Repository navigation
[FEATURE]: GUI for per-model profile/context parameters (system prompt, temperature, context window) #46153
Description
Summary
- I have searched for existing issues before filing this
Allow users to configure per-model context parameters (system prompt / "bio", temperature, max tokens, context window size, reasoning effort, and more) directly from the new GUI — without touching opencode.jsonc.
Problem
OpenCode supports many LLMs with wildly different capabilities, but configuring them requires manual JSON editing. Users who aren't comfortable in config files cannot:
- Set a custom system prompt / "bio" / profile for a specific model
- Override the context window size for models that report incorrect metadata
- Adjust temperature or reasoning effort for model-specific behavior
- Set model-specific parameters that control how the model generates output
This Request Subsumes Existing Issues
This intentionally broadens and subsumes the following open feature requests into a unified concept — a Model Profile:
| Existing Issue | Scope | This Issue Adds |
|---|---|---|
| #40818 | Context window per model | + system prompt, temperature, max tokens |
| #41287 | Show model capabilities in picker | + editable fields, not just display |
| #42929 | Show context window limit in TUI | + writable profile config in GUI |
| #43601 | 设置大模型上下文窗口大小 | + full profile, not just window size |
| #38057 | Incorrect context params for StepFun | + the ability to override and fix it |
Proposed Solution
A Model Profile Configuration UI accessible from:
- The model picker (
...menu per model) - Settings → Model Profiles
- A
/profileslash command
Each profile:
{
"model": "gpt-4o",
"systemPrompt": "[custom bio/profile text]",
"temperature": 0.7,
"maxTokens": 4096,
"contextWindow": 128000,
"reasoningEffort": "medium"
}Profiles must be:
- Named and persistent — saved to config
- Assignable per-model — one profile per model
- Editable from GUI — no JSON required
- Exportable/importable — share profiles across machines
Plugin Angle
A modelProfile(params) plugin hook would let plugins:
- Inject or override profile parameters dynamically
- Enforce enterprise compliance system prompts
- Cap
maxTokensper model for cost control - Sanitize system prompts via security plugins
OS
Cross-platform (Desktop + TUI).
🦙 Creative Monkey at FrenzyPenguin Media — open-source security hardening & privacy tools
Reacted by METAPOD
Metadata
Metadata
Assignees
Labels
No labels
Activity
neohiro commentedon Aug 29, 2026
Hey @thdxr @adamdotdevin @fwang @Brendonovich — this subsumes #40818, #41287, #42929, #43601, #38057 into a unified Model Profile concept: system prompt, temperature, max tokens, context window, reasoning effort — all editable from GUI per model. Would also enable a modelProfile() plugin hook. 🦙 Creative Monkey at FrenzyPenguin Media
github-actions commentedon Aug 29, 2026
This issue does not fully meet our contributing guidelines.
What needs to be fixed:
[ENHANCEMENT]:prefix but does not follow any of the three required issue templates (Bug Report, Feature Request, or Question). Feature requests must use the Feature Request template, include a verification checkbox, and have a title starting with[FEATURE]:.Please edit this issue to address the above within 2 hours, or it will be automatically closed.
This issue might be a duplicate of existing issues. Please check:
If you believe this was flagged incorrectly, please let a maintainer know.
github-actions commentedon Aug 29, 2026
Thank you for editing the issue, but it still does not meet our contributing guidelines.
What still needs to be fixed:
[ENHANCEMENT]:but Feature Requests must start with[FEATURE]:. Please rename the title accordingly.Please edit the issue to address both points. Keep it simple and focused.
[-][ENHANCEMENT]: GUI for per-model profile/context parameters (system prompt, temperature, context window)[/-][+][FEATURE]: GUI for per-model profile/context parameters (system prompt, temperature, context window)[/+]github-actions commentedon Aug 29, 2026
Thank you for updating the title to
[FEATURE]:— that part is now correct.However, the issue still does not meet our contributing guidelines:
What still needs to be fixed:
Please replace the body with a concise, human-written description of:
Keep it focused and in your own words. No tables, schemas, or plugin specs needed.
github-actions commentedon Aug 29, 2026
Thank you for updating the issue. The title prefix is now correct and the verification checkbox is present.
However, the issue still does not meet our contributing guidelines:
What still needs to be fixed:
Please rewrite the body in plain, human-written text. A good feature request simply needs:
No tables, JSON schemas, plugin specifications, or promotional footers are needed.
github-actions commentedon Aug 29, 2026
Thanks for updating the issue! It now meets the contributing guidelines and looks good. We appreciate the detailed write-up.
opencodejr commentedon Oct 28, 2026
Thanks for the detailed write-up, @neohiro! The request is a GUI for creating and editing per-model profiles (system prompt, temperature, max tokens, context window override, reasoning effort). Profiles would be saved to config and could be exported and imported, so nobody has to edit
opencode.jsoncby hand. Maintainers will consider it.On the plugin idea: v2's plugin API already has session request hooks. They receive the request's
modeland can change itssystemparts and generationoptions(temperature, max tokens, provider options), so plugins can already inject or cap per-model parameters. The context window is the exception: it comes from the model catalog/config (limit.context/limit.output), not from each request.Related issues that each cover part of this:
limit(context/output)