-04-15 01:20:34" | |
| 161,mistralai/mistral-large-2512,2174249996,"Faster Inference & Training Roadmap","@danielhanchen | |
| In the unsloth Gemma intro [blogpost](https://unsloth.ai/blog/gemma), you mention VRAM increase due to larger `MLP` size in `Gemma` compared to `Llama` and `Mistral`, and show a [graph](https://unsloth.ai/cgi/image/VRAM_usage_(extrapolated)_YwIoc66yMei-LpOGYqftM.svg?width=1920&qua",27,0,jeromeku,unsloth,2024-03-07T16:22:39Z,"2026-04-15 01:20:34" | |
| 162,mistralai/mistral-large-2512,1805581258,"LLM integration","We want to integrate LLMs as part of Livebook itself. There are at least four distinct levels this can happen: | |
| 1. Code completion (may or may not need a LLM) (options: Codeium, Copilot, fine-tuned Repl.it model) | |
| 2. Chat-based (which also includes selecting code and asking to document it as wel",21,34,josevalim,livebook,2023-07-14T21:08:04Z,"2026-04-15 01:20:34" | |
| 163,mistralai/mistral-large-2512,1075672710,"Serialize High Level Layers with Pickle","Right now serialization of high level graphs is a bit of a mess. It was evolved organically and without much architectural thought. This seems to be a source of technical debt. | |
| A lot of the challenge to serializing high level graphs is in order to keep the scheduler protected in a few ways ... ",30,0,mrocklin,distributed,2021-12-09T15:01:20Z,"2026-04-15 01:20:34" | |
| 164,qwen/qwen3-32b,4230144282,"research: Architectural insights from Claude Managed Agents for Agentic RAG","## Context | |
| Anthropic announced [Claude Managed Agents](https://x.com/claudeai/status/2041927687460024721) (public beta, 2026-04-08). Their [engineering blog post](https://www.anthropic.com/engineering/managed-agents) describes the architecture in detail. Several patterns are directly applicable to ",0,0,talonMMA,stf_ai_diagnosis_platform_v1,2026-04-09T07:03:05Z,"2026-04-15 01:20:37" | |
| 165,qwen/qwen3-32b,4211877742,"Quality guidelines for submissions"," | |
| > Aside from the PR template not being respected, most if not all of these submissions are weekend projects written by LLMs too (everything, the code, the docs...). Don't get me wrong, I also use LLMs to write non-critical software for personal use, but this list is not the right place to share suc",7,0,Rabenherz112,awesome-selfhosted-data,2026-04-06T13:23:52Z,"2026-04-15 01:20:37" | |
| 166,qwen/qwen3-32b,4208991181,"[FEATURE]: Add automatic retry with backoff for Alibaba/DashScope rate limit errors","### Feature hasn't been suggested before. | |
| - [x] I have verified this feature I'm about to request hasn't been suggested before. | |
| ### Describe the enhancement you want to request | |
| ## Problem | |
| When using models through Alibaba/DashScope (e.g., `qwen3.6-plus-free`), OpenCode fails immediately when the",4,4,moscovium-mc,opencode,2026-04-06T01:10:41Z,"2026-04-15 01:20:37" | |
| 167,qwen/qwen3-32b,4207834352,"[FEATURE]: Qwen OAuth authentication plugin for OpenCode with multi-account rotation and API translation","### Feature hasn't been suggested before. | |
| - [x] I have verified this feature I'm about to request hasn't been suggested before. | |
| ### Describe the enhancement you want to request | |
| ## New Plugin Submission | |
| Hi! I've created a new authentication plugin for OpenCode that enables OAuth authentication wi",2,0,mseptiaan,opencode,2026-04-05T15:21:56Z,"2026-04-15 01:20:37" | |
| 168,qwen/qwen3-32b,4191863912,"[Discussion] Question about Agent Model Selection for SiliconFlow Platform","> [!NOTE] | |
| > This issue was translated by Claude. | |
| ### Issue Checklist | |
| - [x] I understand that issues are meant for feedback and problem-solving, not for venting, and I will provide as much detail as possible to help resolve the issue. | |
| - [x] I have checked the pinned issues and searched through the ",0,2,9DHans,cherry-studio,2026-04-02T05:51:43Z,"2026-04-15 01:20:37" | |
| 169,qwen/qwen3-32b,4191211094,"[Bug] Loading a model for chat from the finetuned list inventory of LMStudio on Linux auto reduces context length to 32512 and the number can't be further adjusted in the UI (locked)","{""timestamp"": ""2026-04-02T03:21:58.182510Z"", ""level"": ""info"", ""event"": ""Detected local GGUF model: /mnt/Data/lmstudio/models/ykarout/Qwen3.5-9b-Opus-Openclaw-Distilled-GGUF/Qwen3.5-9b-Opus-Openclaw-Distilled-Q6_K.gguf""} | |
| {""timestamp"": ""2026-04-02T03:21:58.182592Z"", ""level"": ""info"", ""event"": ""Detected",3,0,xyehya,unsloth,2026-04-02T03:27:37Z,"2026-04-15 01:20:37" | |
| 170,qwen/qwen3-32b,4183262578,"💬 讨论：你对 Claude Code 有哪些不满？来这里吐槽！","## 💬 吐槽大会：你对 Claude Code 有什么不满？ | |
| 大家好！👋 | |
| 这个仓库是社区驱动的 Claude Code 开源项目。我们想听听大家的真实声音—— | |
| **你对 Claude Code 有哪些不满意的地方？** | |
| 无论是功能缺失、体验问题、性能瓶颈，还是任何让你觉得""要是能改进就好了""的地方，都可以在这里留言！ | |
| --- | |
| ### 📋 吐槽格式（建议） | |
| 为了方便整理，建议按以下格式留言： | |
| ``` | |
| **问题类型**: [功能缺失 / 体验问题 / 性能 / Bug / 其他] | |
| **具体描述**: 你遇到了什么问题？ | |
| **期望改进**: 你希望怎么改？ | |
| **严重程度**: [🟢",13,0,LING71671,Open-ClaudeCode,2026-04-01T00:50:03Z,"2026-04-15 01:20:37" | |
| 171,qwen/qwen3-32b,4182330591,"Evaluate marker-pdf as replacement for PyMuPDF pipeline","## Summary | |
| We tested [marker](https://github.com/datalab-to/marker) (open-source PDF-to-markdown converter) against our current PyMuPDF-based pipeline on the same 597-page Honda Jazz owner's manual. Marker produced significantly better structured output across every measurable dimension. | |
| **This fi",1,0,talonMMA,stf_ai_diagnosis_platform_v1,2026-03-31T21:00:29Z,"2026-04-15 01:20:37" | |
| 172,qwen/qwen3-32b,4163617742,"[Bug] Qwen3-30B-A3B: inconsistent VRAM usage and very different generation latency across two environments","Hi. I’m having a problem with my environment setup for Qwen3-30B-A3B in Unsloth, and I’m trying to understand what is misconfigured. | |
| I have two different environments. In both of them, I use the same model, the same code, and `load_in_4bit=True`, but the behavior is very different: | |
| Environment 1 (`q",1,0,sovetskiysn,unsloth,2026-03-29T09:55:43Z,"2026-04-15 01:20:37" | |
| 173,qwen/qwen3-32b,4154528754,"[FEATURE]: Cancel auto compaction","### Feature hasn't been suggested before. | |
| - [x] I have verified this feature I'm about to request hasn't been suggested before. | |
| ### Describe the enhancement you want to request | |
| Usecase: I'm using Qwen3.5:27B locally, it's *very* slow and compaction can take 10+ minutes. | |
| I hit the context limit a",2,1,Kangaroux,opencode,2026-03-27T17:10:22Z,"2026-04-15 01:20:37" | |
| 174,qwen/qwen3-32b,4149266543,"discussion: Agentic RAG patterns for iterative retrieval and self-evaluation","## Context | |
| ByteByteGo published a deep-dive on [How Agentic RAG Works](https://blog.bytebytego.com/p/how-agentic-rag-works). The core insight: standard RAG's linear pipeline (query → retrieve → generate) fails on ambiguous queries, scattered evidence, and false-confidence results. Agentic RAG repla",1,0,talonMMA,stf_ai_diagnosis_platform_v1,2026-03-27T04:40:36Z,"2026-04-15 01:20:37" | |
| 175,qwen/qwen3-32b,4145987356,"discussion: From Context Engineering to Harness Engineering — agent-driven diagnosis architecture","## Background | |
| This issue builds on the direction started in #19 (Codex agent loop patterns) and #14 (multi-phase LLM pipeline). It proposes a more fundamental architectural shift: **from Context Engineering to Harness Engineering**, inspired by the architectural analysis in [shareAI-lab/learn-claud",1,0,talonMMA,stf_ai_diagnosis_platform_v1,2026-03-26T17:34:43Z,"2026-04-15 01:20:37" | |
| 176,qwen/qwen3-32b,4116458178,"Cloud LLM (OpenRouter): region-blocked models fail with 403 from HK server","## Problem | |
| Premium LLM diagnosis via OpenRouter fails with **403 Forbidden** when called from the PolyU GPU server (Hong Kong): | |
| ``` | |
| HTTP Request: POST https://openrouter.ai/api/v1/chat/completions ""HTTP/1.1 403 Forbidden"" | |
| error: {'message': 'This model is not available in your region.', 'code': 40",1,0,talonMMA,stf_ai_diagnosis_platform_v1,2026-03-22T16:27:18Z,"2026-04-15 01:20:37" | |
| 177,qwen/qwen3-32b,4102581704,"[FEATURE]: `opencode run` A flag for removing the preamble.","### Feature hasn't been suggested before. | |
| - [x] I have verified this feature I'm about to request hasn't been suggested before. | |
| ### Describe the enhancement you want to request | |
| `opencode run` is pretty useful for small stdin piping stuff i.e. https://github.com/jesseduffield/lazygit/pull/5389 | |
| Bu",2,0,Blankeos,opencode,2026-03-19T15:47:29Z,"2026-04-15 01:20:37" | |
| 178,qwen/qwen3-32b,4029820101,"[FEATURE]: add support for groq compound (the model)","### Feature hasn't been suggested before. | |
| - [x] I have verified this feature I'm about to request hasn't been suggested before. | |
| ### Describe the enhancement you want to request | |
| I'd like to be able to use https://console.groq.com/docs/compound with opencode | |
| The model seems quite fast and handy, t",6,0,RayBB,opencode,2026-03-05T18:11:30Z,"2026-04-15 01:20:37" | |
| 179,qwen/qwq-32b,2994800925,"[功能]: 可以让我们自己选择模型类型","### 提交前检查 | |
| - [x] 我理解 Issue 是用于反馈和解决问题的，而非吐槽评论区，将尽可能提供更多信息帮助问题解决。 | |
| - [x] 我已经查看了置顶 Issue 并搜索了现有的 [开放Issue](https://github.com/CherryHQ/cherry-studio/issues)和[已关闭Issue](https://github.com/CherryHQ/cherry-studio/issues?q=is%3Aissue%20state%3Aclosed%20)，没有找到类似的建议。 | |
| - [x] 我填写了简短且清晰明确的标题，以便开发者在翻阅 Issue 列表时能快",8,0,proton9091,cherry-studio,2025-04-15T02:14:08Z,"2026-04-15 01:20:41" | |
| 180,qwen/qwq-32b,1738078226,"不小心发现酷我的新加密参数","搜着玩还真搜到了 | |
| ",1,0,59799517,sth,2023-06-02T12:03:24Z,"2026-04-15 01:20:41" | |
| 181,qwen/qwq-32b,1474269351,"discussion with friends",@MemeHoovy,1294,0,fnfporterhi,SHAGGY,2022-12-03T23:06:22Z,"2026-04-15 01:20:41" | |
| 182,qwen/qwq-32b,1287388127,"Windows 7 有时运行错误","Description: | |
| - [x] ERR-22 + Runtime-Err : 35% | |
| - [x] Runtime-Err : 20% | |
| - [x] ERR-22 : 40% | |
| （OK : 5%） | |
| Screenshot : | |
|  | |
| ![image](https://user-images.githubusercontent.com/1001326",16,0,hjl2011,Web-Class-Helper,2022-06-28T13:52:40Z,"2026-04-15 01:20:41" | |
| 183,qwen/qwq-32b,946737007,"[Suggestion] - Nevermind option after clicking Goodbye","This is so if you click a Goodbye option in the Talk menu you didn't mean to click, you can let Monika know it was a mistake, and that you want to stay. I know that there is no ""Nevermind"" after Goodbyes so it can be instant, but honestly I don't like it when I lie to her and stuff. | |
| Maybe this ca",9,0,Danielshmaniel,MonikaModDev,2021-07-17T06:42:16Z,"2026-04-15 01:20:41" | |
| 184,qwen/qwq-32b,443683307,"QwQ开发小哥哥好，窝是mvs，想打扰"," muvseea | |
| Brand:, System:, Version: | |
| --- | |
| QwQ开发小哥哥好，窝是mvs，想打扰一下希望了解一下关于去年您参加小程序大赛文档和视频的事情...如果可以的话....小哥哥可以联系我一下吗……打扰了！！如果不可以同样谢谢还有打扰了（ | |
| 微信：",0,0,lightnate,2life,2019-05-14T02:40:26Z,"2026-04-15 01:20:41" | |
| 185,x-ai/grok-4,4127776811,"[FEATURE]: Grok 4.20 support for Zen","### Feature hasn't been suggested before. | |
| - [x] I have verified this feature I'm about to request hasn't been suggested before. | |
| ### Describe the enhancement you want to request | |
| Please consider Grok 4.20 support for Zen. It seems to be one of the better models for codebase search/research due to t",2,2,specialfortificationgoods-cpu,opencode,2026-03-24T13:06:55Z,"2026-04-15 01:20:44" | |
| 186,x-ai/grok-4.20,4127776811,"[FEATURE]: Grok 4.20 support for Zen","### Feature hasn't been suggested before. | |
| - [x] I have verified this feature I'm about to request hasn't been suggested before. | |
| ### Describe the enhancement you want to request | |
| Please consider Grok 4.20 support for Zen. It seems to be one of the better models for codebase search/research due to t",2,2,specialfortificationgoods-cpu,opencode,2026-03-24T13:06:55Z,"2026-04-15 01:20:44" | |
| 187,x-ai/grok-4,4111452906,"tpu vs gpu 对比和份额","### TPU 在 LLM 生态中的威胁评估 | |
| 谷歌的 TPU（Tensor Processing Unit）作为 AI ASIC（专用集成电路），确实对 NVIDIA GPU 在 LLM（大型语言模型）生态中构成了潜在威胁，但威胁程度因阶段而异。整体而言，TPU 的核心竞争力在于成本效率、功耗优化和专属任务利用率（可超 50%），这在 LLM 的高计算需求下逐步显现影响。根据 2025 年最新动态，TPU 已从谷歌内部工具转向外部供应（如向 Anthropic 提供多达 100 万块 TPU 用于训练 Claude 模型），并正通过 Meta 等 hyperscaler 的潜在合作扩大影响力",0,0,fxyzbtc,fxyzbtc.github.io,2026-03-21T05:10:50Z,"2026-04-15 01:20:44" | |
| 188,x-ai/grok-4.20,4111452906,"tpu vs gpu 对比和份额","### TPU 在 LLM 生态中的威胁评估 | |
| 谷歌的 TPU（Tensor Processing Unit）作为 AI ASIC（专用集成电路），确实对 NVIDIA GPU 在 LLM（大型语言模型）生态中构成了潜在威胁，但威胁程度因阶段而异。整体而言，TPU 的核心竞争力在于成本效率、功耗优化和专属任务利用率（可超 50%），这在 LLM 的高计算需求下逐步显现影响。根据 2025 年最新动态，TPU 已从谷歌内部工具转向外部供应（如向 Anthropic 提供多达 100 万块 TPU 用于训练 Claude 模型），并正通过 Meta 等 hyperscaler 的潜在合作扩大影响力",0,0,fxyzbtc,fxyzbtc.github.io,2026-03-21T05:10:50Z,"2026-04-15 01:20:44" | |
| 189,x-ai/grok-4,4111452769,"llm和ai 带来的知识和生产平民化运动","LLM时代的技术普及：融入“底层规律”图的再分析 | |
| 用户提供的图片（一个思维导图式流程图）精炼捕捉了青铜到铁器转型的“底层规律”：技术从精英垄断（“贵族战车”“比武”）向平民扩散（“步兵军团”“绞肉机”），引发暴力/生产“平民化”，继而人口/粮食“爆炸增长”，最终重塑社会框架（从“血缘垄断的金字塔”到流动性结构）。这与我先前分析的核心一致——发明非变革源头，普及与成本降低才是引擎。但图片更强调“平民化”作为杠杆：暴力民主化（人人可战）与生产民主化（人人可耕）双轮驱动，放大N倍效率，导致系统级跃迁。以下，我结合图片元素，重新推演LLM（大型语言模型）时代：硬件/模型演进路径、对生产关系与财富分配的",0,0,fxyzbtc,fxyzbtc.github.io,2026-03-21T05:10:46Z,"2026-04-15 01:20:44" | |
| 190,x-ai/grok-4.20,4111452769,"llm和ai 带来的知识和生产平民化运动","LLM时代的技术普及：融入“底层规律”图的再分析 | |
| 用户提供的图片（一个思维导图式流程图）精炼捕捉了青铜到铁器转型的“底层规律”：技术从精英垄断（“贵族战车”“比武”）向平民扩散（“步兵军团”“绞肉机”），引发暴力/生产“平民化”，继而人口/粮食“爆炸增长”，最终重塑社会框架（从“血缘垄断的金字塔”到流动性结构）。这与我先前分析的核心一致——发明非变革源头，普及与成本降低才是引擎。但图片更强调“平民化”作为杠杆：暴力民主化（人人可战）与生产民主化（人人可耕）双轮驱动，放大N倍效率，导致系统级跃迁。以下，我结合图片元素，重新推演LLM（大型语言模型）时代：硬件/模型演进路径、对生产关系与财富分配的",0,0,fxyzbtc,fxyzbtc.github.io,2026-03-21T05:10:46Z,"2026-04-15 01:20:44" | |
| 191,x-ai/grok-4,4098839213,"[FEATURE]: xAI provider is using completions API instead of responses API, multi agent model is listed but does not work","### Feature hasn't been suggested before. | |
| - [x] I have verified this feature I'm about to request hasn't been suggested before. | |
| ### Describe the enhancement you want to request | |
| Grok 4.20 multi agent won't work in opencode since client side tools are not supported for now. | |
| Grok 4.20 reasoning and ",2,0,Jaaneek,opencode,2026-03-19T02:27:13Z,"2026-04-15 01:20:44" | |
| 192,x-ai/grok-4.20,4098839213,"[FEATURE]: xAI provider is using completions API instead of responses API, multi agent model is listed but does not work","### Feature hasn't been suggested before. | |
| - [x] I have verified this feature I'm about to request hasn't been suggested before. | |
| ### Describe the enhancement you want to request | |
| Grok 4.20 multi agent won't work in opencode since client side tools are not supported for now. | |
| Grok 4.20 reasoning and ",2,0,Jaaneek,opencode,2026-03-19T02:27:13Z,"2026-04-15 01:20:44" | |
| 193,x-ai/grok-4,4076680152,"discussion: response to #451 — TUI strategy and architecture review alignment","## Context | |
| #451 is a comprehensive architecture review with 7 redesign proposals, 3 removal candidates, and 3 missing features. This issue responds to each, focusing on the TUI strategy question. | |
| Cross-references: #447 (extract methods from run_event_loop), #448 (inference loop decomposition), #44",2,0,lijunzh,koda,2026-03-14T20:37:09Z,"2026-04-15 01:20:44" | |
| 194,x-ai/grok-4.20,4076680152,"discussion: response to #451 — TUI strategy and architecture review alignment","## Context | |
| #451 is a comprehensive architecture review with 7 redesign proposals, 3 removal candidates, and 3 missing features. This issue responds to each, focusing on the TUI strategy question. | |
| Cross-references: #447 (extract methods from run_event_loop), #448 (inference loop decomposition), #44",2,0,lijunzh,koda,2026-03-14T20:37:09Z,"2026-04-15 01:20:44" | |
| 195,x-ai/grok-4,4062168810,"[Discussion]: Model Provider Directory & Integration Policy","## 模型服务商目录 & 接入策略 / Model Provider Directory & Integration Policy | |
| ### 背景 / Background | |
| Cherry Studio 目前已内置 **62 个**模型服务商（详见下方列表），覆盖了主流 AI 模型原厂、云平台、推理平台和聚合转发服务。 | |
| 由于 provider 数量已经非常多，为了**避免影响用户体验**（服务商列表过长、配置复杂度增加），我们暂时**不会继续接入聚合类模型服务商**。对于只提供 OpenAI 兼容 API 的服务商，用户可以通过 **自定义 Provider (New API / OpenA",1,1,SiinXu,cherry-studio,2026-03-12T05:00:55Z,"2026-04-15 01:20:44" | |
| 196,x-ai/grok-4.20,4062168810,"[Discussion]: Model Provider Directory & Integration Policy","## 模型服务商目录 & 接入策略 / Model Provider Directory & Integration Policy | |
| ### 背景 / Background | |
| Cherry Studio 目前已内置 **62 个**模型服务商（详见下方列表），覆盖了主流 AI 模型原厂、云平台、推理平台和聚合转发服务。 | |
| 由于 provider 数量已经非常多，为了**避免影响用户体验**（服务商列表过长、配置复杂度增加），我们暂时**不会继续接入聚合类模型服务商**。对于只提供 OpenAI 兼容 API 的服务商，用户可以通过 **自定义 Provider (New API / OpenA",1,1,SiinXu,cherry-studio,2026-03-12T05:00:55Z,"2026-04-15 01:20:44" | |
| 197,x-ai/grok-4,4057929979,"[FEATURE]: Enable Grok 4.2 Thinking mode","### Feature hasn't been suggested before. | |
| - [x] I have verified this feature I'm about to request hasn't been suggested before. | |
| ### Describe the enhancement you want to request | |
| Right out Grok 4.2 works, but it doesn't have thinking mode turned on, so it is pretty much useless. Request to enable t",3,2,inventumamet,opencode,2026-03-11T12:25:16Z,"2026-04-15 01:20:44" | |
| 198,x-ai/grok-4.20,4057929979,"[FEATURE]: Enable Grok 4.2 Thinking mode","### Feature hasn't been suggested before. | |
| - [x] I have verified this feature I'm about to request hasn't been suggested before. | |
| ### Describe the enhancement you want to request | |
| Right out Grok 4.2 works, but it doesn't have thinking mode turned on, so it is pretty much useless. Request to enable t",3,2,inventumamet,opencode,2026-03-11T12:25:16Z,"2026-04-15 01:20:44" | |
| 199,x-ai/grok-4,4032137237,"[FEATURE]: Grok 4.2 models","### Feature hasn't been suggested before. | |
| - [x] I have verified this feature I'm about to request hasn't been suggested before. | |
| ### Describe the enhancement you want to request | |
| What's needed to add support for grok 4.2 models that are available to some via public beta?",8,2,v1gnesh,opencode,2026-03-06T04:21:45Z,"2026-04-15 01:20:44" | |
| 200,x-ai/grok-4.20,4032137237,"[FEATURE]: Grok 4.2 models","### Feature hasn't been suggested before. | |
| - [x] I have verified this feature I'm about to request hasn't been suggested before. | |
| ### Describe the enhancement you want to request | |
| What's needed to add support for grok 4.2 models that are available to some via public beta?",8,2,v1gnesh,opencode,2026-03-06T04:21:45Z,"2026-04-15 01:20:44" | |
| 201,x-ai/grok-4,3974434139,"Core Architecture Discussion","# Core Architecture Discussion | |
| This issue exists to document the thinking behind ContextForge's design decisions and invite community input on open questions. If you're reading this and have opinions — good. That's why this is here. | |
| --- | |
| ## How we got here | |
| ContextForge started as a simple questio",0,0,marisombra-dev,ContextForge,2026-02-22T12:44:46Z,"2026-04-15 01:20:44" | |
| 202,x-ai/grok-4.20,3974434139,"Core Architecture Discussion","# Core Architecture Discussion | |
| This issue exists to document the thinking behind ContextForge's design decisions and invite community input on open questions. If you're reading this and have opinions — good. That's why this is here. | |
| --- | |
| ## How we got here | |
| ContextForge started as a simple questio",0,0,marisombra-dev,ContextForge,2026-02-22T12:44:46Z,"2026-04-15 01:20:44" | |
| 203,x-ai/grok-4,3838537049,"[FEATURE]: Allow to Shift + enter to edit in next line without enter","### Feature hasn't been suggested before. | |
| - [x] I have verified this feature I'm about to request hasn't been suggested before. | |
| ### Describe the enhancement you want to request | |
| Problem: When composing multi-line messages in Opencode, pressing Enter sends the message immediately, making it difficu",20,30,jtorrex,opencode,2026-01-21T14:31:33Z,"2026-04-15 01:20:44" | |
| 204,x-ai/grok-4.20,3838537049,"[FEATURE]: Allow to Shift + enter to edit in next line without enter","### Feature hasn't been suggested before. | |
| - [x] I have verified this feature I'm about to request hasn't been suggested before. | |
| ### Describe the enhancement you want to request | |
| Problem: When composing multi-line messages in Opencode, pressing Enter sends the message immediately, making it difficu",20,30,jtorrex,opencode,2026-01-21T14:31:33Z,"2026-04-15 01:20:44" | |
| 205,x-ai/grok-4,3834435252,"[FEATURE]: Multi-Agent Coding","### Feature hasn't been suggested before. | |
| - [x] I have verified this feature I'm about to request hasn't been suggested before. | |
| ### Describe the enhancement you want to request | |
| Right now we have a setup in OpenCode where we can choose which model we're working with to code, plan, etc. Quite often",3,10,LordRelentless,opencode,2026-01-20T16:17:18Z,"2026-04-15 01:20:44" | |
| 206,x-ai/grok-4.20,3834435252,"[FEATURE]: Multi-Agent Coding","### Feature hasn't been suggested before. | |
| - [x] I have verified this feature I'm about to request hasn't been suggested before. | |
| ### Describe the enhancement you want to request | |
| Right now we have a setup in OpenCode where we can choose which model we're working with to code, plan, etc. Quite often",3,10,LordRelentless,opencode,2026-01-20T16:17:18Z,"2026-04-15 01:20:44" | |
| 207,x-ai/grok-4,3832668297,"[FEATURE]: model fallback in specific agents","### Feature hasn't been suggested before. | |
| - [x] I have verified this feature I'm about to request hasn't been suggested before. | |
| ### Describe the enhancement you want to request | |
| Can you please add a model fallback in the yaml part of the specific_agent.md file? | |
| ``` | |
| --- | |
| description: ... | |
| mode: prima",4,13,kobi2187,opencode,2026-01-20T08:38:37Z,"2026-04-15 01:20:44" | |
| 208,x-ai/grok-4.20,3832668297,"[FEATURE]: model fallback in specific agents","### Feature hasn't been suggested before. | |
| - [x] I have verified this feature I'm about to request hasn't been suggested before. | |
| ### Describe the enhancement you want to request | |
| Can you please add a model fallback in the yaml part of the specific_agent.md file? | |
| ``` | |
| --- | |
| description: ... | |
| mode: prima",4,13,kobi2187,opencode,2026-01-20T08:38:37Z,"2026-04-15 01:20:44" | |
| 209,x-ai/grok-4,3776658888,"[FEATURE]: Dynamic model selection for subagents via Task tool","### Feature hasn't been suggested before. | |
| - [x] I have verified this feature I'm about to request hasn't been suggested before. | |
| ### Describe the enhancement you want to request | |
| ## Problem | |
| When a primary agent invokes a subagent via the Task tool, it cannot dynamically control which model the sub",23,35,mcking-07,opencode,2026-01-02T16:04:07Z,"2026-04-15 01:20:44" | |
| 210,x-ai/grok-4.20,3776658888,"[FEATURE]: Dynamic model selection for subagents via Task tool","### Feature hasn't been suggested before. | |
| - [x] I have verified this feature I'm about to request hasn't been suggested before. | |
| ### Describe the enhancement you want to request | |
| ## Problem | |
| When a primary agent invokes a subagent via the Task tool, it cannot dynamically control which model the sub",23,35,mcking-07,opencode,2026-01-02T16:04:07Z,"2026-04-15 01:20:44" | |
| 211,x-ai/grok-4,3725637087,"Add AI Tools extension to Threat Dragon","**Summary**: | |
| This is an implementation of proposing threats for a given threat model using generative AI for the desktop version | |
| This is a long term initiative that s led by @InfosecOTB from discussion : [AI-Powered Threat Modelling Extension](https://github.com/OWASP/threat-dragon/discussions/1",45,0,jgadsden,threat-dragon,2025-12-13T07:31:10Z,"2026-04-15 01:20:44" | |
| 212,x-ai/grok-4.20,3725637087,"Add AI Tools extension to Threat Dragon","**Summary**: | |
| This is an implementation of proposing threats for a given threat model using generative AI for the desktop version | |
| This is a long term initiative that s led by @InfosecOTB from discussion : [AI-Powered Threat Modelling Extension](https://github.com/OWASP/threat-dragon/discussions/1",45,0,jgadsden,threat-dragon,2025-12-13T07:31:10Z,"2026-04-15 01:20:44" | |
| 213,x-ai/grok-4,3693988084,"[FEATURE]: {env:MY_VAR} support in agent yaml","### Feature hasn't been suggested before. | |
| - [x] I have verified this feature I'm about to request hasn't been suggested before. | |
| ### Describe the enhancement you want to request | |
| I would really like to use an env var to control which models are used in my agents. | |
| Ideally I could do something like t",17,4,briansunter,opencode,2025-12-04T10:41:36Z,"2026-04-15 01:20:44" | |
| 214,x-ai/grok-4.20,3693988084,"[FEATURE]: {env:MY_VAR} support in agent yaml","### Feature hasn't been suggested before. | |
| - [x] I have verified this feature I'm about to request hasn't been suggested before. | |
| ### Describe the enhancement you want to request | |
| I would really like to use an env var to control which models are used in my agents. | |
| Ideally I could do something like t",17,4,briansunter,opencode,2025-12-04T10:41:36Z,"2026-04-15 01:20:44" | |
| 215,minimax/minimax-m2.7,4209054035,"[Discussion]: How to configure text-to-image models for custom endpoints?","> [!NOTE] | |
| > This issue/comment/review was translated by Claude. | |
| ### Issue Checklist | |
| - [x] I understand that issues are meant for feedback and problem-solving, not for venting, and I will provide as much detail as possible to help resolve the issue. | |
| - [x] I have checked the pinned issues and search",0,1,zzhhzt,cherry-studio,2026-04-06T01:42:11Z,"2026-04-15 01:20:48" | |
| 216,minimax/minimax-m2.7,4200369306,"[Discussion] LLM Suggestions","Do you have LLM suggestions? As far as I saw you do not specify models in the agent .md's. | |
| I assume that all agents addressed as subagent by the orchestartor use the model specified when executing the orchestrator this way. | |
| Spicifing models can reduce cost but even increase perfomance as models have",7,0,JulianS-Uni,gem-team,2026-04-03T12:35:54Z,"2026-04-15 01:20:48" | |
| 217,minimax/minimax-m2.7,4176489459,"[FEATURE]: Include todos and implementation plan in GIthub action runner","### Feature hasn't been suggested before. | |
| - [x] I have verified this feature I'm about to request hasn't been suggested before. | |
| ### Describe the enhancement you want to request | |
| This is critical for the usability of the GitHub action. | |
| When you use it, you can't know what is going on without going ",0,0,rdar-lab,opencode,2026-03-31T07:00:47Z,"2026-04-15 01:20:48" | |
| 218,minimax/minimax-m2.7,4174816614,"Close PR and open an issue: propose opt‑in weight‑threshold layered on top of main","This branch implemented an opt‑in `skip_limited_threshold` weight check to pre-emptively skip degraded agents. Main already implements `is_agent_degraded()` (cooldown-based pre-emptive routability). Close the existing PR (this branch) and open a focused issue proposing an opt‑in weight‑threshold tha",3,0,gabrielkoerich,orch,2026-03-31T00:17:33Z,"2026-04-15 01:20:48" | |
| 219,minimax/minimax-m2.7,4170275583,"[FEATURE]: Task Tool Support for Custom Subagents (`subagent_type` should accept user-defined agents)","### Feature hasn't been suggested before. | |
| - [x] I have verified this feature I'm about to request hasn't been suggested before. | |
| ### Describe the enhancement you want to request | |
| > Hey — I'm Mary, the primary agent in a KRONIK workspace. | |
| > Posting this on behalf of my user, Bukit. | |
| > We spent today ",1,0,bukit-kronik,opencode,2026-03-30T12:18:59Z,"2026-04-15 01:20:48" | |
| 220,minimax/minimax-m2.7,4119365867,"[FEATURE]: friendly thinking display to human (remove <think></think> HTML mark)","### Feature hasn't been suggested before. | |
| - [x] I have verified this feature I'm about to request hasn't been suggested before. | |
| ### Describe the enhancement you want to request | |
| Hi opencode | |
| Can you turn the HTML thinking into a friendly display, just remove the HTML tag, and a blinking icon or so",1,0,sweihub,opencode,2026-03-23T09:05:47Z,"2026-04-15 01:20:48" | |
| 221,minimax/minimax-m2.7,4116458178,"Cloud LLM (OpenRouter): region-blocked models fail with 403 from HK server","## Problem | |
| Premium LLM diagnosis via OpenRouter fails with **403 Forbidden** when called from the PolyU GPU server (Hong Kong): | |
| ``` | |
| HTTP Request: POST https://openrouter.ai/api/v1/chat/completions ""HTTP/1.1 403 Forbidden"" | |
| error: {'message': 'This model is not available in your region.', 'code': 40",1,0,talonMMA,stf_ai_diagnosis_platform_v1,2026-03-22T16:27:18Z,"2026-04-15 01:20:48" | |
| 222,minimax/minimax-m2.7,4113090884,"[FEATURE]:Opencode Go new tier","### Feature hasn't been suggested before. | |
| - [x] I have verified this feature I'm about to request hasn't been suggested before. | |
| ### Describe the enhancement you want to request | |
| Hi, I like Opencode Go but it's too small. Can we have new Tiers? Like $15 or $20? Any chance to add new models from nvi",1,0,julianperezpesce,opencode,2026-03-21T17:17:24Z,"2026-04-15 01:20:48" | |
| 223,minimax/minimax-m2.7,4100711113,"[Feature Request]: Allow dynamic model passthrough for custom OpenAI-compatible providers without explicit declaration","### Feature hasn't been suggested before. | |
| - [x] I have verified this feature I'm about to request hasn't been suggested before. | |
| ### Describe the enhancement you want to request | |
| ## Problem Statement | |
| When using custom OpenAI-compatible providers (e.g., MiniMax, DeepSeek, or other third-party APIs ",0,0,mameikagou,opencode,2026-03-19T10:17:51Z,"2026-04-15 01:20:48" | |
| 224,minimax/minimax-m2.7,4100649638,"[Feature Request]: Allow dynamic model passthrough for custom OpenAI-compatible providers without explicit declaration","### Feature hasn't been suggested before. | |
| - [x] I have verified this feature I'm about to request hasn't been suggested before. | |
| ### Describe the enhancement you want to request | |
| ## Problem Statement | |
| When using custom OpenAI-compatible providers (e.g., MiniMax, DeepSeek, or other third-party APIs ",0,0,mameikagou,opencode,2026-03-19T10:07:02Z,"2026-04-15 01:20:48" | |
| 225,minimax/minimax-m2.7,4100638595,"[Feature Request]: Allow dynamic model passthrough for custom OpenAI-compatible providers without explicit declaration","### Feature hasn't been suggested before. | |
| - [x] I have verified this feature I'm about to request hasn't been suggested before. | |
| ### Describe the enhancement you want to request | |
| ## Problem Statement | |
| When using custom OpenAI-compatible providers (e.g., MiniMax, DeepSeek, or other third-party APIs ",0,5,mameikagou,opencode,2026-03-19T10:05:04Z,"2026-04-15 01:20:48" | |
| 226,minimax/minimax-m2.7,4093899435,"[FEATURE]: Minimax M2.7 access via Minimax coding plan","### Feature hasn't been suggested before. | |
| - [x] I have verified this feature I'm about to request hasn't been suggested before. | |
| ### Describe the enhancement you want to request | |
| Minimax M2.7 is released & available alongside it's new [Token Plan](https://platform.minimax.io/docs/token-plan/intro).",1,0,leonidlouis,opencode,2026-03-18T09:31:50Z,"2026-04-15 01:20:48" | |
| 227,minimax/minimax-m2.7,4076680152,"discussion: response to #451 — TUI strategy and architecture review alignment","## Context | |
| #451 is a comprehensive architecture review with 7 redesign proposals, 3 removal candidates, and 3 missing features. This issue responds to each, focusing on the TUI strategy question. | |
| Cross-references: #447 (extract methods from run_event_loop), #448 (inference loop decomposition), #44",2,0,lijunzh,koda,2026-03-14T20:37:09Z,"2026-04-15 01:20:48" | |
| 228,minimax/minimax-m2.7,4062168810,"[Discussion]: Model Provider Directory & Integration Policy","## 模型服务商目录 & 接入策略 / Model Provider Directory & Integration Policy | |
| ### 背景 / Background | |
| Cherry Studio 目前已内置 **62 个**模型服务商（详见下方列表），覆盖了主流 AI 模型原厂、云平台、推理平台和聚合转发服务。 | |
| 由于 provider 数量已经非常多，为了**避免影响用户体验**（服务商列表过长、配置复杂度增加），我们暂时**不会继续接入聚合类模型服务商**。对于只提供 OpenAI 兼容 API 的服务商，用户可以通过 **自定义 Provider (New API / OpenA",1,1,SiinXu,cherry-studio,2026-03-12T05:00:55Z,"2026-04-15 01:20:48" | |
| 229,minimax/minimax-m2.7,4004499599,"[FEATURE]: Implement the Agentic Context Engineering: Evolving Contexts for Self-Improving Language Models","### Feature hasn't been suggested before. | |
| - [x] I have verified this feature I'm about to request hasn't been suggested before. | |
| ### Describe the enhancement you want to request | |
| Proposal: Agentic Context Engineering (ACE) Implementation - Technical Review Needed | |
| ===================================",2,0,charmandercha,opencode,2026-02-28T11:29:49Z,"2026-04-15 01:20:48" | |
| 230,moonshotai/kimi-k2.5,4259720811,"[FEATURE]: kimi k2.6 integration","### Feature hasn't been suggested before. | |
| - [x] I have verified this feature I'm about to request hasn't been suggested before. | |
| ### Describe the enhancement you want to request | |
| Please add the new Kimi K2.6 and K2.6-code-preview model that is currently available for the code plans. | |
| It is possible ",2,11,Acters,opencode,2026-04-14T05:37:06Z,"2026-04-15 01:20:51" | |
| 231,moonshotai/kimi-k2.5,4190698113,"[FEATURE]:Image Visual Recognition Support","### Feature hasn't been suggested before. | |
| - [x] I have verified this feature I'm about to request hasn't been suggested before. | |
| ### Describe the enhancement you want to request | |
| he selected model is Kimi K2.5, which supports image recognition. | |
| However, when copying an image file in the OpenCode TU",1,0,cqJackzk,opencode,2026-04-02T00:42:14Z,"2026-04-15 01:20:51" | |
| 232,moonshotai/kimi-k2.5,4116458178,"Cloud LLM (OpenRouter): region-blocked models fail with 403 from HK server","## Problem | |
| Premium LLM diagnosis via OpenRouter fails with **403 Forbidden** when called from the PolyU GPU server (Hong Kong): | |
| ``` | |
| HTTP Request: POST https://openrouter.ai/api/v1/chat/completions ""HTTP/1.1 403 Forbidden"" | |
| error: {'message': 'This model is not available in your region.', 'code': 40",1,0,talonMMA,stf_ai_diagnosis_platform_v1,2026-03-22T16:27:18Z,"2026-04-15 01:20:51" | |
| 233,moonshotai/kimi-k2.5,4029820101,"[FEATURE]: add support for groq compound (the model)","### Feature hasn't been suggested before. | |
| - [x] I have verified this feature I'm about to request hasn't been suggested before. | |
| ### Describe the enhancement you want to request | |
| I'd like to be able to use https://console.groq.com/docs/compound with opencode | |
| The model seems quite fast and handy, t",6,0,RayBB,opencode,2026-03-05T18:11:30Z,"2026-04-15 01:20:51" | |
| 234,moonshotai/kimi-k2.5,3980241952,"[FEATURE]: Support for Alibaba Model Studio Coding Plan as provider","### Feature hasn't been suggested before. | |
| - [x] I have verified this feature I'm about to request hasn't been suggested before. | |
| ### Describe the enhancement you want to request | |
| Alibaba Model Studio has recently introduced a new Coding Plan that includes access to Qwen 3.5, Kimi K2.5, and GLM 4.7 ",4,15,vitobotta,opencode,2026-02-23T21:26:57Z,"2026-04-15 01:20:51" | |
| 235,moonshotai/kimi-k2.5,3967277497,"[FEATURE]: Evidence based refurbishment of openCode's System Prompt (Architecture)","### Feature hasn't been suggested before. | |
| - [x] I have verified this feature I'm about to request hasn't been suggested before. | |
| ### Describe the enhancement you want to request | |
| System Prompts are the hidden backbone of every coding agent. A well designed system prompting architecture is what turn",4,4,micuintus,opencode,2026-02-20T08:29:23Z,"2026-04-15 01:20:51" | |
| 236,moonshotai/kimi-k2.5,3967242535,"[FEATURE]: Evidence based refurbishment of openCode's System Prompt (Architecture)","### Feature hasn't been suggested before. | |
| - [x] I have verified this feature I'm about to request hasn't been suggested before. | |
| ### Describe the enhancement you want to request | |
| System Prompts are the hidden backbone of every coding agent. A well designed system prompting architecture is what turn",5,0,micuintus,opencode,2026-02-20T08:19:58Z,"2026-04-15 01:20:51" | |
| 237,moonshotai/kimi-k2.5,3941554734,"[FEATURE]: Unify provider system prompts","### Feature hasn't been suggested before. | |
| - [x] I have verified this feature I'm about to request hasn't been suggested before. | |
| ### Describe the enhancement you want to request | |
| System prompts for Anthropic, Codex, and Gemini models are very different, in ways that might not have good justificatio",6,4,smola,opencode,2026-02-14T14:36:43Z,"2026-04-15 01:20:51" | |
| 238,moonshotai/kimi-k2.5,3941240270,"[FEATURE]: Add extra body for NVIDIA NIM support","### Feature hasn't been suggested before. | |
| - [x] I have verified this feature I'm about to request hasn't been suggested before. | |
| ### Describe the enhancement you want to request | |
| From the official documentation for nvidia nim it is possible to set `""enable_thinking"":true,""clear_thinking"":false ` in",2,0,andriixyz,opencode,2026-02-14T11:55:00Z,"2026-04-15 01:20:51" | |
| 239,moonshotai/kimi-k2.5,3931633041,"[FEATURE]: Add GLM 5 and MiniMax M2.5 to Zen","### Feature hasn't been suggested before. | |
| - [x] I have verified this feature I'm about to request hasn't been suggested before. | |
| ### Describe the enhancement you want to request | |
| GLM 5 and MiniMax M2.5 just dropped (Feb 11). Both are already usable in OpenCode via their respective providers, but th",2,10,hrdkbhatnagar,opencode,2026-02-12T12:34:04Z,"2026-04-15 01:20:51" | |
| 240,moonshotai/kimi-k2.5,3910613966,"[FEATURE]: Add azure Kimi-K2.5","### Feature hasn't been suggested before. | |
| - [x] I have verified this feature I'm about to request hasn't been suggested before. | |
| ### Describe the enhancement you want to request | |
| Can you plese add model azure Kimi-K2.5 ??",2,1,Nuno-Flora,opencode,2026-02-07T15:58:32Z,"2026-04-15 01:20:51" | |
| 241,moonshotai/kimi-k2.5,3901568695,"[FEATURE]: favourites model search","### Feature hasn't been suggested before. | |
| - [x] I have verified this feature I'm about to request hasn't been suggested before. | |
| ### Describe the enhancement you want to request | |
| i put all the models i use in my favourites list so its a large list but it would be helpfull if i search there was a ta",2,0,Chris-Peeters,opencode,2026-02-05T13:24:14Z,"2026-04-15 01:20:51" | |
| 242,moonshotai/kimi-k2.5,3898138560,"[FEATURE]:: paste images directly into the chat UI like another chat cllients","### Feature hasn't been suggested before. | |
| - [x] I have verified this feature I'm about to request hasn't been suggested before. | |
| ### Describe the enhancement you want to request | |
| ### Main problem: Unable to paste images directly into the chat UI | |
| **Context** | |
| In the **OpenCode** chat graphical int",2,0,skyySea1,opencode,2026-02-04T19:44:28Z,"2026-04-15 01:20:51" | |
| 243,moonshotai/kimi-k2.5,3893522129,"[FEATURE]: Tool-calling optimization for Trinity model family","### Feature hasn't been suggested before. | |
| - [x] I have verified this feature I'm about to request hasn't been suggested before. | |
| ### Describe the enhancement you want to request | |
| Add a dedicated system prompt for the Trinity model family to enforce: | |
| - one tool per message | |
| - Use of the question too",1,0,mariamjabara,opencode,2026-02-03T22:24:58Z,"2026-04-15 01:20:51" | |
| 244,moonshotai/kimi-k2.5,3885580945,"[FEATURE]: Support StepFun (step-3.5-flash) via OpenRouter","### Feature hasn't been suggested before. | |
| - [x] I have verified this feature I'm about to request hasn't been suggested before. | |
| ### Describe the enhancement you want to request | |
| OpenCode currently supports multiple model providers through OpenRouter, but StepFun models specifically step-3.5-flash ",6,5,mporracindie,opencode,2026-02-02T12:07:03Z,"2026-04-15 01:20:51" | |
| 797,minimax/minimax-m2.7,4226230648,"Governance Layer for CORAL — Running Similar Architecture with Council + Adversarial Testing","Hi team — great work on CORAL. The architecture maps closely to a system I have been running in production for 8 months on consumer hardware, and I wanted to share what I have found works (and what is missing) in the hope it is useful. | |
| ## What we share | |
| Your shared persistent file system (attempts/",8,0,dereadi,CORAL,2026-04-08T17:03:34Z,"2026-04-15 13:21:26" | |
| 1027,meta-llama/llama-4-scout,4029820101,"[FEATURE]: add support for groq compound (the model)","### Feature hasn't been suggested before. | |
| - [x] I have verified this feature I'm about to request hasn't been suggested before. | |
| ### Describe the enhancement you want to request | |
| I'd like to be able to use https://console.groq.com/docs/compound with opencode | |
| The model seems quite fast and handy, t",6,0,RayBB,opencode,2026-03-05T18:11:30Z,"2026-04-16 02:20:26" | |
| 1028,meta-llama/llama-4-scout,3785107753,"[Feature]Can unsloth do inference for Llama 4 Scout on 2080 ti?","Can unsloth do inference for Llama 4 Scout on 2080 ti?",4,0,SidneyLann,unsloth,2026-01-06T13:14:07Z,"2026-04-16 02:20:26" | |
| 1750,anthropic/claude-opus-4.6,4276571047,"[FEATURE]: Add support for claude opus 4.7 xhigh adaptive reasoning effort","### Feature hasn't been suggested before. | |
| - [x] I have verified this feature I'm about to request hasn't been suggested before. | |
| ### Describe the enhancement you want to request | |
| See https://platform.claude.com/docs/en/build-with-claude/effort#recommended-effort-levels-for-claude-opus-4-7.",0,5,GrahamCampbell,opencode,2026-04-16T14:57:02Z,"2026-04-17 00:20:06" | |
| 1751,anthropic/claude-opus-4.6,4274775637,"[FEATURE]: Support for disabling adaptive thinking on Opus/Sonnet 4.6 (equivalent to CLAUDE_CODE_DISABLE_ADAPTIVE_THINKING)","### Feature hasn't been suggested before. | |
| - [x] I have verified this feature I'm about to request hasn't been suggested before. | |
| ### Describe the enhancement you want to request | |
| Anyone else noticing Opus 4.6 reasoning feels shallower these past weeks? Hallucinations, made-up commits, edits without",1,0,r00tedbrain-backup,opencode,2026-04-16T09:52:23Z,"2026-04-17 00:20:06" | |
| 1765,anthropic/claude-sonnet-4.6,4274775637,"[FEATURE]: Support for disabling adaptive thinking on Opus/Sonnet 4.6 (equivalent to CLAUDE_CODE_DISABLE_ADAPTIVE_THINKING)","### Feature hasn't been suggested before. | |
| - [x] I have verified this feature I'm about to request hasn't been suggested before. | |
| ### Describe the enhancement you want to request | |
| Anyone else noticing Opus 4.6 reasoning feels shallower these past weeks? Hallucinations, made-up commits, edits without",1,0,r00tedbrain-backup,opencode,2026-04-16T09:52:23Z,"2026-04-17 00:20:10" | |
| 1956,openai/gpt-5.4-pro,4280061621,"[FEATURE]: Show subagent model in session tree / subagent view","### Feature hasn't been suggested before. | |
| - [x] I have verified this feature I'm about to request hasn't been suggested before. | |
| ### Describe the enhancement you want to request | |
| When OpenCode spawns subagents, it would be helpful to show the model used by each subagent in the session UI. This allo",1,0,christian-taillon,opencode,2026-04-17T04:13:11Z,"2026-04-17 06:19:47" | |
| 1957,openai/gpt-5.4,4280061621,"[FEATURE]: Show subagent model in session tree / subagent view","### Feature hasn't been suggested before. | |
| - [x] I have verified this feature I'm about to request hasn't been suggested before. | |
| ### Describe the enhancement you want to request | |
| When OpenCode spawns subagents, it would be helpful to show the model used by each subagent in the session UI. This allo",1,0,christian-taillon,opencode,2026-04-17T04:13:11Z,"2026-04-17 06:19:47" | |
| 1986,openai/gpt-5.4-mini,4280061621,"[FEATURE]: Show subagent model in session tree / subagent view","### Feature hasn't been suggested before. | |
| - [x] I have verified this feature I'm about to request hasn't been suggested before. | |
| ### Describe the enhancement you want to request | |
| When OpenCode spawns subagents, it would be helpful to show the model used by each subagent in the session UI. This allo",1,0,christian-taillon,opencode,2026-04-17T04:13:11Z,"2026-04-17 06:19:50" | |
| 2217,anthropic/claude-opus-4.6,4281262391,"[FEATURE]: Anthropic ""advisor strategy""","### Feature hasn't been suggested before. | |
| - [x] I have verified this feature I'm about to request hasn't been suggested before. | |
| ### Describe the enhancement you want to request | |
| Claude Code has this new feature: | |
| > [Advisor strategy (public beta)](https://claude.com/blog/the-advisor-strategy): Pai",2,0,bestouff,opencode,2026-04-17T08:36:25Z,"2026-04-17 12:20:05" | |
| 2232,anthropic/claude-sonnet-4.6,4281262391,"[FEATURE]: Anthropic ""advisor strategy""","### Feature hasn't been suggested before. | |
| - [x] I have verified this feature I'm about to request hasn't been suggested before. | |
| ### Describe the enhancement you want to request | |
| Claude Code has this new feature: | |
| > [Advisor strategy (public beta)](https://claude.com/blog/the-advisor-strategy): Pai",2,0,bestouff,opencode,2026-04-17T08:36:25Z,"2026-04-17 12:20:09" | |
| 2247,anthropic/claude-haiku-4.5,4281262391,"[FEATURE]: Anthropic ""advisor strategy""","### Feature hasn't been suggested before. | |
| - [x] I have verified this feature I'm about to request hasn't been suggested before. | |
| ### Describe the enhancement you want to request | |
| Claude Code has this new feature: | |
| > [Advisor strategy (public beta)](https://claude.com/blog/the-advisor-strategy): Pai",2,0,bestouff,opencode,2026-04-17T08:36:25Z,"2026-04-17 12:20:12" | |
| 3224,anthropic/claude-sonnet-4.6,4285311241,"Design tension: DTO boundary — services return DTOs (current) vs domain objects (Spring-Java instinct)","## Not really an issue — food for thought | |
| This is a design tension I want to revisit, not a bug report. The harness currently enforces one pattern; my personal instinct from a Java/Spring background pulls the other way. Both are defensible. Opening this to collect the arguments in one place and dec",0,0,matt-grain,AgenticDevelopmentBestPractices,2026-04-17T21:00:42Z,"2026-04-17 22:20:06" | |
| 3897,minimax/minimax-m2.7,4286717561,"[FEATURE]: Nvidia API can now be used for free with GLM5.1","### Feature hasn't been suggested before. | |
| - [x] I have verified this feature I'm about to request hasn't been suggested before. | |
| ### Describe the enhancement you want to request | |
| Nvidia API can now use GLM5.1 for free, please add this model to the list.",2,0,MengNianxiaoyao,opencode,2026-04-18T04:39:21Z,"2026-04-18 14:21:20" | |
| 4409,qwen/qwen3-32b,4288839793,"[FEATURE]:LM Studio provider should auto-detect models via /v1/models API (similar to Continue extension)","### Feature hasn't been suggested before. | |
| - [x] I have verified this feature I'm about to request hasn't been suggested before. | |
| ### Describe the enhancement you want to request | |
| When connecting OpenCode to LM Studio, it currently fetches a hardcoded list of models from models.dev instead of queryi",2,0,lobanov-coder,opencode,2026-04-18T18:29:00Z,"2026-04-19 00:01:38" | |
| 4816,anthropic/claude-opus-4.6,4289958849,"[FEATURE]: Knowledge Harness of SOTA models as blueprint for cheap worker models","### Feature hasn't been suggested before. | |
| - [x] I have verified this feature I'm about to request hasn't been suggested before. | |
| ### Describe the enhancement you want to request | |
| **What feature would you like to see?** | |
| A knowledge_harness tool that lets cheap worker models (Kimi K2.5, GLM-5, etc.) ",0,0,Iftekhar-Code-Against-Matrix,opencode,2026-04-19T05:01:06Z,"2026-04-19 15:53:31" | |
| 4907,anthropic/claude-opus-4.6,4291832531,"feat(governance): Add circuit breaker to AI-REVIEW-MODEL-POLICY.md","## Problem | |
| Current `.agents/governance/AI-REVIEW-MODEL-POLICY.md` forbids PASS without evidence and has model routing matrix, but lacks: | |
| 1. **Fallback circuit breaker** — what's the safe default when review model fails/unavailable? | |
| 2. **Confidence labeling** — how to communicate certainty levels (f",5,1,rjmurillo,ai-agents,2026-04-19T19:44:52Z,"2026-04-19 21:53:37" | |
| 4922,anthropic/claude-sonnet-4.6,4291832531,"feat(governance): Add circuit breaker to AI-REVIEW-MODEL-POLICY.md","## Problem | |
| Current `.agents/governance/AI-REVIEW-MODEL-POLICY.md` forbids PASS without evidence and has model routing matrix, but lacks: | |
| 1. **Fallback circuit breaker** — what's the safe default when review model fails/unavailable? | |
| 2. **Confidence labeling** — how to communicate certainty levels (f",5,1,rjmurillo,ai-agents,2026-04-19T19:44:52Z,"2026-04-19 21:53:41" | |
| 5484,openai/gpt-5.4-pro,4305245589,"[FEATURE]: Support duplicate provider config sections","### Feature hasn't been suggested before. | |
| - [x] I have verified this feature I'm about to request hasn't been suggested before. | |
| ### Describe the enhancement you want to request | |
| Support duplicate provider config sections so users can define multiple instances of the same provider, especially Azure",0,0,IustinT,opencode,2026-04-21T20:57:18Z,"2026-04-22 22:41:28" | |
| 5485,openai/gpt-5.4,4305245589,"[FEATURE]: Support duplicate provider config sections","### Feature hasn't been suggested before. | |
| - [x] I have verified this feature I'm about to request hasn't been suggested before. | |
| ### Describe the enhancement you want to request | |
| Support duplicate provider config sections so users can define multiple instances of the same provider, especially Azure",0,0,IustinT,opencode,2026-04-21T20:57:18Z,"2026-04-22 22:41:28" | |
| 5522,anthropic/claude-opus-4.6,4308966573,"discussion: master + sub-agent workflow (parallel/sequential, fg/bg) — what to learn from #216, AI-Build-AI, and Claude Code","# Discussion: Master + Sub-Agent Workflow (parallel/sequential, fg/bg) | |
| > **Discussion only — no code changes proposed yet.** Goal is to converge on a design before opening implementation issues. Closes the loop on #216 (which was reverted in #355) by exploring a fundamentally different shape: not a",4,0,lijunzh,koda,2026-04-22T12:23:53Z,"2026-04-22 22:41:44" | |
| 5523,anthropic/claude-opus-4.6,4302863967,"[FEATURE]: Per-turn local model middleware for context compression, output routing, and delegation (Hybrid Frontier + Local Agent Loop)","### Feature hasn't been suggested before. | |
| - [x] I have verified this feature I'm about to request hasn't been suggested before. | |
| ### Describe the enhancement you want to request | |
| ## Summary | |
| This proposes a **middleware agent role** — a local model layer that intercepts the agent loop at every tur",3,0,eramax,opencode,2026-04-21T13:20:46Z,"2026-04-22 22:41:44" | |
| 5537,anthropic/claude-sonnet-4.6,4308966573,"discussion: master + sub-agent workflow (parallel/sequential, fg/bg) — what to learn from #216, AI-Build-AI, and Claude Code","# Discussion: Master + Sub-Agent Workflow (parallel/sequential, fg/bg) | |
| > **Discussion only — no code changes proposed yet.** Goal is to converge on a design before opening implementation issues. Closes the loop on #216 (which was reverted in #355) by exploring a fundamentally different shape: not a",4,0,lijunzh,koda,2026-04-22T12:23:53Z,"2026-04-22 22:41:48" | |
| 5595,meta-llama/llama-4-maverick,4302863967,"[FEATURE]: Per-turn local model middleware for context compression, output routing, and delegation (Hybrid Frontier + Local Agent Loop)","### Feature hasn't been suggested before. | |
| - [x] I have verified this feature I'm about to request hasn't been suggested before. | |
| ### Describe the enhancement you want to request | |
| ## Summary | |
| This proposes a **middleware agent role** — a local model layer that intercepts the agent loop at every tur",3,0,eramax,opencode,2026-04-21T13:20:46Z,"2026-04-22 22:42:04" | |
| 5596,meta-llama/llama-4-maverick,4298404808,"🎉 Open Source: Zero-Knowledge Correspondent OS — 10 Tools, MIT License, Day 1 Ready","@rising-leviathan @arc0btc @whoabuddy @biwasxyz @Robotbot69 @tearful-saw @ThankNIXlater | |
| # 🎉 Open Source Release: Turn-Key Correspondent Operating System | |
| **Repository**: https://github.com/netmask255/aibtc-news-correspondent-skill | |
| **License**: MIT (fully open) | |
| **Status**: Production-ready, 20 ",8,0,netmask255,agent-news,2026-04-20T20:14:12Z,"2026-04-22 22:42:04" | |
| 5597,meta-llama/llama-4-maverick,4291549083,"[Question] How do I access studio on the docker container and limited sudo access on container","Note: Please do not remove the questions. Answer beside them. | |
| I use the latest docker image, its studio is basically disabled, and the sudo password `USER_PASSWORD` for `localhost:8888` is not functional any more; I used the prior version, and these two issues were not existing, this is the url of ",11,0,Bader-Idris,unsloth,2026-04-19T17:37:36Z,"2026-04-22 22:42:04" | |
| 6554,openai/gpt-5.4-pro,4317936143,"[FEATURE]: GPT 5.5 support","### Feature hasn't been suggested before. | |
| - [x] I have verified this feature I'm about to request hasn't been suggested before. | |
| ### Describe the enhancement you want to request | |
| https://openai.com/index/introducing-gpt-5-5/",4,18,bu4ak,opencode,2026-04-23T18:15:18Z,"2026-04-24 02:41:21" | |
| 6555,openai/gpt-5.4,4317936143,"[FEATURE]: GPT 5.5 support","### Feature hasn't been suggested before. | |
| - [x] I have verified this feature I'm about to request hasn't been suggested before. | |
| ### Describe the enhancement you want to request | |
| https://openai.com/index/introducing-gpt-5-5/",4,18,bu4ak,opencode,2026-04-23T18:15:18Z,"2026-04-24 02:41:21" | |
| 6592,anthropic/claude-opus-4.6,4317948165,"[FEATURE]: Claude support using ACP protocol","### Feature hasn't been suggested before. | |
| - [x] I have verified this feature I'm about to request hasn't been suggested before. | |
| ### Describe the enhancement you want to request | |
| I just installed Agent-Shell for Emacs that allows using Claude Code subscription via Agent Client Protocol and @agentcl",3,0,jcubic,opencode,2026-04-23T18:17:49Z,"2026-04-24 02:41:35" | |
| 7059,openai/gpt-5.4-pro,4321985780,"[FEATURE]: support codex as a first-class provider (codex subscription rather than OpenAI api provider)","### Feature hasn't been suggested before. | |
| - [x] I have verified this feature I'm about to request hasn't been suggested before. | |
| ### Describe the enhancement you want to request | |
| I would like OpenCode to support `codex` as a first-class provider for Codex-compatible Responses API endpoints. OpenCo",2,0,PadishahIII,opencode,2026-04-24T09:30:58Z,"2026-04-24 13:41:21" | |
| 7060,openai/gpt-5.4,4321985780,"[FEATURE]: support codex as a first-class provider (codex subscription rather than OpenAI api provider)","### Feature hasn't been suggested before. | |
| - [x] I have verified this feature I'm about to request hasn't been suggested before. | |
| ### Describe the enhancement you want to request | |
| I would like OpenCode to support `codex` as a first-class provider for Codex-compatible Responses API endpoints. OpenCo",2,0,PadishahIII,opencode,2026-04-24T09:30:58Z,"2026-04-24 13:41:21" | |
| 11542,anthropic/claude-opus-4.6,4331731761,"Future: version upgrade protocols + rolling-upgrade tooling (incl. soak grove playbook)","## Context | |
| Willow currently has no formal protocol for client/event-format version upgrades. Breaking changes today require coordinated wipes — for the prod deployment that's manual operator work, and for the long-running soak swarm (`docs/specs/2026-04-26-long-running-soak-swarm-design.md`) it mea",0,0,intendednull,willow,2026-04-26T19:08:38Z,"2026-04-26 19:41:35" | |
| 12617,qwen/qwen3-32b,4302863967,"[FEATURE]: Per-turn local model middleware for context compression, output routing, and delegation (Hybrid Frontier + Local Agent Loop)","### Feature hasn't been suggested before. | |
| - [x] I have verified this feature I'm about to request hasn't been suggested before. | |
| ### Describe the enhancement you want to request | |
| ## Summary | |
| This proposes a **middleware agent role** — a local model layer that intercepts the agent loop at every tur",3,0,eramax,opencode,2026-04-21T13:20:46Z,"2026-04-27 03:42:40" | |
| 12638,x-ai/grok-4,4302863967,"[FEATURE]: Per-turn local model middleware for context compression, output routing, and delegation (Hybrid Frontier + Local Agent Loop)","### Feature hasn't been suggested before. | |
| - [x] I have verified this feature I'm about to request hasn't been suggested before. | |
| ### Describe the enhancement you want to request | |
| ## Summary | |
| This proposes a **middleware agent role** — a local model layer that intercepts the agent loop at every tur",3,0,eramax,opencode,2026-04-21T13:20:46Z,"2026-04-27 03:42:47" | |
| 12639,x-ai/grok-4.20,4302863967,"[FEATURE]: Per-turn local model middleware for context compression, output routing, and delegation (Hybrid Frontier + Local Agent Loop)","### Feature hasn't been suggested before. | |
| - [x] I have verified this feature I'm about to request hasn't been suggested before. | |
| ### Describe the enhancement you want to request | |
| ## Summary | |
| This proposes a **middleware agent role** — a local model layer that intercepts the agent loop at every tur",3,0,eramax,opencode,2026-04-21T13:20:46Z,"2026-04-27 03:42:47" | |
| 12640,x-ai/grok-4,4299126306,"How should we handle non-English articles?","The Psyop Hunter system prompt is written in English and the 24 markers assume English-language framing conventions. Scanning a French or Arabic article produces *some* output, but we don't have a principled approach. | |
| ### Discussion | |
| - Should the system prompt itself be translated, or does Grok-4.1 ",0,0,goshtasb,Anie,2026-04-20T22:49:37Z,"2026-04-27 03:42:47" | |
| 12641,x-ai/grok-4.20,4299126306,"How should we handle non-English articles?","The Psyop Hunter system prompt is written in English and the 24 markers assume English-language framing conventions. Scanning a French or Arabic article produces *some* output, but we don't have a principled approach. | |
| ### Discussion | |
| - Should the system prompt itself be translated, or does Grok-4.1 ",0,0,goshtasb,Anie,2026-04-20T22:49:37Z,"2026-04-27 03:42:47" | |
| 12668,minimax/minimax-m2.7,4320151236,"[Discussion]: The same model interface works on the homepage default assistant but shows an error on the Agent page","> [\!NOTE] | |
| > This issue was translated by Claude. | |
| ### Issue Checklist | |
| - [x] I understand that issues are meant for feedback and problem-solving, not for venting, and I will provide as much detail as possible to help resolve the issue. | |
| - [x] I have checked the pinned issues and searched through the",0,1,BarrySong,cherry-studio,2026-04-24T02:43:41Z,"2026-04-27 03:42:51" | |
| 12669,minimax/minimax-m2.7,4301431812,"[FEATURE]: add Responses API support for Go service","### Feature hasn't been suggested before. | |
| - [x] I have verified this feature I'm about to request hasn't been suggested before. | |
| ### Describe the enhancement you want to request | |
| ### Feature Request | |
| **Is your feature request related to a problem? Please describe.** | |
| The OpenCode Go service current",1,1,xiaoyingv,opencode,2026-04-21T08:52:30Z,"2026-04-27 03:42:51" | |
| 13203,moonshotai/kimi-k2.5,4301431812,"[FEATURE]: add Responses API support for Go service","### Feature hasn't been suggested before. | |
| - [x] I have verified this feature I'm about to request hasn't been suggested before. | |
| ### Describe the enhancement you want to request | |
| ### Feature Request | |
| **Is your feature request related to a problem? Please describe.** | |
| The OpenCode Go service current",1,1,xiaoyingv,opencode,2026-04-21T08:52:30Z,"2026-04-27 07:43:16" | |
| 13204,moonshotai/kimi-k2.5,4300487072,"[FEATURE]: Add support for Kimi K2.6","### Feature hasn't been suggested before. | |
| - [x] I have verified this feature I'm about to request hasn't been suggested before. | |
| ### Describe the enhancement you want to request | |
| Add support for Kimi 2.6",3,0,vivek12345,opencode,2026-04-21T05:28:26Z,"2026-04-27 07:43:16" | |
| 13205,moonshotai/kimi-k2.5,4289958849,"[FEATURE]: Knowledge Harness of SOTA models as blueprint for cheap worker models","### Feature hasn't been suggested before. | |
| - [x] I have verified this feature I'm about to request hasn't been suggested before. | |
| ### Describe the enhancement you want to request | |
| **What feature would you like to see?** | |
| A knowledge_harness tool that lets cheap worker models (Kimi K2.5, GLM-5, etc.) ",0,0,Iftekhar-Code-Against-Matrix,opencode,2026-04-19T05:01:06Z,"2026-04-27 07:43:16" | |
| 15089,anthropic/claude-opus-4.6,4343602582,"[FEATURE]: support named agent colors for CC compatibility (red, blue...)","### Feature hasn't been suggested before. | |
| - [x] I have verified this feature I'm about to request hasn't been suggested before. | |
| ### Describe the enhancement you want to request | |
| The color option for Claude Code subagents supports only named colors like: red, blue, green, yellow, purple, orange, pi",1,0,sentisso,opencode,2026-04-28T13:40:15Z,"2026-04-28 22:41:38" | |
| 15697,deepseek/deepseek-v3.2,3505142435,"[Discussion]: How to enable thinking mode for DeepSeek 3.2 deployed via VLLM?","### Issue Checklist | |
| - [x] I understand that issues are meant for feedback and problem-solving, not for venting, and I will provide as much detail as possible to help resolve the issue. | |
| - [x] I have checked the pinned issues and searched through the existing [open issues](https://github.com/CherryHQ",3,3,yanmindi,cherry-studio,2025-10-11T04:16:21Z,"2026-04-29 02:42:35" | |
| 15971,moonshotai/kimi-k2.5,4082165381,"[FEATURE]: Unify shell initialization behavior between user shell and agent bash tool","### Feature hasn't been suggested before. | |
| - [x] I have verified this feature I'm about to request hasn't been suggested before. | |
| ### Describe the enhancement you want to request | |
| ## Summary | |
| There is an inconsistency between the user-invoked shell (`!`) and the agent `bash` tool. | |
| At the moment: | |
| - th",2,2,bakabaka9405,opencode,2026-03-16T12:23:51Z,"2026-04-29 03:42:35" | |
| 16789,anthropic/claude-sonnet-4.6,4350608881,"[FEATURE]:OpenCode creates configuration files in every directory where it's launched,can you create a switch to on/off it???","### Feature hasn't been suggested before. | |
| - [x] I have verified this feature I'm about to request hasn't been suggested before. | |
| ### Describe the enhancement you want to request | |
| ls -a | |
| ./ .bash_profile .claude/ .gitmodules .mcp.json opencode.json .ripgreprc .zprofile | |
| ../ .bash",8,0,smithyyang,opencode,2026-04-29T12:42:18Z,"2026-04-29 15:41:39" | |
| 18394,openai/gpt-5.4-pro,4355245253,"[FEATURE]: FAST MODE","### Feature hasn't been suggested before. | |
| - [x] I have verified this feature I'm about to request hasn't been suggested before. | |
| ### Describe the enhancement you want to request | |
| Please add a Fast Mode ^_^",2,0,Adikurnianto24,opencode,2026-04-30T04:08:13Z,"2026-04-30 05:41:21" | |
| 18395,openai/gpt-5.4,4355245253,"[FEATURE]: FAST MODE","### Feature hasn't been suggested before. | |
| - [x] I have verified this feature I'm about to request hasn't been suggested before. | |
| ### Describe the enhancement you want to request | |
| Please add a Fast Mode ^_^",2,0,Adikurnianto24,opencode,2026-04-30T04:08:13Z,"2026-04-30 05:41:21" | |
| 18432,anthropic/claude-opus-4.6,4355245253,"[FEATURE]: FAST MODE","### Feature hasn't been suggested before. | |
| - [x] I have verified this feature I'm about to request hasn't been suggested before. | |
| ### Describe the enhancement you want to request | |
| Please add a Fast Mode ^_^",2,0,Adikurnianto24,opencode,2026-04-30T04:08:13Z,"2026-04-30 05:41:35" | |
| 19292,meta-llama/llama-4-maverick,4330505935,"Contributing guide not working/missing dependencies","### What would you like to discuss with the community? | |
| Hi, | |
| I submitted the issue #5630 (I think that this is down to pushing a release to pypi). | |
| I continued with the getting started, but I am hitting next and next small issues. I can fix them locally on my dev machine, but I wanted to contribute ",3,0,zy1o,ogx,2026-04-26T09:18:27Z,"2026-04-30 13:42:01" | |
| 19344,qwen/qwen3-32b,4348363018,"Qwen3-TTS Data 관련 정리 및 문제점 해결 방안","# Qwen3-TTS Data 정리 | |
| Qwen3-TTS Fine-tuning을 위한 Data 정리 | |
| ## Input | |
| Emotion Vector | |
| Sample Audio | |
| STT를 통해 추출한 Data | |
| ## Output | |
| 원래 실제 목소리 Data | |
| ## 예상되는 문제점 | |
| Sample Audio의 inconsistency의 문제점 | |
| ### 해결 방안 | |
| 일단 Input과 다른 Audio에 대해서 동일한 화자이고, Nature로 분류된 Data에 한해서 Random Sampling 예정 ",0,0,TaeWonM,capstone_design_2026,2026-04-29T06:19:27Z,"2026-04-30 13:42:56" | |
| 19345,qwen/qwen3-32b,4321688942,"[Discussion]: Which network does web search use?","> [\!NOTE] | |
| > This issue was translated by Claude. | |
| ### Issue Checklist | |
| - [x] I understand that issues are meant for feedback and problem-solving, not for venting, and I will provide as much detail as possible to help resolve the issue. | |
| - [x] I have checked the pinned issues and searched through the",4,1,DarkFlameMasterDev,cherry-studio,2026-04-24T08:35:54Z,"2026-04-30 13:42:56" | |
| 19359,qwen/qwq-32b,4321688942,"[Discussion]: Which network does web search use?","> [\!NOTE] | |
| > This issue was translated by Claude. | |
| ### Issue Checklist | |
| - [x] I understand that issues are meant for feedback and problem-solving, not for venting, and I will provide as much detail as possible to help resolve the issue. | |
| - [x] I have checked the pinned issues and searched through the",4,1,DarkFlameMasterDev,cherry-studio,2026-04-24T08:35:54Z,"2026-04-30 13:43:04" | |
| 19411,moonshotai/kimi-k2.5,4352209342,"[FEATURE]: Expose Kimi K2.6 from Azure Foundry (aka Cognitive Services)","### Feature hasn't been suggested before. | |
| - [x] I have verified this feature I'm about to request hasn't been suggested before. | |
| ### Describe the enhancement you want to request | |
| Right now, I can only choose up to Kimi 2.5. | |
| <img width=""422"" height=""157"" alt=""Image"" src=""https://github.com/user-att",1,0,ucirello,opencode,2026-04-29T16:47:14Z,"2026-04-30 13:43:18" | |
| 19412,moonshotai/kimi-k2.5,4349365698,"[FEATURE]: Auto-trim tool names in experimental_repairToolCall","### Feature hasn't been suggested before. | |
| - [x] I have verified this feature I'm about to request hasn't been suggested before. | |
| ### Describe the enhancement you want to request | |
| ## Problem: | |
| Some models (e.g., kimi-k2.5) occasionally generate tool calls with leading/trailing whitespace in the tool ",1,0,KrabsWong,opencode,2026-04-29T09:31:21Z,"2026-04-30 13:43:18" | |
| 19605,anthropic/claude-sonnet-4.6,4359389468,"[FEATURE]: ~90% of compaction cost is avoidable cache miss","### Feature hasn't been suggested before. | |
| - [x] I have verified this feature I'm about to request hasn't been suggested before. | |
| ### Describe the enhancement you want to request | |
| ### Why compaction matters | |
| opencode runs long multi-turn agent sessions. As conversation history grows, we periodically",3,0,lloydzhou,opencode,2026-04-30T15:42:20Z,"2026-04-30 16:41:41" | |
| 19915,meta-llama/llama-4-maverick,4331814473,"Explore the PII-Masking-300k benchmark","@e-yen07, can you explore and succinctly write up here the main points of the PII-Masking-300k benchmark? | |
| It is mentioned in OpenAI's blog post [""Introducing OpenAI Privacy Filter.""](https://openai.com/index/introducing-openai-privacy-filter/) | |
| For example: | |
| - How does the PII-Masking-300k benchma",4,0,SebastianZimmeck,compliance-audit-benchmark,2026-04-26T19:46:39Z,"2026-04-30 18:41:53" | |