fix(opencode): show Reasoning selector for OpenCode models

[maria-rcks]

Note

Written by muse-spark-1.2-contributor on behalf of Maria

OpenCode showed only the Agent chip (e.g. Build) while Codex/Grok/Cursor all expose a Reasoning control. Variant was labeled "Variant" and hidden when a model advertised no variants, leaving the composer with no effort selector.

Label the variant descriptor as "Reasoning" and synthesize standard levels (low/medium/high/xhigh, default medium) when a model reports no variants. Provide the same fallback for custom models and update the saved-option label for unavailable models. The composer now shows the reasoning effort next to the agent (e.g. "Medium · Build").

Before	After
	Reasoning menu (Reasoning / Agent groups)

Verified by inspecting provider capabilities and TraitsPicker logic; reasoning selector now renders for OpenCode models with and without advertised variants.

Model: muse-spark-1.2-contributor via opencode

---

Note

Medium Risk
Synthesized reasoning levels may be sent for models that never advertised variants; behavior depends on OpenCode accepting those values at request time.

Overview
OpenCode’s composer now aligns with Codex/Grok by exposing a Reasoning control (renamed from “Variant”) alongside Agent.

Server (OpenCodeProvider): When a model reports no variants, capabilities now synthesizelow / medium / high / xhigh (default still via inferDefaultVariant). The dynamic descriptor label is Reasoning. Custom / fallbackDEFAULT_OPENCODE_MODEL_CAPABILITIES includes the same Reasoning and Agent selectors instead of an empty options list.

Web (TraitsPicker): Saved-option display maps variant → Reasoning when model metadata is unavailable.

Users should see effort in the traits chip (e.g. Medium · Build) and in the traits menu for models with or without advertised variants.

^{Reviewed by Cursor Bugbot for commit 6fe6ced. Bugbot is set up for automated code reviews on this repo. Configure here.}

[macroscopeapp]

Approvability

Verdict: Not approved

Macroscope's review found this PR not approvable — The PR changes OpenCode’s production capability metadata by introducing default Medium/Build selections and synthesizing reasoning levels for models without advertised variants. Those defaults can affect the composer state and the variant sent on subsequent OpenCode requests, so the change warrants human review.

Not approved because:

• Monthly spending limit reached (workspace setting). Approvability relies on correctness review in order to determine eligibility

^{Review your spending limits in Billing settings. You can add or adjust custom eligibility rules. Learn more.}

[maria-rcks]

Note

Written by muse-spark-1.2-contributor on behalf of Maria

OpenCode showed only the Agent chip (e.g. Build) while Codex/Grok/Cursor all expose a Reasoning control. Variant was labeled "Variant" and hidden when a model advertised no variants, leaving the composer with no effort selector.

Label the variant descriptor as "Reasoning" and synthesize standard levels (low/medium/high/xhigh, default medium) when a model reports no variants. Provide the same fallback for custom models and update the saved-option label for unavailable models. The composer now shows the reasoning effort next to the agent (e.g. "Medium · Build").

Before	After
	Reasoning menu (Reasoning / Agent groups)

Verified by inspecting provider capabilities and TraitsPicker logic; reasoning selector now renders for OpenCode models with and without advertised variants.

Model: muse-spark-1.2-contributor via opencode

---

Note

Medium Risk
Synthesized reasoning levels may be sent for models that never advertised variants; behavior depends on OpenCode accepting those values at request time.

Overview
OpenCode’s composer now aligns with Codex/Grok by exposing a Reasoning control (renamed from “Variant”) alongside Agent.

Server (OpenCodeProvider): When a model reports no variants, capabilities now synthesizelow / medium / high / xhigh (default still via inferDefaultVariant). The dynamic descriptor label is Reasoning. Custom / fallbackDEFAULT_OPENCODE_MODEL_CAPABILITIES includes the same Reasoning and Agent selectors instead of an empty options list.

Web (TraitsPicker): Saved-option display maps variant → Reasoning when model metadata is unavailable.

Users should see effort in the traits chip (e.g. Medium · Build) and in the traits menu for models with or without advertised variants.

^{Reviewed by Cursor Bugbot for commit 6fe6ced. Bugbot is set up for automated code reviews on this repo. Configure here.}

[macroscopeapp]

Approvability

Verdict: Not approved

Macroscope's review found this PR not approvable — The PR changes OpenCode’s production capability metadata by introducing default Medium/Build selections and synthesizing reasoning levels for models without advertised variants. Those defaults can affect the composer state and the variant sent on subsequent OpenCode requests, so the change warrants human review.

Not approved because:

• Monthly spending limit reached (workspace setting). Approvability relies on correctness review in order to determine eligibility

^{Review your spending limits in Billing settings. You can add or adjust custom eligibility rules. Learn more.}

[maria-rcks]

Note

Written by muse-spark-1.2-contributor on behalf of Maria

OpenCode showed only the Agent chip (e.g. Build) while Codex/Grok/Cursor all expose a Reasoning control. Variant was labeled "Variant" and hidden when a model advertised no variants, leaving the composer with no effort selector.

Label the variant descriptor as "Reasoning" and synthesize standard levels (low/medium/high/xhigh, default medium) when a model reports no variants. Provide the same fallback for custom models and update the saved-option label for unavailable models. The composer now shows the reasoning effort next to the agent (e.g. "Medium · Build").

Before	After
	Reasoning menu (Reasoning / Agent groups)

Verified by inspecting provider capabilities and TraitsPicker logic; reasoning selector now renders for OpenCode models with and without advertised variants.

Model: muse-spark-1.2-contributor via opencode

---

Note

Medium Risk
Synthesized reasoning levels may be sent for models that never advertised variants; behavior depends on OpenCode accepting those values at request time.

Overview
OpenCode’s composer now aligns with Codex/Grok by exposing a Reasoning control (renamed from “Variant”) alongside Agent.

Server (OpenCodeProvider): When a model reports no variants, capabilities now synthesizelow / medium / high / xhigh (default still via inferDefaultVariant). The dynamic descriptor label is Reasoning. Custom / fallbackDEFAULT_OPENCODE_MODEL_CAPABILITIES includes the same Reasoning and Agent selectors instead of an empty options list.

Web (TraitsPicker): Saved-option display maps variant → Reasoning when model metadata is unavailable.

Users should see effort in the traits chip (e.g. Medium · Build) and in the traits menu for models with or without advertised variants.

^{Reviewed by Cursor Bugbot for commit 6fe6ced. Bugbot is set up for automated code reviews on this repo. Configure here.}

[macroscopeapp]

Approvability

Verdict: Not approved

Macroscope's review found this PR not approvable — The PR changes OpenCode’s production capability metadata by introducing default Medium/Build selections and synthesizing reasoning levels for models without advertised variants. Those defaults can affect the composer state and the variant sent on subsequent OpenCode requests, so the change warrants human review.

Not approved because:

• Monthly spending limit reached (workspace setting). Approvability relies on correctness review in order to determine eligibility

^{Review your spending limits in Billing settings. You can add or adjust custom eligibility rules. Learn more.}

[maria-rcks]

Note

Written by muse-spark-1.2-contributor on behalf of Maria

OpenCode showed only the Agent chip (e.g. Build) while Codex/Grok/Cursor all expose a Reasoning control. Variant was labeled "Variant" and hidden when a model advertised no variants, leaving the composer with no effort selector.

Label the variant descriptor as "Reasoning" and synthesize standard levels (low/medium/high/xhigh, default medium) when a model reports no variants. Provide the same fallback for custom models and update the saved-option label for unavailable models. The composer now shows the reasoning effort next to the agent (e.g. "Medium · Build").

Before	After
	Reasoning menu (Reasoning / Agent groups)

Verified by inspecting provider capabilities and TraitsPicker logic; reasoning selector now renders for OpenCode models with and without advertised variants.

Model: muse-spark-1.2-contributor via opencode

---

Note

Medium Risk
Synthesized reasoning levels may be sent for models that never advertised variants; behavior depends on OpenCode accepting those values at request time.

Overview
OpenCode’s composer now aligns with Codex/Grok by exposing a Reasoning control (renamed from “Variant”) alongside Agent.

Server (OpenCodeProvider): When a model reports no variants, capabilities now synthesizelow / medium / high / xhigh (default still via inferDefaultVariant). The dynamic descriptor label is Reasoning. Custom / fallbackDEFAULT_OPENCODE_MODEL_CAPABILITIES includes the same Reasoning and Agent selectors instead of an empty options list.

Web (TraitsPicker): Saved-option display maps variant → Reasoning when model metadata is unavailable.

Users should see effort in the traits chip (e.g. Medium · Build) and in the traits menu for models with or without advertised variants.

^{Reviewed by Cursor Bugbot for commit 6fe6ced. Bugbot is set up for automated code reviews on this repo. Configure here.}

[macroscopeapp]

Approvability

Verdict: Not approved

Macroscope's review found this PR not approvable — The PR changes OpenCode’s production capability metadata by introducing default Medium/Build selections and synthesizing reasoning levels for models without advertised variants. Those defaults can affect the composer state and the variant sent on subsequent OpenCode requests, so the change warrants human review.

Not approved because:

• Monthly spending limit reached (workspace setting). Approvability relies on correctness review in order to determine eligibility

^{Review your spending limits in Billing settings. You can add or adjust custom eligibility rules. Learn more.}

[maria-rcks]

Note

Written by muse-spark-1.2-contributor on behalf of Maria

OpenCode showed only the Agent chip (e.g. Build) while Codex/Grok/Cursor all expose a Reasoning control. Variant was labeled "Variant" and hidden when a model advertised no variants, leaving the composer with no effort selector.

Label the variant descriptor as "Reasoning" and synthesize standard levels (low/medium/high/xhigh, default medium) when a model reports no variants. Provide the same fallback for custom models and update the saved-option label for unavailable models. The composer now shows the reasoning effort next to the agent (e.g. "Medium · Build").

Before	After
	Reasoning menu (Reasoning / Agent groups)

Verified by inspecting provider capabilities and TraitsPicker logic; reasoning selector now renders for OpenCode models with and without advertised variants.

Model: muse-spark-1.2-contributor via opencode

---

Note

Medium Risk
Synthesized reasoning levels may be sent for models that never advertised variants; behavior depends on OpenCode accepting those values at request time.

Overview
OpenCode’s composer now aligns with Codex/Grok by exposing a Reasoning control (renamed from “Variant”) alongside Agent.

Server (OpenCodeProvider): When a model reports no variants, capabilities now synthesizelow / medium / high / xhigh (default still via inferDefaultVariant). The dynamic descriptor label is Reasoning. Custom / fallbackDEFAULT_OPENCODE_MODEL_CAPABILITIES includes the same Reasoning and Agent selectors instead of an empty options list.

Web (TraitsPicker): Saved-option display maps variant → Reasoning when model metadata is unavailable.

Users should see effort in the traits chip (e.g. Medium · Build) and in the traits menu for models with or without advertised variants.

^{Reviewed by Cursor Bugbot for commit 6fe6ced. Bugbot is set up for automated code reviews on this repo. Configure here.}

[macroscopeapp]

Approvability

Verdict: Not approved

Macroscope's review found this PR not approvable — The PR changes OpenCode’s production capability metadata by introducing default Medium/Build selections and synthesizing reasoning levels for models without advertised variants. Those defaults can affect the composer state and the variant sent on subsequent OpenCode requests, so the change warrants human review.

Not approved because:

• Monthly spending limit reached (workspace setting). Approvability relies on correctness review in order to determine eligibility

^{Review your spending limits in Billing settings. You can add or adjust custom eligibility rules. Learn more.}

[maria-rcks]

Note

Written by muse-spark-1.2-contributor on behalf of Maria

OpenCode showed only the Agent chip (e.g. Build) while Codex/Grok/Cursor all expose a Reasoning control. Variant was labeled "Variant" and hidden when a model advertised no variants, leaving the composer with no effort selector.

Label the variant descriptor as "Reasoning" and synthesize standard levels (low/medium/high/xhigh, default medium) when a model reports no variants. Provide the same fallback for custom models and update the saved-option label for unavailable models. The composer now shows the reasoning effort next to the agent (e.g. "Medium · Build").

Before	After
	Reasoning menu (Reasoning / Agent groups)

Verified by inspecting provider capabilities and TraitsPicker logic; reasoning selector now renders for OpenCode models with and without advertised variants.

Model: muse-spark-1.2-contributor via opencode

---

Note

Medium Risk
Synthesized reasoning levels may be sent for models that never advertised variants; behavior depends on OpenCode accepting those values at request time.

Overview
OpenCode’s composer now aligns with Codex/Grok by exposing a Reasoning control (renamed from “Variant”) alongside Agent.

Server (OpenCodeProvider): When a model reports no variants, capabilities now synthesizelow / medium / high / xhigh (default still via inferDefaultVariant). The dynamic descriptor label is Reasoning. Custom / fallbackDEFAULT_OPENCODE_MODEL_CAPABILITIES includes the same Reasoning and Agent selectors instead of an empty options list.

Web (TraitsPicker): Saved-option display maps variant → Reasoning when model metadata is unavailable.

Users should see effort in the traits chip (e.g. Medium · Build) and in the traits menu for models with or without advertised variants.

^{Reviewed by Cursor Bugbot for commit 6fe6ced. Bugbot is set up for automated code reviews on this repo. Configure here.}

[macroscopeapp]

Approvability

Verdict: Not approved

Macroscope's review found this PR not approvable — The PR changes OpenCode’s production capability metadata by introducing default Medium/Build selections and synthesizing reasoning levels for models without advertised variants. Those defaults can affect the composer state and the variant sent on subsequent OpenCode requests, so the change warrants human review.

Not approved because:

• Monthly spending limit reached (workspace setting). Approvability relies on correctness review in order to determine eligibility

^{Review your spending limits in Billing settings. You can add or adjust custom eligibility rules. Learn more.}

[maria-rcks]

Note

Written by muse-spark-1.2-contributor on behalf of Maria

OpenCode showed only the Agent chip (e.g. Build) while Codex/Grok/Cursor all expose a Reasoning control. Variant was labeled "Variant" and hidden when a model advertised no variants, leaving the composer with no effort selector.

Label the variant descriptor as "Reasoning" and synthesize standard levels (low/medium/high/xhigh, default medium) when a model reports no variants. Provide the same fallback for custom models and update the saved-option label for unavailable models. The composer now shows the reasoning effort next to the agent (e.g. "Medium · Build").

Before	After
	Reasoning menu (Reasoning / Agent groups)

Verified by inspecting provider capabilities and TraitsPicker logic; reasoning selector now renders for OpenCode models with and without advertised variants.

Model: muse-spark-1.2-contributor via opencode

---

Note

Medium Risk
Synthesized reasoning levels may be sent for models that never advertised variants; behavior depends on OpenCode accepting those values at request time.

Overview
OpenCode’s composer now aligns with Codex/Grok by exposing a Reasoning control (renamed from “Variant”) alongside Agent.

Server (OpenCodeProvider): When a model reports no variants, capabilities now synthesizelow / medium / high / xhigh (default still via inferDefaultVariant). The dynamic descriptor label is Reasoning. Custom / fallbackDEFAULT_OPENCODE_MODEL_CAPABILITIES includes the same Reasoning and Agent selectors instead of an empty options list.

Web (TraitsPicker): Saved-option display maps variant → Reasoning when model metadata is unavailable.

Users should see effort in the traits chip (e.g. Medium · Build) and in the traits menu for models with or without advertised variants.

^{Reviewed by Cursor Bugbot for commit 6fe6ced. Bugbot is set up for automated code reviews on this repo. Configure here.}

[macroscopeapp]

Approvability

Verdict: Not approved

Macroscope's review found this PR not approvable — The PR changes OpenCode’s production capability metadata by introducing default Medium/Build selections and synthesizing reasoning levels for models without advertised variants. Those defaults can affect the composer state and the variant sent on subsequent OpenCode requests, so the change warrants human review.

Not approved because:

• Monthly spending limit reached (workspace setting). Approvability relies on correctness review in order to determine eligibility

^{Review your spending limits in Billing settings. You can add or adjust custom eligibility rules. Learn more.}

[maria-rcks]

Note

Written by muse-spark-1.2-contributor on behalf of Maria

OpenCode showed only the Agent chip (e.g. Build) while Codex/Grok/Cursor all expose a Reasoning control. Variant was labeled "Variant" and hidden when a model advertised no variants, leaving the composer with no effort selector.

Label the variant descriptor as "Reasoning" and synthesize standard levels (low/medium/high/xhigh, default medium) when a model reports no variants. Provide the same fallback for custom models and update the saved-option label for unavailable models. The composer now shows the reasoning effort next to the agent (e.g. "Medium · Build").

Before	After
	Reasoning menu (Reasoning / Agent groups)

Verified by inspecting provider capabilities and TraitsPicker logic; reasoning selector now renders for OpenCode models with and without advertised variants.

Model: muse-spark-1.2-contributor via opencode

---

Note

Medium Risk
Synthesized reasoning levels may be sent for models that never advertised variants; behavior depends on OpenCode accepting those values at request time.

Overview
OpenCode’s composer now aligns with Codex/Grok by exposing a Reasoning control (renamed from “Variant”) alongside Agent.

Server (OpenCodeProvider): When a model reports no variants, capabilities now synthesizelow / medium / high / xhigh (default still via inferDefaultVariant). The dynamic descriptor label is Reasoning. Custom / fallbackDEFAULT_OPENCODE_MODEL_CAPABILITIES includes the same Reasoning and Agent selectors instead of an empty options list.

Web (TraitsPicker): Saved-option display maps variant → Reasoning when model metadata is unavailable.

Users should see effort in the traits chip (e.g. Medium · Build) and in the traits menu for models with or without advertised variants.

^{Reviewed by Cursor Bugbot for commit 6fe6ced. Bugbot is set up for automated code reviews on this repo. Configure here.}

[macroscopeapp]

Approvability

Verdict: Not approved

Macroscope's review found this PR not approvable — The PR changes OpenCode’s production capability metadata by introducing default Medium/Build selections and synthesizing reasoning levels for models without advertised variants. Those defaults can affect the composer state and the variant sent on subsequent OpenCode requests, so the change warrants human review.

Not approved because:

• Monthly spending limit reached (workspace setting). Approvability relies on correctness review in order to determine eligibility

^{Review your spending limits in Billing settings. You can add or adjust custom eligibility rules. Learn more.}