Add reasoningEffort to setModel/session.model.switchTo across all SDKs by patniko · Pull Request #712 · github/copilot-sdk

patniko · 2026-03-08T04:55:37Z

Summary

All four SDKs now support passing reasoningEffort when switching models mid-session via setModel(). The parameter is optional and fully backward-compatible.

API Changes

SDK	Signature
Node.js	`setModel(model, options?: { reasoningEffort? })`
Python	`set_model(model, *, reasoning_effort=None)`
Go	`SetModel(ctx, model, opts ...*SetModelOptions)`
.NET	`SetModelAsync(model, reasoningEffort?, cancellationToken?)`

Usage Examples

// Node.js
await session.setModel("claude-sonnet-4.6", { reasoningEffort: "high" });

# Python
await session.set_model("claude-sonnet-4.6", reasoning_effort="high")

// Go
session.SetModel(ctx, "claude-sonnet-4.6", &copilot.SetModelOptions{ReasoningEffort: "high"})

// .NET
await session.SetModelAsync("claude-sonnet-4.6", reasoningEffort: "high");

Files Changed

nodejs/src/session.ts + nodejs/src/generated/rpc.ts
python/copilot/session.py + python/copilot/generated/rpc.py
go/session.go + go/rpc/generated_rpc.go
dotnet/src/Session.cs + dotnet/src/Generated/Rpc.cs

Companion PR

Runtime changes: https://github.com/github/copilot-agent-runtime/pull/4408

Fixes #687

Copilot

Pull request overview

Adds support for passing an optional reasoningEffort when switching models mid-session (session.model.switchTo / setModel / set_model) across the Node.js, Python, Go, and .NET SDKs, aligning with the runtime capability introduced in the companion runtime PR.

Changes:

Extend each SDK’s session-level “switch model” API to accept optional reasoningEffort.
Update generated RPC parameter types in each language to serialize reasoningEffort as reasoningEffort on the wire.
Update inline docs/examples to demonstrate setting reasoningEffort during a model switch.

Reviewed changes

Copilot reviewed 5 out of 8 changed files in this pull request and generated 4 comments.

Show a summary per file

File	Description
python/copilot/session.py	Adds keyword-only `reasoning_effort` to `set_model()` and forwards it to RPC.
python/copilot/generated/rpc.py	Extends `SessionModelSwitchToParams` to include optional `reasoning_effort` serialization.
nodejs/src/session.ts	Extends `setModel()` to accept an options bag and forwards `reasoningEffort`.
nodejs/src/generated/rpc.ts	Extends `SessionModelSwitchToParams` to include optional `reasoningEffort` union type.
go/session.go	Adds `SetModelOptions` and forwards `ReasoningEffort` (as optional) to RPC.
go/rpc/generated_rpc.go	Extends `SessionModelSwitchToParams` and request building to include `reasoningEffort`.
dotnet/src/Session.cs	Adds optional `reasoningEffort` parameter and forwards it to RPC.
dotnet/src/Generated/Rpc.cs	Extends the generated request type and `SwitchToAsync` signature to include `reasoningEffort`.

Copilot · 2026-03-08T04:58:59Z

dotnet/src/Generated/Rpc.cs

+    public async Task<SessionModelSwitchToResult> SwitchToAsync(string modelId, string? reasoningEffort = null, CancellationToken cancellationToken = default)
    {
-        var request = new SessionModelSwitchToRequest { SessionId = _sessionId, ModelId = modelId };
+        var request = new SessionModelSwitchToRequest { SessionId = _sessionId, ModelId = modelId, ReasoningEffort = reasoningEffort };
        return await CopilotClient.InvokeRpcAsync<SessionModelSwitchToResult>(_rpc, "session.model.switchTo", [request], cancellationToken);
    }


ModelApi.SwitchToAsync now takes (string modelId, string? reasoningEffort = null, CancellationToken cancellationToken = default), which similarly breaks any existing callers that passed a CancellationToken as the second positional argument. Consider adding a backward-compatible overload SwitchToAsync(string modelId, CancellationToken cancellationToken = default) that forwards to the new method (or otherwise avoid inserting parameters before cancellationToken in public APIs).

Copilot · 2026-03-08T04:59:00Z

python/copilot/session.py

+            >>> await session.set_model("claude-sonnet-4.6", reasoning_effort="high")
        """
-        await self.rpc.model.switch_to(SessionModelSwitchToParams(model_id=model))
+        await self.rpc.model.switch_to(SessionModelSwitchToParams(model_id=model, reasoning_effort=reasoning_effort))


This call is long enough that it will be reformatted by ruff format (the repo enforces ruff format --check in CI with a 100 character line length). To avoid CI formatting diffs/failures, wrap the SessionModelSwitchToParams(...) construction and/or the switch_to(...) call across multiple lines in the style used elsewhere in this file.

Suggested change

await self.rpc.model.switch_to(SessionModelSwitchToParams(model_id=model, reasoning_effort=reasoning_effort))

await self.rpc.model.switch_to(

SessionModelSwitchToParams(

model_id=model,

reasoning_effort=reasoning_effort,

)

)

Copilot · 2026-03-08T04:59:00Z

go/session.go

+func (s *Session) SetModel(ctx context.Context, model string, opts ...*SetModelOptions) error {
+	params := &rpc.SessionModelSwitchToParams{ModelID: model}
+	if len(opts) > 0 && opts[0] != nil && opts[0].ReasoningEffort != "" {
+		re := opts[0].ReasoningEffort
+		params.ReasoningEffort = &re
+	}


SetModel accepts a variadic opts ...*SetModelOptions, but only opts[0] is ever read. This makes it easy to accidentally pass multiple option structs and have all but the first silently ignored. Consider either validating len(opts) <= 1 (and returning an error if more are provided) or changing the API shape to avoid a variadic when only a single options value is supported.

github-actions · 2026-03-08T05:13:49Z

✅ Cross-SDK Consistency Review

This PR demonstrates excellent cross-SDK consistency for adding reasoningEffort support to the model switching APIs. All four SDKs (Node.js, Python, Go, .NET) receive the same feature with appropriate language-specific patterns.

Consistency Highlights

✅ Feature parity: All SDKs now support reasoningEffort when switching models
✅ Naming conventions: Each language follows its idiomatic casing (camelCase, snake_case, PascalCase)
✅ Type safety: All SDKs use proper enums/unions for the four effort levels (low, medium, high, xhigh)
✅ Documentation: Examples updated across all SDKs showing the new parameter
✅ Backward compatibility: Most SDKs maintain full backward compatibility via optional parameters

API Signatures (all semantically equivalent)

SDK	Signature
Node.js	`setModel(model, options?: { reasoningEffort? })`
Python	`set_model(model, *, reasoning_effort=None)`
Go	`SetModel(ctx, model, opts *SetModelOptions)`
.NET	`SetModelAsync(model, reasoningEffort?, cancellationToken?)`

Note: Go Breaking Change

⚠️ The Go SDK signature changes from SetModel(ctx, model) to SetModel(ctx, model, opts), requiring existing callers to pass nil for the options parameter if they don't need reasoning effort. This is acceptable for Go's idioms, and the PR description clearly documents the new signature.

Additional Feature

This PR also consistently adds session.log() RPC support across all generated files, maintaining the same level of cross-SDK parity for this bonus feature.

Recommendation: Approve — this PR maintains excellent consistency across all SDK implementations. 🚀

Generated by SDK Consistency Review Agent for issue #712 · ◷

github-actions · 2026-03-08T05:19:28Z

✅ SDK Consistency Review: APPROVED with one minor observation

Great work on maintaining consistency across all four SDKs! This PR successfully adds the reasoningEffort parameter to the setModel/set_model/SetModel/SetModelAsync methods across all four language implementations.

✅ What's Consistent

Feature parity: All SDKs (Node.js, Python, Go, .NET) now support reasoningEffort
API naming: Follows proper language conventions:
- Node.js: reasoningEffort (camelCase)
- Python: reasoning_effort (snake_case)
- Go: ReasoningEffort (PascalCase)
- .NET: reasoningEffort parameter, ReasoningEffort enum (PascalCase)
Optional parameter: All implementations make it optional/nullable
Enum values: Consistent across all SDKs: "low" | "medium" | "high" | "xhigh"
Documentation: Examples added for all languages
Backward compatibility:
- Node.js/Python: Optional parameter (backward compatible)
- Go: Uses options pattern with nil default (requires signature change but idiomatic)
- .NET: Provides overload for backward compatibility
Tests updated: All breaking test calls updated to pass nil/null for the new parameter

📋 Minor Observation

I noticed that the generated RPC files include a new session.log RPC method in Node.js, Python, and Go, but this doesn't appear in the .NET generated file. This is likely expected if the .NET code generation hasn't been run yet, or if there's a difference in what schema version each SDK is using for codegen.

Recommendation: If session.log is intended as a new feature, ensure all SDKs regenerate from the same schema version. If it's not part of this PR's scope, you can ignore this note.

Verdict: The reasoningEffort feature is excellently consistent across all four SDKs. No action required for the core feature of this PR. 🎉

Generated by SDK Consistency Review Agent for issue #712 · ◷

github-actions · 2026-03-08T05:25:52Z

✅ Cross-SDK Consistency Review: PASSED

I've reviewed this PR for consistency across all four SDK implementations (Node.js, Python, Go, and .NET). The changes maintain excellent cross-SDK consistency.

Summary

All four SDKs now support the reasoningEffort parameter when switching models via setModel/set_model/SetModel/SetModelAsync, with:

✅ Full feature parity across all languages
✅ Appropriate language-specific idioms (optional param in TS, keyword-only in Python, options struct in Go, nullable enum in .NET)
✅ Consistent valid values: "low", "medium", "high", "xhigh"
✅ Backward compatibility maintained in all implementations
✅ Updated documentation and examples in all SDKs
✅ Test coverage added/updated appropriately

Implementation Details

SDK	Signature	Pattern
Node.js	`setModel(model, options?: { reasoningEffort? })`	Optional parameter object
Python	`set_model(model, *, reasoning_effort=None)`	Keyword-only argument
Go	`SetModel(ctx, model, opts *SetModelOptions)`	Options struct (idiomatic Go)
.NET	`SetModelAsync(model, reasoningEffort?, cancellationToken?)`	Nullable enum + overload

Each approach follows the established conventions for that language while maintaining semantic equivalence.

No consistency issues found. 🎉

Generated by SDK Consistency Review Agent for issue #712 · ◷

All four SDKs now support passing reasoningEffort when switching models mid-session via setModel(). The parameter is optional and backward-compatible. - Node.js: setModel(model, { reasoningEffort? }) - Python: set_model(model, *, reasoning_effort=None) - Go: SetModel(ctx, model, opts ...*SetModelOptions) - .NET: SetModelAsync(model, reasoningEffort?, cancellationToken?) Fixes #687 Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>

patniko · 2026-03-08T05:39:54Z

@SteveSandersonMS need to get #4408 in and released first.

IeuanWalker · 2026-03-08T14:57:44Z

dotnet/src/Generated/Rpc.cs


    /// <summary>Calls "session.model.switchTo".</summary>
-    public async Task<SessionModelSwitchToResult> SwitchToAsync(string modelId, CancellationToken cancellationToken = default)
+    public async Task<SessionModelSwitchToResult> SwitchToAsync(string modelId, SessionModelSwitchToRequestReasoningEffort? reasoningEffort, CancellationToken cancellationToken = default)


Just a note, the reasoning effort is currently a string? everywhere else (getting list of models, creating/ resuming session, etc), so feels a bit weird that its an enum here.

As it would require us to parse the string? that the SDK returns into this enum to change it mid session

Wouldnt be an issue if it was an enum throughout SDK

Copilot AI review requested due to automatic review settings March 8, 2026 04:55

patniko requested a review from a team as a code owner March 8, 2026 04:55

Copilot started reviewing on behalf of patniko March 8, 2026 04:56 View session

Copilot AI reviewed Mar 8, 2026

View reviewed changes

patniko force-pushed the fix/setmodel-reasoning-effort-and-byok branch from fbffd30 to 2910573 Compare March 8, 2026 05:10

patniko force-pushed the fix/setmodel-reasoning-effort-and-byok branch from 2910573 to f35bf5c Compare March 8, 2026 05:15

github deleted a comment from Copilot AI Mar 8, 2026

patniko force-pushed the fix/setmodel-reasoning-effort-and-byok branch from f35bf5c to f6ee586 Compare March 8, 2026 05:22

patniko force-pushed the fix/setmodel-reasoning-effort-and-byok branch from f6ee586 to 518256d Compare March 8, 2026 05:37

IeuanWalker reviewed Mar 8, 2026

View reviewed changes

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

Add reasoningEffort to setModel/session.model.switchTo across all SDKs#712

Add reasoningEffort to setModel/session.model.switchTo across all SDKs#712
patniko wants to merge 1 commit intomainfrom
fix/setmodel-reasoning-effort-and-byok

patniko commented Mar 8, 2026

Uh oh!

Copilot AI left a comment

Uh oh!

Copilot AI Mar 8, 2026

Uh oh!

Copilot AI Mar 8, 2026

Uh oh!

Copilot AI Mar 8, 2026

Uh oh!

github-actions bot commented Mar 8, 2026

Uh oh!

github-actions bot commented Mar 8, 2026

Uh oh!

github-actions bot commented Mar 8, 2026

Uh oh!

patniko commented Mar 8, 2026

Uh oh!

IeuanWalker Mar 8, 2026 •

edited

Loading

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

3 participants

Conversation

patniko commented Mar 8, 2026

Summary

API Changes

Usage Examples

Files Changed

Companion PR

Uh oh!

Copilot AI left a comment

Choose a reason for hiding this comment

Pull request overview

Reviewed changes

Uh oh!

Copilot AI Mar 8, 2026

Choose a reason for hiding this comment

Uh oh!

Copilot AI Mar 8, 2026

Choose a reason for hiding this comment

Uh oh!

Copilot AI Mar 8, 2026

Choose a reason for hiding this comment

Uh oh!

github-actions bot commented Mar 8, 2026

✅ Cross-SDK Consistency Review

Consistency Highlights

API Signatures (all semantically equivalent)

Note: Go Breaking Change

Additional Feature

Uh oh!

github-actions bot commented Mar 8, 2026

✅ SDK Consistency Review: APPROVED with one minor observation

✅ What's Consistent

📋 Minor Observation

Uh oh!

github-actions bot commented Mar 8, 2026

✅ Cross-SDK Consistency Review: PASSED

Summary

Implementation Details

Uh oh!

patniko commented Mar 8, 2026

Uh oh!

IeuanWalker Mar 8, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

3 participants

IeuanWalker Mar 8, 2026 •

edited

Loading